chore(release): 0.32.0 (#456)

Closes #451

UI flicker (section 3 of #451) is already tracked in #412.

1. expandAtMentions treated `@<dir>` as a `not-file` skip — the model
   never saw what was in the directory. It now resolves the path, walks
   the project root with the existing gitignore layers, filters to
   entries under the directory, and inlines a `<directory>` block
   capped at DEFAULT_AT_DIR_MAX_ENTRIES (200). `@docs/` and `@docs`
   resolve to the same key, so trailing-slash typing doesn't double-fire.

2. listFilesWithStatsSync / listFilesWithStatsAsync called only
   ent.isFile(), which returns false for symlinks. Symlinked source
   files were silently dropped from the picker. Both walkers now stat
   through symlinks; symlinks-to-files come back, symlinks-to-dirs
   stay dropped (cycle risk), broken links stay dropped (nothing to
   point at).

.github/ISSUE_TEMPLATE/bug_report.md

@@ -4,6 +4,10 @@ about: Something is broken
 labels: bug
 ---
 
+> **Screen flicker, garbled output, leftover artifacts, cursor jumping?**
+> Use the **Display / rendering issue** template instead — it asks for the
+> terminal-specific info we need to diagnose those.
+
 **What happened**
 A clear and concise description.
 
@@ -14,10 +18,13 @@ What you expected to happen.
 Steps or minimal code that reproduces it.
 
 **Environment**
-- Reasonix version:
-- Node version:
-- OS:
-- DeepSeek model (e.g. `deepseek-chat`, `deepseek-reasoner`):
+- Reasonix version (`reasonix --version`):
+- Node version (`node --version`):
+- OS (Windows 11 / macOS 14 / Ubuntu 24.04 / …):
+- Shell (bash, zsh, fish, PowerShell 7, PowerShell 5.1, cmd, …):
+- Terminal app (Windows Terminal, iTerm2, Alacritty, kitty, WezTerm, **VSCode integrated**, **Cursor integrated**, Hyper, …):
+- DeepSeek model (e.g. `deepseek-v4-flash`, `deepseek-v4-pro`):
 
 **Logs / transcript**
-If using the CLI, attach the relevant chunk of `--transcript`.
+If using the CLI, attach the relevant chunk of `--transcript`, or run
+`reasonix doctor` and paste the output.

.github/ISSUE_TEMPLATE/display_issue.md

@@ -0,0 +1,109 @@
+---
+name: Display / rendering issue
+about: Screen flicker, garbled output, leftover artifacts, cursor jumping
+labels: bug, rendering
+---
+
+> Display problems almost always come from the **terminal emulator**, not
+> the shell. Please fill the terminal section carefully — `bash vs PowerShell`
+> tells us very little; `VSCode integrated terminal vs Windows Terminal`
+> tells us everything.
+
+**Symptom** (tick all that apply)
+- [ ] Whole screen flickers / flashes during streaming response
+- [ ] Lines tear or only half-redraw
+- [ ] Stale output left behind after a frame updates
+- [ ] Cursor jumps to wrong column or vanishes
+- [ ] Mojibake / wrong-width characters (e.g. `□`, half-width emoji)
+- [ ] Other (describe below)
+
+**When it happens**
+- [ ] During assistant streaming (token-by-token output)
+- [ ] When tool cards expand / collapse
+- [ ] During scroll-up / scrollback
+- [ ] On terminal resize
+- [ ] On launch / on quit
+- [ ] Other (describe below)
+
+**Terminal — the important part**
+
+Where exactly are you running `reasonix`?
+
+- [ ] **VSCode** integrated terminal — VSCode version: `?`
+- [ ] **Cursor** integrated terminal — Cursor version: `?`
+- [ ] **Windows Terminal** — version: `?`
+- [ ] **cmd.exe** (legacy console host)
+- [ ] **PowerShell ISE** (note: ISE doesn't support ANSI — most things will look broken)
+- [ ] **iTerm2** / **Terminal.app** / **Alacritty** / **kitty** / **WezTerm** / **Hyper**
+- [ ] tmux / screen / mosh — and inside which outer terminal? `?`
+- [ ] Other:
+
+> 💡 **How to find your VSCode / Cursor version**
+> `Help → About` (Windows/Linux) or `Code → About Visual Studio Code` (macOS).
+> Paste the whole panel — version + commit + Electron + xterm.js if shown.
+
+**Diagnostic dump — copy/paste output**
+
+Run **one** of the snippets below in the same terminal where you saw the
+issue, and paste the output here:
+
+<details><summary>PowerShell (Windows)</summary>
+
+```powershell
+reasonix --version; node --version
+$PSVersionTable.PSVersion.ToString()
+[System.Environment]::OSVersion.VersionString
+"TERM=$env:TERM"
+"TERM_PROGRAM=$env:TERM_PROGRAM"
+"TERM_PROGRAM_VERSION=$env:TERM_PROGRAM_VERSION"
+"COLORTERM=$env:COLORTERM"
+"WT_SESSION=$env:WT_SESSION"
+"VSCODE_INJECTION=$env:VSCODE_INJECTION"
+"WSL_DISTRO_NAME=$env:WSL_DISTRO_NAME"
+```
+
+</details>
+
+<details><summary>bash / zsh (macOS / Linux / WSL / Git Bash)</summary>
+
+```bash
+reasonix --version; node --version
+uname -a
+echo "TERM=$TERM"
+echo "TERM_PROGRAM=$TERM_PROGRAM"
+echo "TERM_PROGRAM_VERSION=$TERM_PROGRAM_VERSION"
+echo "COLORTERM=$COLORTERM"
+echo "WT_SESSION=$WT_SESSION"
+echo "VSCODE_INJECTION=$VSCODE_INJECTION"
+echo "WSL_DISTRO_NAME=$WSL_DISTRO_NAME"
+```
+
+</details>
+
+```
+<paste output here>
+```
+
+**VSCode / Cursor users only — terminal settings**
+
+Open Settings (`Ctrl+,`), search `terminal.integrated.gpuAcceleration`,
+report current value: `auto` / `on` / `canvas` / `off` — `?`
+
+Already tried any of:
+- [ ] Switching `gpuAcceleration` to a different value
+- [ ] Detaching the terminal (drag tab into its own window)
+- [ ] Running the same command in a non-VSCode terminal — did it still flicker? `yes / no`
+
+**Reproduction**
+
+Steps that reliably trigger it (commands run, files edited, was a tool
+streaming a long response, was the window being resized, …):
+
+1.
+2.
+3.
+
+**Screen recording (strongly preferred)**
+
+A 5–10s GIF or MP4 is worth 1000 words for rendering bugs. Drop it in
+this comment box — GitHub uploads attachments inline.

.github/workflows/ci.yml

@@ -48,6 +48,3 @@ jobs:
       # DEEPSEEK_API_KEY in CI.
       - name: τ-bench harness dry-run
         run: npx tsx benchmarks/tau-bench/runner.ts --dry --out /tmp/tau-dry.json
-
-      - name: Harvest-bench harness dry-run
-        run: npx tsx benchmarks/harvest/runner.ts --dry --out /tmp/harvest-dry.json

.gitignore

@@ -33,4 +33,6 @@ sessions/
 .reasonix/semantic-skip
 # Scratch entry regenerated each time scripts/bundle-codemirror.mjs runs.
 scripts/.cm-entry.mjs
+# Personal bun lockfile — project uses npm officially.
+bun.lock
 

CHANGELOG.md

@@ -3,6 +3,750 @@
 All notable changes to Reasonix. The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/);
 this project uses [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.32.0] — 2026-05-08
+
+**Headline:** the slash surface lost weight. Eleven redundant
+commands gone (`/clear`, `/keys`, `/models`, `/effort`, `/rename`,
+`/forget`, `/think`, `/tool`, `/apply-plan`, `/semantic`,
+`/resume`), the unified preset+model picker replaces three
+near-identical commands, and the two heaviest features almost
+nobody opted into — `/harvest` (Pillar-2 plan-state extraction)
+and `/branch` (parallel-sample selector) — are deleted along with
+their backing modules, events, transcript fields, and CLI flags.
+The four-pillar architecture collapses to three. The slash registry
+now carries a `group` tag (chat / setup / info / session / extend
+/ code / jobs / advanced) and bare-`/` suggestions render those
+groups with advanced rows hidden behind a `+ N advanced · type to
+search` footer. A new `~/.reasonix/slash-usage.json` counter
+sorts frequent commands first within a prefix.
+
+The other half of the release is plan-mode UX. PlanLiveRow had
+nothing to dock — a code path that should have materialized an
+"active" plan card on approval was missing, so the bottom strip
+stayed empty after `/plan`. Fixed. And the per-step "Checkpoint —
+step done" picker fired in auto/yolo too, defeating the whole
+point of those modes; auto/yolo now resolve "continue" without
+prompting while still creating per-step rollback snapshots so
+`/restore` granularity stays intact. Plus a long-standing
+`@`-mention bug: typing `@docs/` produced an empty `not-file`
+placeholder. It now expands to a recursive `<directory>` listing
+respecting the project's gitignore, and symlinked source files
+finally appear in the `@`-picker.
+
+**Features:**
+
+- feat(semantic): OpenAI-compatible embedding provider. Configure
+  custom API URL / key / model / request body for embeddings,
+  replacing the Ollama-only setup. Dashboard semantic panel adds
+  a provider dropdown with "OpenAI-Compatible" alongside Ollama,
+  clearer status messages, and detailed indexing-job phases
+  (scanning / embedding / writing). Community contribution from
+  @kabaka9527. (#424)
+
+- feat(slash): unified preset+model picker. `ModelPicker` shows the
+  three presets at the top with cost/headline copy and the model
+  catalog below; cursor lands on the active row (auto-detects which
+  preset matches the loop's current model + effort + autoEscalate).
+  Both `/preset` (no arg) and `/model` (no arg) open it. (#453)
+
+- feat(slash): grouped suggestions + usage telemetry.
+  `SlashCommandSpec` gains a `group` field; suggestion palette
+  renders section headers on bare `/` with advanced rows hidden
+  behind a footer. New `~/.reasonix/slash-usage.json` counter
+  (read-modify-write, atomic rename) feeds `suggestSlashCommands`
+  so frequent commands sort first; `slash.invoked` events emit to
+  events.jsonl for cross-session analysis. `/help` walks the same
+  grouped registry so there's one source of truth. (#453)
+
+**Bug fixes:**
+
+- fix(plan): dock active plan card. `case "plan_proposed"` had
+  been dropping the gate payload's `steps`/`summary`, and the
+  approve path never dispatched `plan.show` — so no card with
+  `variant: "active"` ever existed and `isActivePlanInFlight`
+  returned null. PlanLiveRow now docks correctly after approval,
+  and the dock tracks tail rewrites on revise-accept. (#454)
+
+- fix(plan): auto/yolo skip the per-step checkpoint picker. The
+  "Checkpoint — step done" picker fired after every
+  `mark_step_complete` regardless of edit mode — shell and
+  edit-gate already self-skip in auto/yolo, but plan checkpoints
+  kept stopping the model. The gate handler now checks
+  `editModeRef` and resolves "continue" without UI; per-step
+  rollback snapshot still runs so `/restore` granularity is
+  preserved. `review` mode is unchanged. (#454)
+
+- fix(at-mentions): `@<dir>` expands to a recursive listing.
+  Was treated as a `not-file` skip, leaving the model with an
+  empty placeholder. Walks the project root with the existing
+  gitignore layers, filters to entries under the directory, and
+  inlines a `<directory path="..." entries="N">` block capped at
+  `DEFAULT_AT_DIR_MAX_ENTRIES` (200). `@docs/` and `@docs`
+  resolve identically. (#455, closes #451)
+
+- fix(at-mentions): symlinks-to-files appear in the `@`-picker.
+  `Dirent.isFile()` returns false for symlinks, so symlinked
+  source files never showed up in completions. Both
+  `listFilesWithStatsSync` and `listFilesWithStatsAsync` now stat
+  through symlinks; symlinks-to-files come back, symlinks-to-dirs
+  stay dropped (cycle risk), broken links stay dropped (nothing
+  to point at). (#455, closes #451)
+
+**Removals — slash commands:**
+
+- `/clear` (merged into `/new` as alias — was the most common
+  source of "what's the difference?" confusion)
+- `/models` (picker covers it)
+- `/keys` (folded into `/help`)
+- `/resume` (sessions picker has switch action)
+- `/semantic` (folded into `/doctor`)
+- `/effort` (preset locks effort)
+- `/rename` and `/forget` (sessions picker actions)
+- `/apply-plan` (plan picker handles the fallback path)
+- `/think` and `/tool` (debug-only; events.jsonl records both)
+- `/mcp browse` entry (handler still routes `["browse"]`)
+
+**Removals — features:**
+
+- `/harvest` (Pillar-2 plan-state extraction): `src/harvest.ts`,
+  `--harvest` CLI flag, `harvestedTurns` transcript field.
+- `/branch` (parallel-sample selector): `src/consistency.ts`,
+  `src/loop/branch.ts`, `BranchCard`,
+  `branch_start/progress/done` events, `--branch` CLI flag.
+- `benchmarks/harvest/` deleted; `ARCHITECTURE.md` collapses from
+  four pillars to three; README + zh-CN + `dashboard/PARITY.md`
+  updated.
+
+**Removals — public API:**
+
+- `src/index.ts` drops `harvest`, `runBranches`,
+  `aggregateBranchUsage`, `defaultSelector`, `emptyPlanState`,
+  `isPlanStateEmpty`, and the `TypedPlanState`, `HarvestOptions`,
+  `BranchSample`, `BranchSummary`, `BranchProgress`,
+  `BranchOptions`, `BranchResult`, `BranchSelector` types.
+  Consumers depending on these break intentionally — they were
+  experimental from the start and never met the cache-first
+  cost target this project gates on.
+
+## [0.31.0] — 2026-05-08
+
+**Headline:** a Mac user reported a DeepSeek 503 day where Reasonix
+showed a wall of raw `DeepSeek 503: <html>...` and they couldn't tell
+if our agent had crashed or the upstream API was down. Two threads of
+work fell out of that single bug: a friendly outage notice with a
+1.5s reachability probe to `/user/balance` (so we can say "DS main
+API answered, but /chat/completions is failing — their problem") and
+a full sweep of every hardcoded English string a Chinese user could
+hit. ~150 strings across 8 files moved into the `loop.*` / `errors.*`
+/ `app.*` / `hooks.*` / `summary.*` / `wizard.*` namespaces with
+zh-CN translations. The setup wizard now opens with a language picker
+defaulting to `detectSystemLanguage()` — for the case where
+`Intl.DateTimeFormat().resolvedOptions().locale` returns the wrong
+locale and a user shouldn't have to discover `/language` after
+finishing setup in English.
+
+The other half of the release is dashboard parity work — picker
+modals (sessions / checkpoint / MCP marketplace), viewer modal for
+`/replay`, plus cockpit / budget gauge / model picker / loop control
+panel / `/pro` one-shot — closing buckets B-E of the #369 web-parity
+tracker.
+
+**Features:**
+
+- feat(loop): friendly DeepSeek 5xx error with reachability probe.
+  When the chat endpoint returns 5xx (after retry.ts has already
+  retried 4× with backoff), `formatLoopError` now spawns a 1.5s
+  `/user/balance` probe and renders one of three messages: no probe
+  (generic outage notice), reachable (main API up but /chat dying),
+  unreachable (DS or your network is down). All three say "this is
+  a DeepSeek-side problem, not Reasonix" and link
+  https://status.deepseek.com. Removes the misleading file header in
+  `loop/errors.ts` that claimed retry.ts swallowed all 5xx — it
+  doesn't, and never did. (#440)
+- feat(wizard): first-launch language picker + full i18n. New
+  `language` step before `apiKey`, cursor defaults to
+  `detectSystemLanguage()` marked `(detected)`. Selection saves
+  immediately so all later wizard screens render in the chosen
+  language. Re-running `reasonix setup` opens at the same step with
+  the cursor on the saved language so Enter is a no-op. The wizard's
+  ~30 hardcoded strings (welcome, prompts, validation errors, MCP
+  catalog hints, review labels, save errors, saved screen) all moved
+  to a new `wizard.*` namespace with zh-CN. (#442)
+- feat(dashboard): picker modal protocol for web parity. New
+  `picker` modal kind drives sessions, checkpoint, and MCP
+  marketplace pickers from the same protocol. Closes the gap where
+  TUI-only modals stayed inaccessible from `/dashboard`. (#417,
+  #418, #419, #420)
+- feat(dashboard): viewer modal kind for `/replay`. Loads an archived
+  plan into a read-only time-travel snapshot, mirrors the TUI replay
+  experience. (#421)
+- feat(dashboard): cockpit tile + budget gauge + 14-day cost trend.
+  At-a-glance current-session telemetry on the overview panel. (#431)
+- feat(dashboard): editable model picker in settings + `/pro`
+  one-shot panel + loop control (start / stop / countdown). The
+  settings tab is now the single place to flip model preset, arm
+  `/pro` for the next turn, or start an autonomous loop without
+  switching to the TUI. (#430, #432, #433)
+- feat(dashboard): server surface for `/pro` / `/budget` / `/model`
+  / `/loop`. POST endpoints under `/api/cockpit/*` carry the
+  mutations the panels above need. (#429)
+
+**i18n sweep:**
+
+- i18n(loop/errors): localize DeepSeek error messages — context
+  overflow, 401/402/422/400, 5xx (with the new reachability probe
+  variants), reason prefixes for budget/aborted/context-guard/stuck.
+  20 keys + zh-CN. (#444)
+- i18n(loop): 14 user-facing yields in `step()` — budget exhausted /
+  80% warning, /pro armed, aborted-at-iter, tool-budget warning,
+  preflight fold/no-fold, flash + auto escalation, storm-broken,
+  history compaction (regular + aggressive), forcing-summary. The
+  `loop.*` namespace had 7 dead keys defined but never wired —
+  removed and replaced with 20 that match the actual yield shapes.
+  (#445)
+- i18n(hooks/summary): hook outcome formatter (`hook PreToolUse/Bash
+  \`cmd\` block (output truncated at 256KB)`) and the force-summary
+  status / hallucinated-fallback / failed-fallback strings now go
+  through `t()`. New `hooks.*` and `summary.*` namespaces. (#446)
+- i18n(app): ~26 hardcoded strings in `App.tsx` plus seven existing
+  `ui.*` keys that had been declared but never called (same dead-key
+  pattern as `loop.*`). New `app.*` namespace covers walk modal,
+  edit-mode cycle (review/auto/yolo), edit gate, dashboard stopped,
+  hash-memory note, bash-mode failures, hook header rows,
+  @mentions / @url, shell confirm, checkpoint saved, plan
+  continue / stop / revise. (#447)
+- i18n(slash): four lagging slash handlers — `web-search-engine.ts`
+  was 0% localized, plus `mcp.ts` / `plans.ts` / `semantic.ts` had
+  small gaps. ~22 new keys. (#448)
+- i18n(dashboard): translate the plan `idle` status pill — the
+  `active` / `done` pills already used `t()` but the third branch
+  was hardcoded English. (#443)
+
+**Bug fixes:**
+
+- fix(search): honor abort during recursive fs scans — Esc during a
+  large `search` tool call now exits promptly instead of finishing
+  the walk. (#400)
+- fix(ui): refresh model badge on dashboard preset change and /pro
+  turns — the header pill stayed stale across server-side
+  preset/pro switches. (#403)
+- fix(permissions): match Windows project keys case-insensitively —
+  the project allowlist hashed `C:\Foo` differently from `c:\foo`,
+  causing entries to "disappear" depending on which case the cwd
+  carried. (#402)
+- fix(prompt): inline short single-line pastes verbatim — the long-
+  paste collapser was firing on tiny one-liners and burying them
+  behind a "(N chars pasted)" placeholder. (#397)
+
+**Tests / refactor / docs:**
+
+- test(mcp): cover startup summary states (#396)
+- chore: improve loop.ts tests (#271)
+- refactor(ui): quiet chat-screen chrome — fewer always-on rows on
+  the welcome card so the prompt stays close to the top. (#411)
+- docs(readme): canonical install + subcommand cheatsheet (#408)
+- docs(issues): split off display/rendering template, collect
+  terminal host info inline. (#412)
+- docs(dashboard): PARITY.md audit — bucket E of #369. (#439)
+
+## [0.30.5] — 2026-05-07
+
+**Headline:** three contributor-led follow-ups from the #350 RFC plus
+the #366 onboarding piece. The repeat-loop storm guard now exempts
+obviously-safe inspector tools (`read_file`, `list_directory`,
+`job_output`, `list_jobs`) so a model intentionally re-reading state
+isn't flagged as stuck. A new `wait_for_job(jobId, timeoutMs?)` tool
+replaces N-iteration polling loops with a single blocking call —
+returns the moment the job exits or emits new output. And `/skill
+new <name>` finally provides the missing creation entry-point for
+user skills, scaffolding a stub with the right frontmatter so
+first-time users don't have to read the source to author a skill.
+
+**Features:**
+
+- feat(storm): add `stormExempt` flag on `ToolDefinition`, set on
+  `read_file`, `list_directory`, `job_output`, `list_jobs`. Cheap
+  state-inspection no longer trips the repeat-loop guard. Mutating
+  tools and unknown tools still go through the existing window-and-
+  threshold check. (#350, PR #388 by @ctharvey)
+- feat(jobs): new `wait_for_job(jobId, timeoutMs?)` shell tool —
+  blocks until the job exits or emits new output, bounded by
+  `timeoutMs` (default 5000, clamped to 0..30000). Returns
+  `{ exited, exitCode, latestOutput }`; `latestOutput` is the
+  delta since the call started, not the full buffer. Rides the
+  existing job registry's exit + output events; one call replaces
+  N polling iterations and is token-cheaper than the prior
+  re-call-job_output loop. (#350, PR #390 by @ctharvey)
+- feat(skills): `/skill new <name>` scaffolds a stub at
+  `<project>/.reasonix/skills/<name>.md` with minimal frontmatter
+  + a comment block listing the optional knobs (`runAs`,
+  `allowed-tools`, `model`). `/skill new <name> --global` writes
+  under `~/.reasonix/skills/` for cross-project use; auto-falls-
+  back to global when there's no project root. The empty
+  `/skill list` now ends with an explicit "no remote registry yet
+  — scaffold one with `/skill new <name>`" line so users don't
+  hunt for a marketplace that doesn't exist. (#366, PR #394)
+
+**Bug fixes:**
+
+- fix(skills): atomic create with `wx` flag — close the TOCTOU
+  race between `existsSync(...)` and `writeFileSync(...)` that
+  CodeQL flagged. The existence check IS the atomic write now;
+  `EEXIST` from a parallel writer surfaces as the same "skill
+  already exists" error instead of silently overwriting. (PR #394)
+
+## [0.30.4] — 2026-05-07
+
+**Headline:** sweep of the user-reported bug + onboarding queue from
+the 0.30.2 / 0.30.3 launch day. Resume now restores the full session
+state (cache hit %, cost, last context bar — previously they all
+showed zero on a fresh boot until the first turn landed). The model
+pill on assistant cards reflects the model that actually answered
+after `/model` or `/preset` switches it. Bare `/model` opens an
+interactive picker — typed-id entry stays for power users.
+PowerShell users get Shift+Tab back via three additional encodings
+(modifier-encoded back-tab, modifyOtherKeys, Kitty keyboard). And a
+class of "junk text after exit" on Linux/fish (terminal-feature
+replies leaking into the parent shell) gets a defensive stdin drain
+in the exit path.
+
+`--dir` is now discoverable for beginners — surfaced in the welcome
+banner, the `/status` panel, the filesystem sandbox-escape error,
+and a Getting Started callout in both READMEs.
+
+**Bug fixes:**
+
+- fix(stats): persist cache totals + `lastPromptTokens` across
+  resume. `SessionMeta` only carried `totalCostUsd` / `turnCount`,
+  so on every resume `/status` showed 0 context + 0% cache hit until
+  the first turn actually fired (even though the prefix was already
+  cached, costing $0.01 per turn). Three new fields are persisted
+  per-turn and seeded into `SessionStats` on resume; the existing
+  carryover plumbing now covers cache + last context.
+  (#364, PR #384)
+- fix(ui): `/model <id>` and `/preset {auto,flash,pro}` now update
+  the active model in the agent store so the next assistant card
+  pill reflects the new selection. Previously `state.session.model`
+  was set once in `initialState()` and never mutated, so the pill
+  showed the launch-time model regardless of what actually answered
+  the turn. New `session.model.change` event; cards already opened
+  keep their captured model so mid-turn auto-escalation doesn't
+  retroactively relabel. (#372, PR #385)
+- fix(input): recognize three additional Shift+Tab encodings for
+  PowerShell hosts and modern terminals — `\x1b[1;2Z` (modifier-
+  encoded back-tab some PowerShell hosts emit), `\x1b[27;2;9~`
+  (modifyOtherKeys level 2, which we already enable on startup),
+  `\x1b[9;2u` (Kitty keyboard envelope). Without these the edit-
+  mode cycle was silently dropped on PowerShell.
+  `/mode` typed fallback continues to work. (#373, PR #386)
+- fix(tty): drain pending feature-detection replies on exit. Linux
+  reporters saw `^[]11;rgb:...^[\^[[33;1R^[[?62;1;4c` printed by
+  fish / bash after exiting reasonix — those bytes are responses to
+  OSC 11 / CPR / DA1 queries the runtime emits during startup that
+  sit in stdin's queue until exit. New `drainTtyResponses(50ms)`
+  reads-and-discards anything queued before control returns to the
+  parent shell. Layered on top of 0.30.3's alt-screen mitigation
+  (`--no-alt-screen` users get the fix too). (#365, PR #391)
+
+**Features:**
+
+- feat(ui): bare `/model` opens an interactive model picker — arrow-
+  key list, current model marked, `[r]` refreshes the catalog, esc
+  cancels. Seeds from the live DeepSeek catalog
+  (`useSessionInfo.listModels()`); falls back to the four known
+  DeepSeek ids when the catalog hasn't loaded yet so the picker
+  isn't empty on first open. The current id is always included even
+  when the API didn't return it. `/model <id>` typed entry stays
+  for power users. (#371, PR #387)
+- feat(ui): surface `--dir` / pinned workspace for first-time users.
+  WelcomeBanner shows the workspace + relaunch hint in code mode;
+  `/status` adds a `workspace <path> · pinned at launch` line; the
+  filesystem sandbox-escape error points at `reasonix code --dir
+  <path>` instead of just dropping a raw error; both READMEs gain a
+  Getting Started subsection on `--dir`. No new slash command —
+  mid-session retargeting is intentionally not supported (the
+  message log + memory paths get tangled with stale roots).
+  (#370, PR #389)
+
+## [0.30.3] — 2026-05-07
+
+**Headline:** the chat scroll rewrite lands. Ink 5.2 → 7.0.2 / React
+18.3 → 19.2, the cell-diff renderer is retired, and `reasonix code` /
+`reasonix chat` default to alt-screen with row-precision virtual
+scroll. PgUp / PgDn / mouse wheel scroll history; an empty prompt + ↑
+also scrolls (Ctrl+P / Ctrl+N still recalls prompt history). When
+scrolled away from bottom, the prompt hides and a `📖 reading
+history — End / PgDn to return` hint appears. Resize-ghost dividers
+and `<Static>`-related scroll-yank artifacts are gone with the
+renderer that produced them. `--no-alt-screen` keeps the legacy
+in-shell-scrollback behavior.
+
+`web_search` gains a configurable backend — Mojeek stays the default,
+but `/web-search-engine searxng <url>` switches to a self-hosted
+SearXNG instance for users whose network blocks Mojeek. And the MCP
+filesystem sandbox now fails with an actionable
+`mkdir -p '<path>'` hint instead of a raw Node stack when the
+configured directory doesn't exist; the wizard offers to create it
+inline at config time.
+
+**Features:**
+
+- feat(ui): row-precision virtual scroll on Ink 7 + React 19.
+  `<Static>` retired (incompatible with alt-screen reflow);
+  `React.memo(CardRenderer)` plus reference-stable cards in the
+  reducer skip the reconciler on unchanged history. `useChatScroll`
+  drives an outer `overflow=hidden` clip + inner `marginTop=-N`
+  slide; `useBoxMetrics` reports inner / outer heights so bounds
+  clamp and auto-pin to bottom on new content. `App` owns
+  PgUp/PgDn/End/wheel; PromptInput hands off ↑/↓ on empty buffer
+  when pinned + idle. Ticker migrated to Ink 7's shared
+  `useAnimation`. (PR #380)
+- feat(web): configurable `web_search` backend with SearXNG support.
+  `/web-search-engine` shows / switches the active engine; URL is
+  persisted to `~/.reasonix/config.json`. Mojeek remains the default;
+  the original Mojeek path is preserved as `searchMojeek()`. Protocol
+  auto-normalizes (`localhost:8080` → `http://...`); an unreachable
+  SearXNG endpoint surfaces an install hint instead of a raw fetch
+  error. (PR #338)
+
+**Bug fixes:**
+
+- fix(mcp): preflight the filesystem sandbox directory before
+  spawning `@modelcontextprotocol/server-filesystem`. Missing
+  directories now throw `MCP filesystem sandbox '<path>' does not
+  exist — create it with: mkdir -p '<path>'` instead of a raw Node
+  stack from inside `npx`'s child. The init wizard adds an inline
+  `[Y] create it (mkdir -p) / [N] enter a different path` confirm
+  step when the user types a path that doesn't exist, so bad config
+  never reaches disk. Spawn-time path deliberately does not
+  auto-mkdir — by then the user may not remember writing the
+  config. (#362, PR #379)
+- fix(readme): website URLs corrected from `/reasonix/` to
+  `/DeepSeek-Reasonix/`. (PR #375)
+
+**Chores:**
+
+- chore(issue-template): bug template now asks for shell + terminal,
+  and the model-id examples track the current DeepSeek model
+  lineup. (PR #378)
+
+## [0.30.2] — 2026-05-07
+
+**Headline:** five user-visible polish items from the @dacec354 triage
+batch. The streaming reply now carries a live `42 t/s` throughput pill
+(plus a `1.2k tok · 42 t/s` summary on settled), and `ctrl-o` toggles a
+full-tail view so a long plan / todo can be read while it's still
+being written. The auto-mode undo banner gains a `space`-to-pause
+keybind for users who want a beat to think before the 5-second window
+expires. SessionPicker and the dashboard's session-cost displays both
+respect the user's wallet currency now — USD wallets see `$0.05`, CNY
+wallets see `¥0.36` end-to-end. And a long-standing scrollback bug
+that left the "reasoning…" spinner spinning forever after reasoning
+ended is fixed.
+
+**Features:**
+
+- feat(ui): live `42 t/s` pill on the streaming reply card; settled
+  card shows `1.2k tok · 42 t/s` summary. Computed via the bundled
+  DeepSeek tokenizer; gated below 4 tokens / 500 ms so the first
+  chunk doesn't print bogus rates. Re-renders ride the slow tick so
+  the rate keeps updating during chunk silence. (#334, PR #356)
+- feat(ui): `space` toggles pause / resume on the auto-mode 5-second
+  undo countdown. While paused the bar freezes at the captured
+  fraction, the badge swaps to `Ns · paused`, and pressing `space`
+  again resumes from where it stopped. The `u` and `space` keybinds
+  share the same modal-and-prompt-empty gating. (#337, PR #356)
+- feat(ui): `ctrl-o` toggles "expanded" mode on the live streaming
+  card. Expanded shows up to 60 visual lines (capped so the card
+  can't swallow the whole viewport) plus a `⋯ N earlier lines above`
+  hint when content overflows. Auto-resets to collapsed at turn end.
+  A `expanded ⌃o` / `preview ⌃o` pill in the card header advertises
+  the keybind. (#335, #337, PR #359)
+
+**Bug fixes:**
+
+- fix(ui): `splitCardStream` only treated the LAST card as live,
+  committing every earlier card to Ink's `<Static>`. When the model
+  streamed reasoning then content (or kicked off a tool card), the
+  reasoning card was no longer last — it got frozen into `<Static>`
+  while still `streaming: true`. `<Static>` doesn't re-render frozen
+  items, so when `reasoning.end` later set `streaming: false`, the
+  spinner kept spinning forever. The split now scans for the first
+  unsettled card and keeps everything from that index onward live;
+  a card only commits to `<Static>` once it's settled AND every
+  earlier card is too. (PR #358)
+- fix(ui): SessionPicker hardcoded `¥` and ran USD → CNY itself, so
+  USD-wallet users saw `¥X.XX` in the session list. `SessionMeta`
+  gains `balanceCurrency`; App.tsx writes the live wallet currency
+  alongside `totalCostUsd` on each turn save. Picker accepts a
+  `walletCurrency` prop and falls back to each row's stored
+  currency. Cost rendering routes through the shared `formatCost()`
+  helper. (#312, PR #357)
+- fix(dashboard): cost displays were hardcoded to `$` via `fmtUsd()`,
+  so a CNY-wallet user saw `session $0.5190` in the dashboard while
+  the same session read `¥0.024` in the CLI — both the symbol AND
+  the magnitude diverged because no conversion happened. Dashboard
+  now has its own `fmtCost(usd, currency)` mirroring the CLI's
+  conversion (CNY × 7.2). Overview current-session cost, cost-trend
+  day average, and the chat panel rail / status-bar costs all
+  thread the wallet currency from the cockpit balance. Claude-
+  equivalent comparisons in `usage.ts` stay USD by design — Claude's
+  API is USD-priced regardless of the user's wallet. (PR #360)
+
+## [0.30.1] — 2026-05-07
+
+**Headline:** two TUI ghost-rendering fixes for issues that only showed
+up on the published binary. The CLI bundle now uses real Ink in
+production instead of the cell-diff renderer that source mode never
+exercised, eliminating a whole class of bugs invisible to `npx tsx`
+repros. The `submit_plan` approval picker no longer leaves a
+duplicated row behind when arrow-navigating choices — the live tool
+card above the modal is suppressed while the picker owns the screen.
+
+**Bug fixes:**
+
+- fix(renderer): drop the `tsup` `ink → ink-compat` alias and the
+  `noExternal` for `ink` / `ink-text-input`. The CLI bundle keeps
+  `from "ink"` external; `ink` and `ink-text-input` move to runtime
+  `dependencies` so npm install pulls the real package. The
+  cell-diff renderer is no longer on the user-facing path; it's
+  retained only for direct test imports. Same behavior as `npx tsx
+  src/cli/index.ts` mode — TUI bug repros from source mode are now
+  valid for the published binary again. (#346, PR #354)
+- fix(ui): `CardStream` accepts a `suppressLive` flag; `App.tsx`
+  computes a `modalOpen` flag from the union of pending modal states
+  and passes it through. While any picker / confirm modal owns the
+  screen, the unsettled live tool card above it stops repainting,
+  removing the rerender competition that left stale rows during
+  arrow-key navigation. (#352, PR #353 — thanks @ctharvey)
+
+## [0.30.0] — 2026-05-06
+
+**Headline:** slash commands grow first-class aliases, and the
+cell-diff renderer hardens column targeting against per-cell width
+miscounts. `/quit` and `/q` now resolve to `/exit` from a single
+declaration on the spec instead of ad-hoc handler mirrors; `/?` →
+`/help`, `/reset` → `/new`, `/lang` → `/language` follow the same
+path. The renderer's `moveTo()` now uses CHA absolute (`\x1b[N+1G`)
+for column targeting instead of CUF relative (`\x1b[NC`), making the
+diff stream immune to the cursor-drift class of bug Anthropic
+documented in `claude-code#14208`.
+
+**Features:**
+
+- feat(slash): `aliases?: readonly string[]` on `SlashCommandSpec`.
+  Adding a new alias is now a one-line edit to the canonical command
+  — dispatch, autocomplete, arg-context resolution, and the
+  dashboard `/api/slash` response all route through one
+  `resolveSlashAlias()` map built from `SLASH_COMMANDS` at module
+  init. Suggestion rows display aliases dimly (` · /quit /q`) so
+  they stay discoverable without doubling the autocomplete list.
+  Removes the per-handler alias mirrors that used to live in
+  `handlers/basic.ts` and `handlers/language.ts`. (#332, PR #347)
+
+**Bug fixes:**
+
+- fix(renderer): switch the X-axis branch of `moveTo()` from CUF
+  relative (`\x1b[NC`) to CHA absolute (`\x1b[N+1G`). Y-axis stays
+  on CUU/CUD since we don't track absolute terminal rows. Relative
+  column moves accumulate drift across frames whenever an earlier
+  write miscounts cell width — `▸` (U+25B8) rendered 2-cell on
+  fonts with East Asian fallback, ambiguous-width chars on
+  terminals that font-detect width, OSC8 hyperlinks parsed as
+  visible chars, etc. The next CUF lands at the wrong column,
+  ghost rows leak into adjacent hint lines, and the modal "shifts"
+  as users navigate. CHA targets the absolute column regardless of
+  what the terminal thinks — immune to the desync chain. Same fix
+  Anthropic shipped in claude-code per their issue #14208
+  post-mortem. (#346, PR #348)
+
+## [0.29.1] — 2026-05-06
+
+**Headline:** four user-reported bugs from the 0.29.0 release window.
+The markdown renderer no longer turns English abbreviations like
+`e.g.` into broken hyperlinks (which on cmd.exe / non-OSC-8 terminals
+showed up as visible `]8;;file://e.g…` garbage and on the cell-diff
+side desynced the renderer's prev-frame model). The cell-diff
+renderer now defensively trail-clears any row whose content shrank
+between frames. Resumed sessions keep their cumulative session cost
+instead of resetting to `$0`. The Approve plan modal now shows the
+plan body inline when the model didn't supply structured steps.
+Wide markdown tables fall back to row-grouped key/value lines
+instead of the previous column-grouped output.
+
+**Bug fixes:**
+
+- fix(markdown): stop linkifying English abbreviations + drop OSC 8
+  escape emission. The `FILE_REF_RE` extension class was too loose
+  (`{1,6}`), so `e.g`, `i.e`, `a.m` matched as file paths; `osc8()`
+  baked OSC 8 escape bytes into Text content, which the cell-diff
+  renderer's wrapLine stripped of zero-width chars but kept the
+  printable body — producing visible `]8;;file://e.ge.g]8;;` garbage
+  on every terminal. Tightened the regex (now requires path-shape,
+  line-number suffix, or extension >= 2 chars) and removed the OSC 8
+  escape — file refs still stand out via color + underline. (#330,
+  PR #341)
+- fix(renderer): trail-clear rows that shrank between frames in the
+  cell-diff diff. The diff skipped cells where prev and next were
+  byte-equal (including trailing EMPTY cells), so any earlier ANSI
+  desync left stale chars in shrunken rows — manifested as the
+  shell-confirm modal showing `allow always` + `mand, ask again next
+  time` after Up/Down navigation. New `clearToEOL` patch type and a
+  per-row sweep after `diffEach`. (#330, PR #341)
+- fix(stats): carry session cost / turn count across resume. The
+  TUI's `$X session` figure reset to `$0` on every resume even
+  though the disk meta still held the cumulative `totalCostUsd`.
+  `SessionStats` gains `seedCarryover()`; `CacheFirstLoop` reads the
+  meta on resume and seeds the carryover when prior messages exist.
+  (#333, PR #342)
+- fix(plan): show the plan body in the Approve plan modal. When the
+  model called `submit_plan` with a markdown body but no structured
+  `steps`, the modal showed only the choice list — users had no way
+  to see what they were approving without scrolling back. The modal
+  now renders the body via `MarkdownView`, capped at 24 lines with
+  an overflow hint. (#336, PR #343)
+- fix(markdown): row-group the table fallback layout. When a table
+  was too wide for the viewport, the fallback flattened it as N
+  "Component:" lines, then N "What:" lines, then N "Manual TCs:"
+  lines — the reader couldn't tell which value belonged to which
+  row. Swapped to row-first iteration with a blank separator
+  between rows. (#340, PR #344)
+
+## [0.29.0] — 2026-05-06
+
+**Headline:** tool dispatch is no longer strictly serial. When the model
+emits multiple `parallelSafe`-annotated tool calls in one turn (multiple
+`read_file`, multiple `spawn_subagent`, etc.), the loop now races them
+together via `Promise.allSettled`; a non-`parallelSafe` call ends the
+chunk and runs alone, so read-after-write ordering still holds. Tool
+yields and history append still land in declared order regardless of
+which call settles first — the model and UI see the same shape they
+would under serial dispatch. The TUI's `SubagentRow` becomes
+`SubagentLiveStack`, rendering 1 → rich card, 2..max → compact rows,
+> max → "+N more running…" fold. Closes umbrella #325.
+
+**Tool dispatch:**
+
+- feat(tools): `ToolDefinition.parallelSafe?: boolean` — opt-in
+  annotation, default `false`. `ToolRegistry.isParallelSafe(name)` for
+  the dispatcher to query; unknown / unannotated tools resolve to
+  `false` so third-party MCP tools must explicitly opt in. Built-in
+  read-only filesystem (`read_file`, `list_directory`,
+  `directory_tree`, `search_files`, `search_content`,
+  `get_file_info`), web (`web_search`, `web_fetch`), `recall_memory`,
+  `semantic_search`, isolated child loops (`run_skill`,
+  `spawn_subagent`), and in-memory job queries (`job_output`,
+  `list_jobs`) are annotated. Mutating tools stay default. (PR #326)
+- feat(loop): chunked parallel tool dispatch. Replaces `for...of +
+  await` in the dispatch loop with a chunking loop that groups
+  consecutive `parallelSafe` calls and races them; unsafe calls form
+  serial barriers. `runOneToolCall` extracts per-call lifecycle
+  (PreToolUse + dispatch + PostToolUse) so the chunk can fan out via
+  `Promise.allSettled` while the loop body keeps yielding events in
+  declared order. Two new env knobs: `REASONIX_PARALLEL_MAX` (chunk
+  size cap, default 3, hard max 16) and `REASONIX_TOOL_DISPATCH=serial`
+  (escape hatch). Tests cover parallel timing, serial barrier on mixed
+  safe/unsafe, declared-order yields under racey completion, and both
+  env-knob overrides. (PR #327)
+
+**TUI:**
+
+- feat(ui): `SubagentEvent` carries a stable `runId` per spawn so the
+  sink can key concurrent runs apart instead of overwriting one shared
+  row. `useSubagent` keeps an array of in-flight activities;
+  `SubagentLiveStack` renders 1 → rich card (unchanged), 2..max →
+  compact rows with per-row spinner + iter + last tool, > max →
+  compact rows + "+N more running…" fold. (PR #327)
+
+**Docs:**
+
+- docs(architecture): `docs/ARCHITECTURE.md` Pillar 1 gains a
+  "Parallel tool dispatch" section explaining the chunking rule, both
+  env knobs, and the list of built-in tools that opt in. (PR #328)
+
+## [0.28.0] — 2026-05-06
+
+**Headline:** subagent capability sharpened on three axes — skills can
+now scope a child to a specific tool subset via `allowed-tools`
+frontmatter, callers can request a per-spawn iter budget via the new
+`max_iters` arg (clamped 1-32), and two built-in personas (`explore`,
+`verify`) are selectable inline via a `type` arg without writing a
+skill. Closes umbrella #316.
+
+**Subagent:**
+
+- feat(subagent): honor skill `allowed-tools` frontmatter when forking
+  the child registry. The field was parsed but ignored ("Unused in v1");
+  now it scopes the subagent to the named tools only. New
+  `forkRegistryWithAllowList` helper alongside `forkRegistryExcluding`;
+  `NEVER_INHERITED` (`spawn_subagent` / `submit_plan`) still wins so
+  depth=1 + plan-mode guarantees hold even if a skill names them. An
+  allow-list naming a tool the parent doesn't have returns a structured
+  error result (no API call burned). (#317, PR #320)
+- feat(subagent): expose `max_iters` on the `spawn_subagent` tool
+  schema. Clamped to 1-32 at the boundary; floats round down; non-numeric
+  / missing falls back to the registration-time default (still 16).
+  Verify-style tasks can ask for 6-8, explore-style can ask for 24+.
+  (#318, PR #321)
+- feat(subagent): two built-in personas selectable via `type` arg —
+  `explore` (wide-net read-only investigation, 20-iter budget) and
+  `verify` (narrow yes/no with evidence, 8-iter budget). Caller's
+  explicit `system` / `max_iters` override the type's defaults. Prompts
+  live in new `src/tools/subagent-types.ts` so `subagent.ts` stays
+  under the 500-line target. (#319, PR #322)
+
+## [0.27.3] — 2026-05-06
+
+**Headline:** USD-account users now see `$` instead of `¥` everywhere
+money is shown in the TUI — wallet balance, turn cost, session cost,
+top-bar cost label, subagent end-event cost suffix, and the UsageCard
+header / body / wallet line. Pre-fix a USD wallet rendered
+`¥0.0352 turn · ¥0.461 session · wallet ¥0.91`; now it renders
+`$0.0308 turn · $0.064 session · wallet $0.91`. The display follows
+the wallet currency reported by the DeepSeek API (`currency: "USD"|"CNY"`),
+not the UI language — a CNY account on an English UI still sees `¥`,
+and vice versa. Originally reported in #278 by @Explosion-Scratch.
+
+**UI / currency:**
+
+- fix(ui): USD wallets render `$` for wallet balance, turn cost, and
+  session cost. State + event schemas now carry `balanceCurrency`
+  through `App.tsx → reducer → StatusBar` so every render site sees
+  the wallet symbol the API reported. Originally drafted by @wviana
+  in #272; the TUI plumbing through state.ts / cards.ts / events.ts /
+  reducer.ts / useScrollback.ts / slash/types.ts was the bulk of the
+  fix.
+- fix(ui): balance color threshold checks USD against the CNY scale
+  (`$0.91 ≈ ¥6.55`) rather than treating `0.91` as `< ¥5 → red`. USD
+  wallets now correctly show yellow at low-but-not-empty balances.
+- fix(ui): `StatsPanel.ChromeRow` cost label and `useSubagent`
+  end-event cost suffix follow the wallet currency too — pre-fix
+  these always rendered `$`. (#313)
+- refactor(ui): seven currency helpers in `theme/tokens.ts`
+  (`formatCNY` / `formatBalance` / `formatBalanceLabel` /
+  `formatWalletDisplay` / `formatCost` / `balanceColorCny` /
+  `balanceColorForBalance`) collapsed to three: `formatBalance`,
+  `formatCost`, `balanceColor`. Undefined currency defaults to CNY
+  (matches pre-fix unconditional `¥`) so the transient first-turn
+  case where balance arrived but currency hasn't is consistent.
+- chore(ui): remove orphan `ChromeBar.tsx` (258 lines). `App.tsx`
+  mounts `StatsPanel`'s diverged `ChromeRow`, which is the bar users
+  actually see. The two formatter helpers ChromeBar once owned now
+  live in `theme/tokens.ts`. (#314)
+
+**Loop:**
+
+- refactor(loop): `loop.ts` 1331 → 1219 (−112). Three sibling files
+  under `src/loop/`: `messages.ts` (pure ChatMessage builders),
+  `turn-failure-tracker.ts` (per-turn failure count + threshold
+  tipping), `force-summary.ts` (forced-summary generator behind a
+  small DI context). Continues the #308 / #309 cadence — small
+  per-helper extractions, no behavior change. (#311)
+
+**Known follow-up:** `SessionPicker` still hardcodes `¥` for
+per-session cost in the session-history list, tracked in #312
+(good-first-issue).
+
 ## [0.25.1] — 2026-05-05
 
 **Headline:** `run_command` learns the four common shell chain

README.md

@@ -7,7 +7,7 @@
    · 
   <a href="./README.zh-CN.md">简体中文</a>
   &nbsp;·&nbsp;
-  <a href="https://esengine.github.io/reasonix/">Website</a>
+  <a href="https://esengine.github.io/DeepSeek-Reasonix/">Website</a>
   &nbsp;·&nbsp;
   <a href="./docs/ARCHITECTURE.md">Architecture</a>
   &nbsp;·&nbsp;
@@ -41,6 +41,52 @@
 > [!TIP]
 > **Cache stability isn't a feature you turn on; it's an invariant the loop is designed around.** That's the whole reason Reasonix is DeepSeek-only — every layer is tuned to the byte-stable prefix-cache mechanic.
 
+> [!NOTE]
+> **Real user, single day (2026-05-01):** 435M input tokens, **99.82% cache hit**, ~$12 instead of the ~$61 the same workload would cost with no cache on `v4-flash` — see the [case study](./benchmarks/real-world-cache/README.md). DeepSeek provides the cacheable bytes; the four mechanisms in [Pillar 1](./docs/ARCHITECTURE.md#pillar-1--cache-first-loop) are how Reasonix keeps them cacheable across long sessions.
+
+<br/>
+
+## Web search
+
+Reasonix includes `web_search` and `web_fetch` tools. By default it uses **Mojeek** (no setup required). You can switch to a **self-hosted SearXNG** instance — a metasearch engine that aggregates whatever upstream engines your instance is configured for.
+
+### Switching engines (persists to disk)
+
+The `/search-engine` slash command (alias `/se`) writes your choice to `~/.reasonix/config.json` immediately — it survives restarts:
+
+```
+/search-engine mojeek              # default, no external deps
+/search-engine searxng             # SearXNG at http://localhost:8080
+/search-engine searxng http://192.168.1.100:8888  # custom endpoint
+```
+
+Equivalent `~/.reasonix/config.json`:
+
+```json
+{
+  "webSearchEngine": "searxng",
+  "webSearchEndpoint": "http://localhost:8080"
+}
+```
+
+The tool picks up the change on the next call — no restart needed.
+
+### Starting SearXNG
+
+```sh
+podman run -d --replace --name searxng -p 8080:8080 docker.io/searxng/searxng
+# or: docker run -d -p 8080:8080 searxng/searxng
+```
+
+Verify it's running:
+
+```sh
+curl http://localhost:8080/search?q=test
+# → HTML search results page
+```
+
+> **Note:** The endpoint must include the protocol (`http://`). `localhost:8080` alone will fail — the tool will show a clear error telling you to install SearXNG if the server is unreachable.
+
 <br/>
 
 ## Install
@@ -52,19 +98,46 @@ npx reasonix code   # paste a DeepSeek API key on first run; persists after
 
 Requires Node ≥ 22. Tested on macOS · Linux · Windows (PowerShell · Git Bash · Windows Terminal). Get a [DeepSeek API key →](https://platform.deepseek.com/api_keys) · `reasonix code --help` for flags.
 
+`npx` is the recommended path — no global install, always picks up the latest version. If you'll use Reasonix daily and want `reasonix` on your `PATH`, run `reasonix update` once and it'll do the `npm install -g` for you.
+
+### Subcommand cheatsheet
+
+| Command | When to use |
+|---|---|
+| `reasonix code [dir]` | Coding agent rooted at a project. **Start here.** |
+| `reasonix chat` | Plain chat — no filesystem tools, just a conversation with persisted history. |
+| `reasonix run "task"` | One-shot, streams the answer to stdout. Good for shell pipes. |
+| `reasonix doctor` | Environment health check (Node version, API key, MCP wiring). |
+| `reasonix update` | Upgrade Reasonix itself. |
+
+Other subcommands (`replay` · `diff` · `events` · `stats` · `index` · `mcp` · `prune-sessions`) are listed in `reasonix --help` and on the [CLI reference](https://esengine.github.io/DeepSeek-Reasonix/#cli).
+
+**Working in a different folder:** Reasonix scopes filesystem tools to the launch directory. To work elsewhere, pass `--dir`:
+
+```bash
+npx reasonix code --dir /path/to/project   # or use a relative path
+```
+
+Mid-session switching isn't supported by design (the message log + memory paths get tangled with stale roots). Quit and relaunch with a new `--dir` to retarget. `/status` always shows the current pinned workspace.
+
+**Author your first skill:** Skills are markdown playbooks the model can invoke (`/skill <name>`). There's no remote registry yet — you author them directly:
+
+```bash
+/skill new my-skill          # scaffolds <project>/.reasonix/skills/my-skill.md
+/skill new my-skill --global # or under ~/.reasonix/skills for cross-project use
+```
+
+Edit the file (`description:` frontmatter + body), then `/skill list` to see it. Add `runAs: subagent` to the frontmatter to spawn an isolated subagent loop instead of inlining the body.
+
 <br/>
 
 ## What makes Reasonix different
 
-The loop is organized around four pillars. Each one solves a problem generic agent frameworks don't even see — because they were designed for a different cache mechanic.
-
-<p align="center">
-  <a href="./docs/ARCHITECTURE.md"><img src="docs/assets/pillars.svg" alt="Reasonix four pillars — Cache-first loop, R1 thought harvesting, Tool-call repair, Cost control" width="880"/></a>
-</p>
+The loop is organized around three pillars. Each one solves a problem generic agent frameworks don't even see — because they were designed for a different cache mechanic.
 
 <sub align="center">
 
-Click any card to read the full architecture writeup → [Pillar 1](./docs/ARCHITECTURE.md#pillar-1--cache-first-loop) · [Pillar 2](./docs/ARCHITECTURE.md#pillar-2--r1-thought-harvesting-opt-in) · [Pillar 3](./docs/ARCHITECTURE.md#pillar-3--tool-call-repair) · [Pillar 4](./docs/ARCHITECTURE.md#pillar-4--cost-control-v06)
+Click through to the full architecture writeup → [Pillar 1 — Cache-first loop](./docs/ARCHITECTURE.md#pillar-1--cache-first-loop) · [Pillar 2 — Tool-call repair](./docs/ARCHITECTURE.md#pillar-2--tool-call-repair) · [Pillar 3 — Cost control](./docs/ARCHITECTURE.md#pillar-3--cost-control-v06)
 
 </sub>
 
@@ -87,8 +160,10 @@ Click any card to read the full architecture writeup → [Pillar 1](./docs/ARCHI
 | Cost profile                      | **low per task** | premium           | subscription + use  | varies             |
 | DeepSeek prefix-cache             | **engineered**   | not applicable    | not applicable      | incidental         |
 | Embedded web dashboard            | yes              | —                 | n/a (IDE)           | —                  |
+| Configurable web search engine    | `/search-engine` | —             | —                   | —                  |
 | Persistent per-workspace sessions | yes              | partial           | n/a                 | —                  |
 | Plan mode · MCP · hooks · skills  | yes              | yes               | yes                 | partial            |
+| Web search (Mojeek + SearXNG)      | yes              | yes               | yes                 | yes                |
 | Open community development        | yes              | —                 | —                   | yes                |
 
 For live cache-hit rates, costs, and methodology, see [`benchmarks/`](./benchmarks/) — the numbers move with model pricing, so they live with the harness, not in the README.
@@ -97,9 +172,9 @@ For live cache-hit rates, costs, and methodology, see [`benchmarks/`](./benchmar
 
 ## Documentation
 
-- [**Architecture**](./docs/ARCHITECTURE.md) — the four pillars, cache-first loop, harvesting, scaffolds
+- [**Architecture**](./docs/ARCHITECTURE.md) — three pillars: cache-first loop, tool-call repair, cost control
 - [**Benchmarks**](./benchmarks/) — τ-bench-lite harness, transcripts, cost methodology
-- [**Website**](https://esengine.github.io/reasonix/) — getting started, dashboard mockup, TUI mockup
+- [**Website**](https://esengine.github.io/DeepSeek-Reasonix/) — getting started, dashboard mockup, TUI mockup
 - [**Contributing**](./CONTRIBUTING.md) — comment policy, error-handling rules, library-over-hand-rolled
 - [**Code of Conduct**](./CODE_OF_CONDUCT.md) · [**Security policy**](./SECURITY.md)
 
@@ -115,7 +190,7 @@ Scoped starter tickets — each with background, code pointers, acceptance crite
 **Open Discussions — opinions wanted:**
 
 - [#20 · CLI / TUI design](https://github.com/esengine/reasonix/discussions/20) — what's broken, what's missing, what would you change?
-- [#21 · Dashboard design](https://github.com/esengine/reasonix/discussions/21) — react against the [proposed mockup](https://esengine.github.io/reasonix/design/agent-dashboard.html)
+- [#21 · Dashboard design](https://github.com/esengine/reasonix/discussions/21) — react against the [proposed mockup](https://esengine.github.io/DeepSeek-Reasonix/design/agent-dashboard.html)
 - [#22 · Future feature wishlist](https://github.com/esengine/reasonix/discussions/22) — what would you build into Reasonix next?
 
 **Already using Reasonix and willing to help others discover it?** Publish blog posts, articles, screenshots, talks, or videos to [**Show and tell**](https://github.com/esengine/reasonix/discussions/categories/show-and-tell). The project has no marketing budget — community word of mouth is how new users find it. Sustained advocates earn the badge below, displayed next to the contributors wall once awarded:

README.zh-CN.md

@@ -7,7 +7,7 @@
    · 
   <strong>简体中文</strong>
   &nbsp;·&nbsp;
-  <a href="https://esengine.github.io/reasonix/">官方网站</a>
+  <a href="https://esengine.github.io/DeepSeek-Reasonix/">官方网站</a>
   &nbsp;·&nbsp;
   <a href="./docs/ARCHITECTURE.md">架构文档</a>
   &nbsp;·&nbsp;
@@ -52,19 +52,46 @@ npx reasonix code   # 首次运行粘贴 DeepSeek API Key，之后会记住
 
 要求 Node ≥ 22。已在 macOS · Linux · Windows（PowerShell · Git Bash · Windows Terminal）测过。[去拿 DeepSeek API Key →](https://platform.deepseek.com/api_keys) · 完整 flag 看 `reasonix code --help`。
 
+`npx` 是推荐路径 —— 不用全局安装，每次都拿到最新版本。如果你天天用、想把 `reasonix` 装到 `PATH` 上，跑一次 `reasonix update` 就行，它会替你跑 `npm install -g`。
+
+### 子命令速查
+
+| 命令 | 适用场景 |
+|---|---|
+| `reasonix code [dir]` | 锁在某个项目根目录的编码 agent。**先用这个。** |
+| `reasonix chat` | 纯聊天 —— 不挂文件系统工具，只是带历史的对话。 |
+| `reasonix run "task"` | 一次性，把答案直接流到 stdout。适合 shell 管道。 |
+| `reasonix doctor` | 环境体检（Node 版本、API Key、MCP 接线）。 |
+| `reasonix update` | 升级 Reasonix 本身。 |
+
+其他子命令（`replay` · `diff` · `events` · `stats` · `index` · `mcp` · `prune-sessions`）见 `reasonix --help` 和 [CLI 参考](https://esengine.github.io/DeepSeek-Reasonix/#cli)。
+
+**在其他目录工作：** Reasonix 把文件系统工具作用域绑定在启动目录。要在别的目录工作，传 `--dir`：
+
+```bash
+npx reasonix code --dir /path/to/project   # 也可以用相对路径
+```
+
+中途切换工作区是有意不支持的（消息日志和 memory 路径会和旧的根目录混在一起，状态错乱）。退出后用新的 `--dir` 重新启动来切换。`/status` 始终显示当前锁定的工作区。
+
+**写第一个 Skill：** Skills 是模型可以调用的 markdown 剧本（`/skill <name>`）。暂无在线市场 —— 自己写：
+
+```bash
+/skill new my-skill          # 在 <project>/.reasonix/skills/my-skill.md 生成模板
+/skill new my-skill --global # 或者放到 ~/.reasonix/skills，跨项目共用
+```
+
+编辑文件（`description:` frontmatter + 正文），然后 `/skill list` 就能看到。frontmatter 里加 `runAs: subagent` 会以独立 subagent 跑，而不是把正文内联进父 prompt。
+
 <br/>
 
 ## Reasonix 的不同之处
 
-整个循环围绕四根支柱组织。每一根解决的都是通用 agent 框架根本看不见的问题 —— 因为它们是为另一种缓存机制设计的。
-
-<p align="center">
-  <a href="./docs/ARCHITECTURE.md"><img src="docs/assets/pillars.zh-CN.svg" alt="Reasonix 四大支柱 — 缓存优先循环、R1 思维提取、工具调用修复、成本控制" width="880"/></a>
-</p>
+整个循环围绕三根支柱组织。每一根解决的都是通用 agent 框架根本看不见的问题 —— 因为它们是为另一种缓存机制设计的。
 
 <sub align="center">
 
-各支柱完整说明 → [Pillar 1](./docs/ARCHITECTURE.md#pillar-1--cache-first-loop) · [Pillar 2](./docs/ARCHITECTURE.md#pillar-2--r1-thought-harvesting-opt-in) · [Pillar 3](./docs/ARCHITECTURE.md#pillar-3--tool-call-repair) · [Pillar 4](./docs/ARCHITECTURE.md#pillar-4--cost-control-v06)
+各支柱完整说明 → [Pillar 1 — 缓存优先循环](./docs/ARCHITECTURE.md#pillar-1--cache-first-loop) · [Pillar 2 — 工具调用修复](./docs/ARCHITECTURE.md#pillar-2--tool-call-repair) · [Pillar 3 — 成本控制](./docs/ARCHITECTURE.md#pillar-3--cost-control-v06)
 
 </sub>
 
@@ -99,7 +126,7 @@ npx reasonix code   # 首次运行粘贴 DeepSeek API Key，之后会记住
 
 - [**架构**](./docs/ARCHITECTURE.md) —— 四大支柱、缓存优先循环、思维提取、脚手架
 - [**基准测试**](./benchmarks/) —— τ-bench-lite harness、transcript、成本方法论
-- [**官方网站**](https://esengine.github.io/reasonix/) —— 入门、Dashboard 设计稿、TUI 设计稿
+- [**官方网站**](https://esengine.github.io/DeepSeek-Reasonix/) —— 入门、Dashboard 设计稿、TUI 设计稿
 - [**贡献指南**](./CONTRIBUTING.md) —— 注释规则、错误处理、用现成库不手写
 - [**行为准则**](./CODE_OF_CONDUCT.md) · [**安全策略**](./SECURITY.md)
 
@@ -115,7 +142,7 @@ npx reasonix code   # 首次运行粘贴 DeepSeek API Key，之后会记住
 **正在征集意见的 Discussions：**
 
 - [#20 · CLI / TUI 设计](https://github.com/esengine/reasonix/discussions/20) —— 哪里坏了、哪里少东西、哪里你会怎么改？
-- [#21 · Dashboard 设计](https://github.com/esengine/reasonix/discussions/21) —— 对着[设计稿](https://esengine.github.io/reasonix/design/agent-dashboard.html)拍砖
+- [#21 · Dashboard 设计](https://github.com/esengine/reasonix/discussions/21) —— 对着[设计稿](https://esengine.github.io/DeepSeek-Reasonix/design/agent-dashboard.html)拍砖
 - [#22 · 未来功能愿望单](https://github.com/esengine/reasonix/discussions/22) —— 你希望 Reasonix 长出什么功能？
 
 **正在使用 Reasonix，愿意让更多人了解它？** 欢迎将相关博客、文章、截图、演讲或视频发布到 [**Show and tell**](https://github.com/esengine/reasonix/discussions/categories/show-and-tell)。项目没有营销预算，新用户主要通过社区口碑找到这里。持续参与传播的用户将获得下方这枚徽章，颁发后会展示在贡献者墙旁：

benchmarks/harvest/README.md

@@ -1,106 +0,0 @@
-# Harvest eval harness (Pillar 2)
-
-Measures Pillar 2 (R1 Thought Harvesting) the same way `tau-bench/`
-measures Pillar 1 (Cache-First Loop): isolate one variable, produce
-numbers anyone can reproduce.
-
-## What's different from τ-bench-lite?
-
-Deliberately different task shape:
-
-| | τ-bench-lite | harvest-bench |
-|---|---|---|
-| task style | multi-turn tool-use | single-turn reasoning |
-| user simulator | yes (LLM) | no |
-| DB state | yes | no |
-| checker | DB-predicate | regex + set/value compare |
-| target | Pillar 1 | Pillar 2 |
-| model | deepseek-chat (default) | deepseek-reasoner (only R1 has `reasoning_content` for harvest to work on) |
-
-The two harnesses answer different questions:
-
-- **τ-bench-lite**: *"Does Cache-First actually cut cost on real tool-use
-  workflows?"* (yes — 47.7pp cache, −39% cost on 48-run data.)
-- **harvest-bench**: *"Does the extra V3 harvest call add measurable
-  value above plain R1?"* (TBD — run it to find out.)
-
-## Modes
-
-Three modes isolate one variable each:
-
-| mode | model | harvest | what it measures |
-|---|---|---|---|
-| `baseline` | deepseek-chat | off | floor reference — V3 at its best |
-| `reasoner` | deepseek-reasoner | off | raw R1 gain over V3 on reasoning |
-| `reasoner-harvest` | deepseek-reasoner | on | R1 + the extra V3 harvest call |
-
-Deltas:
-
-- `baseline → reasoner` answers "is R1 worth the price on these
-  problems?"
-- `reasoner → reasoner-harvest` answers "is the harvest call worth
-  its incremental cost?"
-
-## Quickstart
-
-```bash
-# Dry-run — no API, smoke-test the wiring
-npx tsx benchmarks/harvest/runner.ts --dry
-
-# Full run (live DeepSeek, costs ~$0.20-0.60 for 6 tasks × 3 modes × 1 repeat)
-export DEEPSEEK_API_KEY=sk-...
-npx tsx benchmarks/harvest/runner.ts
-
-# Tighter run with 3 repeats (~$0.60-2.00)
-npx tsx benchmarks/harvest/runner.ts --repeats 3
-
-# Only the hard tasks (skip the easy floor)
-npx tsx benchmarks/harvest/runner.ts --task pseudoprime_base2
-npx tsx benchmarks/harvest/runner.ts --task derangements_d7
-npx tsx benchmarks/harvest/runner.ts --task euler_quadratic_break
-
-# Bump per-call timeout (reasoner + harvest occasionally blow past the 120s default)
-npx tsx benchmarks/harvest/runner.ts --timeout 600
-
-# Per-run transcripts so you can reasonix replay / diff them
-npx tsx benchmarks/harvest/runner.ts --repeats 3 --transcripts-dir transcripts/
-
-# Render report
-npx tsx benchmarks/harvest/report.ts benchmarks/harvest/results-*.json
-```
-
-## Tasks (v0.3 seed)
-
-Split into two bands:
-
-**Easy band** — V3 chat solves these, so baseline has a real pass rate and the bench has a reference floor.
-
-| id | shape | why |
-|---|---|---|
-| `mod7_list` | number theory, 29-element list | R1 often tries enumeration first, then reaches for modular arithmetic — clean rejectedPaths signal |
-| `flips_until_3heads` | probability, single integer | classic recurrence; R1 either derives or recalls, harvest should see hypotheses diverge |
-| `three_hats` | logic puzzle, one-word answer | pure deduction chain, tests harvest's ability to extract the nested reasoning |
-
-**Hard band** — picked specifically for known V3 failure modes. If R1 systematically beats V3 on these, and harvest further beats R1-only, we have our Pillar 2 story.
-
-| id | shape | why V3 tends to fail |
-|---|---|---|
-| `pseudoprime_base2` | smallest Fermat pseudoprime to base 2 = **341** | V3 often answers 561 (Carmichael) or some other composite it remembers — R1 actually checks 2^n mod n |
-| `derangements_d7` | D_7 = **1854** | V3 approximates via n!/e and rounds wrong, or recalls D_6/D_8; R1 uses the recurrence D_n=(n−1)(D_{n−1}+D_{n−2}) |
-| `euler_quadratic_break` | smallest n where n²+n+41 is composite = **40** | V3 frequently confuses with n²−n+41 (first fails at 41); R1 checks small n systematically |
-
-Adding a new task: see `tasks.ts`. Any checker that's deterministic is
-fair game — extract numeric/list/text with regex, compare with set
-equality or string matching.
-
-## Non-goals
-
-- **No LLM-as-judge.** Brittle and expensive; defeats the point of a
-  reproducible bench. If a checker is too hard to write deterministically,
-  the task doesn't belong here.
-- **No tool-use tasks.** Those live in `tau-bench/`. Different story.
-- **No multi-turn.** v0.3 harness is single-turn Q/A. Multi-turn reasoning
-  eval is separate scope.
-- **No benchmark-data cherry-picking.** When results come in, we publish
-  them whether they validate Pillar 2 or not. "harvest didn't help on
-  these 3 tasks" is still useful information.

benchmarks/harvest/report.md

@@ -1,82 +0,0 @@
-# Reasonix harvest eval (Pillar 2)
-
-**Date:** 2026-04-21T16:44:21.011Z
-**Tasks:** 6 · repeats × 1 · modes: baseline, reasoner, reasoner-harvest
-**Reasonix version:** 0.2.2
-
-## Summary by mode
-
-| mode | runs | pass rate | cache hit | cost / run | harvest turns | subgoals | uncertainties |
-|---|---:|---:|---:|---:|---:|---:|---:|
-| baseline | 6 | 100% | 39.5% | $0.001371 | 0.0 | 0.0 | 0.0 |
-| reasoner | 6 | 100% | 74.2% | $0.004170 | 0.0 | 0.0 | 0.0 |
-| reasoner-harvest | 6 | 83% | 61.6% | $0.003304 | 0.8 | 2.8 | 1.3 |
-
-## Deltas
-
-- **baseline → reasoner**
-  - pass rate: 0pp
-  - cost: ×3.04 (each run costs more)
-  - harvest signal / run: 0.0 subgoals, 0.0 uncertainties
-
-- **baseline → reasoner-harvest**
-  - pass rate: -17pp
-  - cost: ×2.41 (each run costs more)
-  - harvest signal / run: 2.8 subgoals, 1.3 uncertainties
-
-## Per-task breakdown
-
-| task | mode | rep | verdict | cache | cost | sg | un | note |
-|---|---|---:|:---:|---:|---:|---:|---:|---|
-| mod7_list | baseline | 1 | ✅ | 86.5% | $0.001361 | 0 | 0 |  |
-| mod7_list | reasoner | 1 | ✅ | 86.5% | $0.004797 | 0 | 0 |  |
-| mod7_list | reasoner-harvest | 1 | ✅ | 86.5% | $0.003045 | 4 | 2 |  |
-| flips_until_3heads | baseline | 1 | ✅ | 92.8% | $0.001198 | 0 | 0 |  |
-| flips_until_3heads | reasoner | 1 | ✅ | 92.8% | $0.002830 | 0 | 0 |  |
-| flips_until_3heads | reasoner-harvest | 1 | ✅ | 92.8% | $0.001691 | 4 | 0 |  |
-| three_hats | baseline | 1 | ✅ | 57.7% | $0.001514 | 0 | 0 |  |
-| three_hats | reasoner | 1 | ✅ | 57.7% | $0.009340 | 0 | 0 |  |
-| three_hats | reasoner-harvest | 1 | ✅ | 57.7% | $0.009651 | 3 | 2 |  |
-| pseudoprime_base2 | baseline | 1 | ✅ | 0.0% | $0.001894 | 0 | 0 |  |
-| pseudoprime_base2 | reasoner | 1 | ✅ | 75.3% | $0.003128 | 0 | 0 |  |
-| pseudoprime_base2 | reasoner-harvest | 1 | ❌ | 0.0% | $0.000000 | 0 | 0 | This operation was aborted |
-| derangements_d7 | baseline | 1 | ✅ | 0.0% | $0.001273 | 0 | 0 |  |
-| derangements_d7 | reasoner | 1 | ✅ | 71.9% | $0.003492 | 0 | 0 |  |
-| derangements_d7 | reasoner-harvest | 1 | ✅ | 71.9% | $0.002843 | 3 | 2 |  |
-| euler_quadratic_break | baseline | 1 | ✅ | 0.0% | $0.000985 | 0 | 0 |  |
-| euler_quadratic_break | reasoner | 1 | ✅ | 61.0% | $0.001431 | 0 | 0 |  |
-| euler_quadratic_break | reasoner-harvest | 1 | ✅ | 61.0% | $0.002594 | 3 | 2 |  |
-
-## Scope
-
-Unlike τ-bench-lite, these tasks are single-turn reasoning problems (no user simulator, no DB, no tool calls). Checkers are deterministic — regex + set / value compare, never an LLM judge. The point is to isolate whether the Pillar 2 harvest step adds measurable value above plain reasoner usage.
-
-Interpretation: `baseline` (chat / V3) is a floor. `reasoner` shows the raw R1 gain. `reasoner-harvest` isolates the cost + quality delta from the extra V3 harvest call.
-## Findings (v0.3, 18-run set)
-
-Two bands of tasks tried (easy: mod7 / flips / hats; hard: pseudoprime341 / D_7 / Euler quadratic). The hard band was chosen for *known V3 failure modes*. Result:
-
-1. **V3 chat passed all six tasks, including the three "hard" ones.** DeepSeek V3 knew 341 is the smallest base-2 pseudoprime, computed D_7=1854, and identified n=40 as the first Euler-polynomial failure. The tasks are well-known enough to be in training; the trap-answer variants (561, 265, 41) that the checkers specifically reject didn't fire.
-2. **Reasoner cost 3.04× baseline for identical pass rates (6/6 both sides).** On this task set, R1 adds cost and latency without measurable quality. The cache-hit story *does* extend to reasoner (74.2% mean), so Pillar 1 generalizes to R1 — but that's not a Pillar 2 story.
-3. **Harvest produced real signal** (mean 2.8 subgoals, 1.3 uncertainties per run where it fired) but didn't improve outcomes. One run hit the 300s timeout (bumped from 120s), suggesting reasoner-harvest latency is still unpredictable on some problems.
-
-### What this tells us about Pillar 2 positioning
-
-We've now given harvest two shots at showing answer-quality value on pure reasoning tasks. Both times V3 ate its lunch. This isn't a framework bug — it's a positioning signal:
-
-- **On well-known math / logic / counting problems, DeepSeek V3 is strong enough that paying for R1 is not justified by accuracy.** Pillar 2's "smart preset" cost multiplier (10×) quoted in the README is real but for most single-question Q/A it buys nothing.
-- **Harvest's plausible value surfaces are not "better math answers".** More likely: (a) transparency / auditability for developers, (b) planning support when tools are in the mix, (c) driving branch-sampling on cases where uncertainty correlates with wrongness. None of those are tested here.
-
-### Recommendation
-
-Don't ship v0.3 on the "harvest makes answers more correct" claim — the data doesn't support it. Two honest paths forward:
-
-- **Reframe Pillar 2 as a developer-facing introspection feature** (debug visibility into R1's planning). Keep harvest opt-in; stop implying it improves accuracy on end-user tasks.
-- **Move harvest eval to tool-use contexts** where uncertainty matters for *which tool to call next*, not *is this math right*. That's a new task shape — different harness, not a bigger task set here.
-
-In either case, the next Reasonix release is better served by MCP client (Pillar 1 leverage across the ecosystem) than by a bigger harvest-bench data push.
-
-### Notes
-
-- Cache hit on baseline runs surprised me (39.5% mean). That's DeepSeek's *cross-session* prompt cache: the same system prompt ran many times across modes and runs warmed the server-side cache. Real cross-session evidence that Cache-First's value isn't only per-session.
-- 5-subgoals cap still hit uniformly. `HarvestOptions.maxItems` default is 5; harvest is truncating. Moot unless we find a setting where that matters.

benchmarks/harvest/report.ts

@@ -1,229 +0,0 @@
-/** Render a harvest-bench results.json into a markdown report. CLI usage in README. */
-
-import { readFileSync, writeFileSync } from "node:fs";
-import { pathToFileURL } from "node:url";
-import {
-  ALL_MODES,
-  type HarvestBenchReport,
-  type HarvestMode,
-  type HarvestRunResult,
-} from "./types.js";
-
-interface CliArgs {
-  input: string;
-  outPath: string;
-}
-
-function parseArgs(argv: string[]): CliArgs {
-  const out: CliArgs = { input: "", outPath: "benchmarks/harvest/report.md" };
-  const positional: string[] = [];
-  for (let i = 0; i < argv.length; i++) {
-    const a = argv[i];
-    if (a === "--out") out.outPath = argv[++i] ?? out.outPath;
-    else if (a && !a.startsWith("--")) positional.push(a);
-  }
-  out.input = positional[0] ?? "";
-  if (!out.input) {
-    throw new Error("usage: npx tsx benchmarks/harvest/report.ts <results.json> [--out report.md]");
-  }
-  return out;
-}
-
-interface Agg {
-  runs: number;
-  passes: number;
-  avgCache: number;
-  avgCost: number;
-  avgClaudeCost: number;
-  avgHarvestedTurns: number;
-  avgSubgoals: number;
-  avgUncertainties: number;
-}
-
-function aggregate(results: HarvestRunResult[]): Agg {
-  if (results.length === 0) {
-    return {
-      runs: 0,
-      passes: 0,
-      avgCache: 0,
-      avgCost: 0,
-      avgClaudeCost: 0,
-      avgHarvestedTurns: 0,
-      avgSubgoals: 0,
-      avgUncertainties: 0,
-    };
-  }
-  const passes = results.filter((r) => r.verdict === "pass").length;
-  const mean = (fn: (r: HarvestRunResult) => number) =>
-    results.reduce((s, r) => s + fn(r), 0) / results.length;
-  return {
-    runs: results.length,
-    passes,
-    avgCache: mean((r) => r.cacheHitRatio),
-    avgCost: mean((r) => r.costUsd),
-    avgClaudeCost: mean((r) => r.claudeEquivalentUsd),
-    avgHarvestedTurns: mean((r) => r.harvestedTurns),
-    avgSubgoals: mean((r) => r.totalSubgoals),
-    avgUncertainties: mean((r) => r.totalUncertainties),
-  };
-}
-
-export function renderReport(report: HarvestBenchReport): string {
-  const lines: string[] = [];
-  lines.push("# Reasonix harvest eval (Pillar 2)");
-  lines.push("");
-  lines.push(`**Date:** ${report.meta.date}`);
-  lines.push(
-    `**Tasks:** ${report.meta.taskCount} · repeats × ${report.meta.repeatsPerTask} · modes: ${report.meta.modesRun.join(", ")}`,
-  );
-  lines.push(`**Reasonix version:** ${report.meta.reasonixVersion}`);
-  lines.push("");
-
-  // Per-mode summary
-  lines.push("## Summary by mode");
-  lines.push("");
-  lines.push(
-    "| mode | runs | pass rate | cache hit | cost / run | harvest turns | subgoals | uncertainties |",
-  );
-  lines.push("|---|---:|---:|---:|---:|---:|---:|---:|");
-  for (const mode of ALL_MODES) {
-    const rs = report.results.filter((r) => r.mode === mode);
-    const a = aggregate(rs);
-    if (a.runs === 0) continue;
-    lines.push(
-      `| ${mode} | ${a.runs} | ${pct(a.passes, a.runs)} | ${pct1(a.avgCache)} | $${a.avgCost.toFixed(6)} | ${a.avgHarvestedTurns.toFixed(1)} | ${a.avgSubgoals.toFixed(1)} | ${a.avgUncertainties.toFixed(1)} |`,
-    );
-  }
-  lines.push("");
-
-  // Pairwise deltas that actually answer the question
-  lines.push("## Deltas");
-  lines.push("");
-  const modes = ALL_MODES.filter((m) => report.results.some((r) => r.mode === m));
-  if (modes.length >= 2) {
-    const baseline = aggregate(report.results.filter((r) => r.mode === modes[0]));
-    for (let i = 1; i < modes.length; i++) {
-      const mode = modes[i]!;
-      const agg = aggregate(report.results.filter((r) => r.mode === mode));
-      const passDelta =
-        agg.passes / Math.max(agg.runs, 1) - baseline.passes / Math.max(baseline.runs, 1);
-      const costRatio = baseline.avgCost > 0 ? agg.avgCost / baseline.avgCost : 0;
-      lines.push(`- **${modes[0]} → ${mode}**`);
-      lines.push(`  - pass rate: ${(passDelta * 100).toFixed(0)}pp`);
-      lines.push(
-        `  - cost: ×${costRatio.toFixed(2)} (each run costs ${costRatio > 1 ? "more" : "less"})`,
-      );
-      lines.push(
-        `  - harvest signal / run: ${agg.avgSubgoals.toFixed(1)} subgoals, ${agg.avgUncertainties.toFixed(1)} uncertainties`,
-      );
-      lines.push("");
-    }
-  }
-
-  // Per-task breakdown
-  lines.push("## Per-task breakdown");
-  lines.push("");
-  const byTask = new Map<string, HarvestRunResult[]>();
-  for (const r of report.results) {
-    const list = byTask.get(r.taskId) ?? [];
-    list.push(r);
-    byTask.set(r.taskId, list);
-  }
-  lines.push("| task | mode | rep | verdict | cache | cost | sg | un | note |");
-  lines.push("|---|---|---:|:---:|---:|---:|---:|---:|---|");
-  for (const [taskId, runs] of byTask) {
-    for (const r of runs) {
-      const mark = r.verdict === "pass" ? "✅" : r.verdict === "fail" ? "❌" : "•";
-      lines.push(
-        `| ${taskId} | ${r.mode} | ${r.repeat} | ${mark} | ${pct1(r.cacheHitRatio)} | $${r.costUsd.toFixed(6)} | ${r.totalSubgoals} | ${r.totalUncertainties} | ${truncate(r.checkNote ?? "", 40)} |`,
-      );
-    }
-  }
-  lines.push("");
-
-  lines.push("## Scope");
-  lines.push("");
-  lines.push(
-    "Unlike τ-bench-lite, these tasks are single-turn reasoning problems (no user simulator, no DB, no tool calls). Checkers are deterministic — regex + set / value compare, never an LLM judge. The point is to isolate whether the Pillar 2 harvest step adds measurable value above plain reasoner usage.",
-  );
-  lines.push("");
-  lines.push(
-    "Interpretation: `baseline` (chat / V3) is a floor. `reasoner` shows the raw R1 gain. `reasoner-harvest` isolates the cost + quality delta from the extra V3 harvest call.",
-  );
-
-  return lines.join("\n");
-}
-
-/** Findings prose lives in report.md (hand-written after rendering); this stub keeps the function shape. */
-export function renderFindings(_report: HarvestBenchReport): string {
-  return [
-    "",
-    "## Findings (v0.3, 18-run set)",
-    "",
-    "Two bands of tasks tried (easy: mod7 / flips / hats; hard: pseudoprime341 / D_7 / Euler quadratic). The hard band was chosen for *known V3 failure modes*. Result:",
-    "",
-    '1. **V3 chat passed all six tasks, including the three "hard" ones.** DeepSeek V3 knew 341 is the smallest base-2 pseudoprime, computed D_7=1854, and identified n=40 as the first Euler-polynomial failure. The tasks are well-known enough to be in training; the trap-answer variants (561, 265, 41) that the checkers specifically reject didn\'t fire.',
-    "2. **Reasoner cost 3.04× baseline for identical pass rates (6/6 both sides).** On this task set, R1 adds cost and latency without measurable quality. The cache-hit story *does* extend to reasoner (74.2% mean), so Pillar 1 generalizes to R1 — but that's not a Pillar 2 story.",
-    "3. **Harvest produced real signal** (mean 2.8 subgoals, 1.3 uncertainties per run where it fired) but didn't improve outcomes. One run hit the 300s timeout (bumped from 120s), suggesting reasoner-harvest latency is still unpredictable on some problems.",
-    "",
-    "### What this tells us about Pillar 2 positioning",
-    "",
-    "We've now given harvest two shots at showing answer-quality value on pure reasoning tasks. Both times V3 ate its lunch. This isn't a framework bug — it's a positioning signal:",
-    "",
-    '- **On well-known math / logic / counting problems, DeepSeek V3 is strong enough that paying for R1 is not justified by accuracy.** Pillar 2\'s "smart preset" cost multiplier (10×) quoted in the README is real but for most single-question Q/A it buys nothing.',
-    '- **Harvest\'s plausible value surfaces are not "better math answers".** More likely: (a) transparency / auditability for developers, (b) planning support when tools are in the mix, (c) driving branch-sampling on cases where uncertainty correlates with wrongness. None of those are tested here.',
-    "",
-    "### Recommendation",
-    "",
-    "Don't ship v0.3 on the \"harvest makes answers more correct\" claim — the data doesn't support it. Two honest paths forward:",
-    "",
-    "- **Reframe Pillar 2 as a developer-facing introspection feature** (debug visibility into R1's planning). Keep harvest opt-in; stop implying it improves accuracy on end-user tasks.",
-    "- **Move harvest eval to tool-use contexts** where uncertainty matters for *which tool to call next*, not *is this math right*. That's a new task shape — different harness, not a bigger task set here.",
-    "",
-    "In either case, the next Reasonix release is better served by MCP client (Pillar 1 leverage across the ecosystem) than by a bigger harvest-bench data push.",
-    "",
-    "### Notes",
-    "",
-    "- Cache hit on baseline runs surprised me (39.5% mean). That's DeepSeek's *cross-session* prompt cache: the same system prompt ran many times across modes and runs warmed the server-side cache. Real cross-session evidence that Cache-First's value isn't only per-session.",
-    "- 5-subgoals cap still hit uniformly. `HarvestOptions.maxItems` default is 5; harvest is truncating. Moot unless we find a setting where that matters.",
-    "",
-  ].join("\n");
-}
-
-function pct(num: number, denom: number): string {
-  if (denom === 0) return "—";
-  return `${((num / denom) * 100).toFixed(0)}%`;
-}
-
-function pct1(x: number): string {
-  return `${(x * 100).toFixed(1)}%`;
-}
-
-function truncate(s: string, n: number): string {
-  return s.length > n ? `${s.slice(0, n)}…` : s;
-}
-
-async function main(): Promise<void> {
-  const args = parseArgs(process.argv.slice(2));
-  const raw = readFileSync(args.input, "utf8");
-  const report = JSON.parse(raw) as HarvestBenchReport;
-  const md = renderReport(report) + renderFindings(report);
-  writeFileSync(args.outPath, md, "utf8");
-  console.log(`wrote ${args.outPath} (${report.results.length} runs)`);
-}
-
-function isMain(): boolean {
-  const entry = process.argv[1];
-  if (!entry) return false;
-  return import.meta.url === pathToFileURL(entry).href;
-}
-
-if (isMain()) {
-  main().catch((err) => {
-    console.error(err);
-    process.exit(1);
-  });
-}
-
-// Suppress unused import warning — kept in scope so ALL_MODES order matches renderer iteration.
-export type { HarvestMode };

benchmarks/harvest/results.json

@@ -1,305 +0,0 @@
-{
-  "meta": {
-    "date": "2026-04-21T16:44:21.011Z",
-    "repeatsPerTask": 1,
-    "taskCount": 6,
-    "modesRun": [
-      "baseline",
-      "reasoner",
-      "reasoner-harvest"
-    ],
-    "reasonixVersion": "0.2.2"
-  },
-  "results": [
-    {
-      "taskId": "mod7_list",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Alright, let's go step-by-step.\n\n---\n\n## **Step 1: Problem restatement**\nWe want positive integers \\( n \\leq 100 \\) such that  \n\\[\nn^2 + n + 1 \\equiv 0 \\pmod{7}.\n\\]\n\n---\n\n## **Step 2: Try residues modulo 7**\nLet \\( n \\equiv r \\pmod{7} \\), where \\( r \\in \\{0, 1, 2, 3, 4, 5, 6\\} \\).\n\nWe compute \\( r^2 + r + 1 \\mod 7 \\):\n\n- \\( r = 0 \\): \\( 0 + 0 + 1 = 1 \\) → not divisible by 7.\n- \\( r = 1 \\): \\( 1 + 1 + 1 = 3 \\) → no.\n- \\( r = 2 \\): \\( 4 + 2 + 1 = 7 \\equiv 0 \\) → yes.\n- \\( r = 3 \\): \\( 9 + 3 + 1 = 13 \\equiv 6 \\) → no.\n- \\( r = 4 \\): \\( 16 + 4 + 1 = 21 \\equiv 0 \\) → yes.\n- \\( r = 5 \\): \\( 25 + 5 + 1 = 31 \\equiv 3 \\) → no.\n- \\( r = 6 \\): \\( 36 + 6 + 1 = 43 \\equiv 1 \\) → no.\n\nSo \\( n \\equiv 2 \\pmod{7} \\) or \\( n \\equiv 4 \\pmod{7} \\).\n\n---\n\n## **Step 3: General form**\n\\[\nn \\equiv 2 \\pmod{7} \\quad \\text{or} \\quad n \\equiv 4 \\pmod{7}.\n\\]\n\nThat means \\( n = 7k + 2 \\) or \\( n = 7k + 4 \\) for some integer \\( k \\ge 0 \\).\n\n---\n\n## **Step 4: Positive integers \\( n \\leq 100 \\)**\n**Case 1:** \\( n = 7k + 2 \\), \\( n \\leq 100 \\), \\( k \\ge 0 \\).\n\n\\( 7k + 2 \\leq 100 \\) → \\( 7k \\leq 98 \\) → \\( k \\leq 14 \\).\n\nSo \\( k = 0, 1, 2, \\dots, 14 \\) gives \\( n = 2, 9, 16, \\dots, 100 \\).\n\nCheck \\( n=100 \\): \\( 7 \\times 14 + 2 = 98 + 2 = 100 \\) okay.\n\n---\n\n**Case 2:** \\( n = 7k + 4 \\), \\( n \\leq 100 \\), \\( k \\ge 0 \\).\n\n\\( 7k + 4 \\leq 100 \\) → \\( 7k \\leq 96 \\) → \\( k \\leq 13 \\) (since \\( 7 \\times 13 = 91 \\), \\( 91+4=95 \\), \\( 7\\times14=98 \\) → \\( 98+4=102>100 \\)).\n\nSo \\( k = 0, 1, 2, \\dots, 13 \\) gives \\( n = 4, 11, 18, \\dots, 95 \\).\n\n---\n\n## **Step 5: List all such \\( n \\leq 100 \\)**\nFrom case 1: \\( 2, 9, 16, 23, 30, 37, 44, 51, 58, 65, 72, 79, 86, 93, 100 \\).\n\nFrom case 2: \\( 4, 11, 18, 25, 32, 39, 46, 53, 60, 67, 74, 81, 88, 95 \\).\n\nCombine and sort:\n\n\\[\n2, 4, 9, 11, 16, 18, 23, 25, 30, 32, 37, 39, 44, 46, 51, 53, 58, 60, 65, 67, 72, 74, 79, 81, 86, 88, 93, 95, 100.\n\\]\n\n---\n\n## **Step 6: Check count**\nEach sequence has 15 and 14 terms respectively, total \\( 15 + 14 = 29 \\) numbers.\n\nCheck: \\( n=2 \\) works? \\( 4+2+1=7 \\) yes. \\( n=4 \\) works? \\( 16+4+1=21 \\) yes. \\( n=100 \\) works? \\( 10000+100+1=10101 \\), divide by 7: \\( 10101/7 = 1443 \\) yes.\n\n---\n\n**Final answer:**\n\\[\n\\boxed{2, 4, 9, 11, 16, 18, 23, 25, 30, 32, 37, 39, 44, 46, 51, 53, 58, 60, 65, 67, 72, 74, 79, 81, 86, 88, 93, 95, 100}\n\\]\n\n---\n\nAnswer: 2, 4, 9, 11, 16, 18, 23, 25, 30, 32, 37, 39, 44, 46, 51, 53, 58, 60, 65, 67, 72, 74, 79, 81, 86, 88, 93, 95, 100",
-      "turns": 1,
-      "cacheHitRatio": 0.8648648648648649,
-      "costUsd": 0.0013612800000000003,
-      "claudeEquivalentUsd": 0.018687,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "mod7_list",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The condition \\(n^2 + n + 1\\) divisible by 7 is equivalent to \\(n^2 + n + 1 \\equiv 0 \\pmod{7}\\). Testing residues modulo 7:\n\n- \\(n \\equiv 0: 0+0+1 \\equiv 1\\)  \n- \\(n \\equiv 1: 1+1+1 \\equiv 3\\)  \n- \\(n \\equiv 2: 4+2+1 \\equiv 7 \\equiv 0\\)  \n- \\(n \\equiv 3: 9+3+1 \\equiv 13 \\equiv 6\\)  \n- \\(n \\equiv 4: 16+4+1 \\equiv 21 \\equiv 0\\)  \n- \\(n \\equiv 5: 25+5+1 \\equiv 31 \\equiv 3\\)  \n- \\(n \\equiv 6: 36+6+1 \\equiv 43 \\equiv 1\\)\n\nThus, \\(n \\equiv 2\\) or \\(n \\equiv 4 \\pmod{7}\\) are the solutions modulo 7. For positive integers \\(n \\leq 100\\), we list all numbers congruent to 2 or 4 modulo 7 in increasing order.\n\nAnswer: 2, 4, 9, 11, 16, 18, 23, 25, 30, 32, 37, 39, 44, 46, 51, 53, 58, 60, 65, 67, 72, 74, 79, 81, 86, 88, 93, 95, 100",
-      "turns": 1,
-      "cacheHitRatio": 0.8648648648648649,
-      "costUsd": 0.00479742,
-      "claudeEquivalentUsd": 0.032982,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "mod7_list",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The congruence \\(n^2 + n + 1 \\equiv 0 \\pmod{7}\\) is solved by testing residues modulo 7. The solutions are \\(n \\equiv 2\\) and \\(n \\equiv 4 \\pmod{7}\\). For positive integers \\(n \\le 100\\), these yield \\(n = 7k + 2\\) with \\(k = 0, 1, \\dots, 14\\) and \\(n = 7k + 4\\) with \\(k = 0, 1, \\dots, 13\\).\n\nAnswer: 2, 4, 9, 11, 16, 18, 23, 25, 30, 32, 37, 39, 44, 46, 51, 53, 58, 60, 65, 67, 72, 74, 79, 81, 86, 88, 93, 95, 100",
-      "turns": 1,
-      "cacheHitRatio": 0.8648648648648649,
-      "costUsd": 0.00304542,
-      "claudeEquivalentUsd": 0.020982,
-      "harvestedTurns": 1,
-      "totalSubgoals": 4,
-      "totalUncertainties": 2,
-      "totalHypotheses": 2,
-      "totalRejectedPaths": 2
-    },
-    {
-      "taskId": "flips_until_3heads",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Alright, let's reason through this step-by-step.\n\n---\n\n## **Step 1: Define states**\nLet’s define states based on how many consecutive heads we have at the moment.\n\n- \\( E_0 \\): Expected flips to reach 3 heads in a row starting from 0 consecutive heads (or just starting, or after a tail).\n- \\( E_1 \\): Expected flips to reach 3 heads in a row starting from 1 consecutive head.\n- \\( E_2 \\): Expected flips to reach 3 heads in a row starting from 2 consecutive heads.\n\nWe want \\( E_0 \\).\n\n---\n\n## **Step 2: Transition equations**\nFrom \\( E_0 \\):\n- Flip H (prob \\( \\frac12 \\)) → go to state \\( E_1 \\), 1 flip used.\n- Flip T (prob \\( \\frac12 \\)) → stay in \\( E_0 \\), 1 flip used.\n\nSo:\n\\[\nE_0 = \\frac12 (1 + E_1) + \\frac12 (1 + E_0)\n\\]\n\n---\n\nFrom \\( E_1 \\) (already have 1 head):\n- Flip H (prob \\( \\frac12 \\)) → go to state \\( E_2 \\), 1 flip used.\n- Flip T (prob \\( \\frac12 \\)) → go to state \\( E_0 \\), 1 flip used.\n\nSo:\n\\[\nE_1 = \\frac12 (1 + E_2) + \\frac12 (1 + E_0)\n\\]\n\n---\n\nFrom \\( E_2 \\) (already have 2 heads):\n- Flip H (prob \\( \\frac12 \\)) → done (3 heads in a row), 1 flip used.\n- Flip T (prob \\( \\frac12 \\)) → go to state \\( E_0 \\), 1 flip used.\n\nSo:\n\\[\nE_2 = \\frac12 (1 + 0) + \\frac12 (1 + E_0)\n\\]\n\\[\nE_2 = \\frac12 \\cdot 1 + \\frac12 (1 + E_0) \\quad\\text{(Wait, careful:)}\n\\]\nActually:  \nIf H: \\( 1 \\) flip and done.  \nIf T: \\( 1 \\) flip and then \\( E_0 \\) more.\n\nSo:\n\\[\nE_2 = \\frac12 \\cdot 1 + \\frac12 (1 + E_0)\n\\]\nCheck:  \nH: 1 flip → total 1, done. Contribution: \\( \\frac12 \\times 1 \\).  \nT: 1 flip + \\( E_0 \\) more → total \\( 1 + E_0 \\). Contribution: \\( \\frac12 \\times (1 + E_0) \\).\n\nYes:\n\\[\nE_2 = \\frac12 \\cdot 1 + \\frac12 (1 + E_0) = \\frac12 + \\frac12 + \\frac12 E_0 = 1 + \\frac{E_0}{2}.\n\\]\n\n---\n\n## **Step 3: Solve system**\nWe have:\n\n(1) \\( E_0 = \\frac12 (1 + E_1) + \\frac12 (1 + E_0) \\)  \nSimplify: \\( E_0 = 1 + \\frac12 E_1 + \\frac12 E_0 \\)  \nMultiply by 2: \\( 2E_0 = 2 + E_1 + E_0 \\)  \nSo \\( E_0 - E_1 = 2 \\) … (Eq A)\n\n---\n\n(2) \\( E_1 = \\frac12 (1 + E_2) + \\frac12 (1 + E_0) \\)  \nSimplify: \\( E_1 = 1 + \\frac12 E_2 + \\frac12 E_0 \\) … (Eq B)\n\n---\n\n(3) \\( E_2 = 1 + \\frac12 E_0 \\) … (Eq C)\n\n---\n\nFrom Eq C into Eq B:\n\\[\nE_1 = 1 + \\frac12 \\left( 1 + \\frac12 E_0 \\right) + \\frac12 E_0\n\\]\n\\[\nE_1 = 1 + \\frac12 + \\frac14 E_0 + \\frac12 E_0\n\\]\n\\[\nE_1 = \\frac32 + \\frac34 E_0\n\\]\n\n---\n\nFrom Eq A: \\( E_0 - E_1 = 2 \\)  \nSubstitute \\( E_1 \\):\n\\[\nE_0 - \\left( \\frac32 + \\frac34 E_0 \\right) = 2\n\\]\n\\[\nE_0 - \\frac32 - \\frac34 E_0 = 2\n\\]\n\\[\n\\frac14 E_0 - \\frac32 = 2\n\\]\n\\[\n\\frac14 E_0 = 2 + \\frac32 = \\frac72\n\\]\n\\[\nE_0 = 14\n\\]\n\n---\n\n## **Step 4: Conclusion**\nThe expected number of flips until we get 3 heads in a row is \\( 14 \\).\n\n---\n\n**Answer:** 14",
-      "turns": 1,
-      "cacheHitRatio": 0.927536231884058,
-      "costUsd": 0.00119823,
-      "claudeEquivalentUsd": 0.016467,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "flips_until_3heads",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The expected number of flips required to get three consecutive heads is 14, derived from solving the system of linear equations for the Markov chain states representing the current streak of heads.\n\nAnswer: 14",
-      "turns": 1,
-      "cacheHitRatio": 0.927536231884058,
-      "costUsd": 0.00283024,
-      "claudeEquivalentUsd": 0.019512,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "flips_until_3heads",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The expected number of flips until three consecutive heads appear is 14.\n\nAnswer: 14",
-      "turns": 1,
-      "cacheHitRatio": 0.927536231884058,
-      "costUsd": 0.00169144,
-      "claudeEquivalentUsd": 0.011712,
-      "harvestedTurns": 1,
-      "totalSubgoals": 4,
-      "totalUncertainties": 0,
-      "totalHypotheses": 2,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "three_hats",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Let's break this down step by step.\n\n---\n\n**Step 1: Understanding the setup**  \nThree people: A (first), B (second), C (third).  \nEach hat is red (R) or blue (B).  \nThey know: at least one hat is red.  \nEach sees the other two hats, not their own.  \nThey are asked in order: \"Do you know your hat color?\"  \nA answers no, B answers no, C answers yes.  \nWe want C’s hat color.\n\n---\n\n**Step 2: From A’s perspective when asked**  \nA sees B and C.  \nIf A had seen two blue hats, then A would know: \"At least one red hat means my hat must be red.\"  \nSo A says \"no\" → A did **not** see two blue hats.  \nThus A sees at least one red hat on {B, C}.\n\nPossible cases for what A sees (B, C):  \n(R, R), (R, B), (B, R) — all have at least one red.\n\n---\n\n**Step 3: B’s reasoning when asked**  \nB hears A say \"no\".  \nB sees A and C.  \n\nLet’s consider B’s possible deductions from A’s \"no\":  \nB knows: If A had seen two blues, A would have said yes. Since A said no, A saw at least one red among B and C.  \nBut B sees A and C.  \n\nFrom B’s viewpoint:  \n- If B saw C with blue and A with blue, then B would know: \"A saw me and C both blue → impossible because A said no.\" But B doesn’t see that because if B saw A blue and C blue, then B would know his own hat must be red (since at least one red total). But then B would answer yes.  \nSo B’s \"no\" means B did **not** see A and C both blue.  \n\nAlso, B considers: Could my hat be blue? Let’s check:  \nSuppose B sees A red, C blue. Then B thinks: \"If my hat were blue, then A would have seen (B blue, C blue) → A would have said yes. But A said no, so my hat cannot be blue.\"  \nSo if B saw (A red, C blue), B would deduce: \"My hat must be red.\" Then B would answer yes.  \nBut B said no → so B did **not** see (A red, C blue).\n\nSimilarly, if B saw (A blue, C red):  \nThen B thinks: \"If my hat were blue, then A would have seen (B blue, C red) → that’s fine (A sees one red on C), so A would still say no. If my hat were red, then A would see (B red, C red) → A would still say no. So I don’t know.\" That’s consistent with B saying no.  \n\nIf B saw (A red, C red):  \nThen B thinks: \"If my hat were blue, A sees (B blue, C red) → A says no. If my hat were red, A sees (B red, C red) → A says no. So I don’t know.\" That’s also consistent with B saying no.\n\nSo B says no in cases:  \nB sees (A blue, C red) or (A red, C red).  \nB would say yes if B saw (A red, C blue) — but that didn’t happen.  \n\n---\n\n**Step 4: C’s reasoning**  \nC hears A says no, B says no.  \nC sees A and B.  \n\nFrom C’s viewpoint:  \nLet’s list all possible (A, B, C) hat colorings with at least one red, consistent with A’s \"no\" (which we already incorporated: A sees at least one red on B/C) and B’s \"no\" (meaning B didn’t see A red, C blue — because that would make B know).\n\nPossible cases for (A, B) that C could see:  \n\n1. C sees (A blue, B red):  \n   Then C thinks:  \n   - If C were blue: (A blue, B red, C blue) → A sees (B red, C blue) → A says no. B sees (A blue, C blue) → B sees two blues? Wait, B sees A blue, C blue → then B knows his own hat is red (since at least one red total) → B would say yes. But B said no → contradiction. So C cannot be blue.  \n   So C must be red.  \n\n2. C sees (A red, B blue):  \n   Then C thinks:  \n   - If C were blue: (A red, B blue, C blue) → A sees (B blue, C blue) → A would say yes (since at least one red total means A’s hat must be red). But A said no → contradiction. So C cannot be blue.  \n   So C must be red.  \n\n3. C sees (A red, B red):  \n   Then C thinks:  \n   - If C were blue: (A red, B red, C blue) → A sees (B red, C blue) → A says no. B sees (A red, C blue) → B sees A red, C blue → earlier we said: If B sees A red, C blue, then B would deduce his own hat is red (reasoning: if B were blue, A would have seen two blues and said yes, but A said no, so B can’t be blue) → so B would say yes. But B actually said no → contradiction. So C cannot be blue.  \n   - If C were red: (A red, B red, C red) → A sees (B red, C red) → A says no. B sees (A red, C red) → B says no. Works fine.  \n\nSo in all cases C sees for (A, B), C can deduce C’s hat is **red** because assuming C is blue leads to contradiction with B’s \"no\" (in cases 1 and 3) or with A’s \"no\" (in case 2).\n\n---\n\n**Step 5: Conclusion**  \nC answers yes, and C’s hat must be red.\n\n---\n\nAnswer: red",
-      "turns": 1,
-      "cacheHitRatio": 0.5765765765765766,
-      "costUsd": 0.0015142700000000003,
-      "claudeEquivalentUsd": 0.020748,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "three_hats",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Answer: red",
-      "turns": 1,
-      "cacheHitRatio": 0.5765765765765766,
-      "costUsd": 0.009340119999999999,
-      "claudeEquivalentUsd": 0.064068,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "three_hats",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The third person deduces their hat color based on the previous responses. Since the first two answer \"no,\" it implies they each see at least one red hat and cannot determine their own color. Specifically, the second person's \"no\" indicates they see the third person's hat as red; otherwise, they would have known their own hat color. Therefore, the third person's hat must be red.\n\nAnswer: red",
-      "turns": 1,
-      "cacheHitRatio": 0.5765765765765766,
-      "costUsd": 0.0096511,
-      "claudeEquivalentUsd": 0.066198,
-      "harvestedTurns": 1,
-      "totalSubgoals": 3,
-      "totalUncertainties": 2,
-      "totalHypotheses": 3,
-      "totalRejectedPaths": 3
-    },
-    {
-      "taskId": "pseudoprime_base2",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Alright, let's go step-by-step.\n\n---\n\n**Step 1: Understanding the problem**\n\nWe want the smallest composite integer \\( n > 1 \\) such that\n\n\\[\n2^n \\equiv 2 \\pmod{n}.\n\\]\n\nEquivalently, \\( 2^{n-1} \\equiv 1 \\pmod{n} \\) for composite \\( n \\).\n\nThis means \\( n \\) is a **Fermat pseudoprime to base 2**.\n\nWe check composite numbers starting from the smallest.\n\n---\n\n**Step 2: Check small composites**\n\n- \\( n = 4 \\): composite.  \n\\( 2^4 = 16 \\), \\( 16 \\mod 4 = 0 \\), not \\( 2 \\). So \\( 2^4 \\equiv 0 \\not\\equiv 2 \\pmod{4} \\). Fails.\n\n- \\( n = 6 \\): composite.  \n\\( 2^6 = 64 \\), \\( 64 \\mod 6 = 4 \\), not \\( 2 \\). Fails.\n\n- \\( n = 8 \\): composite.  \n\\( 2^8 = 256 \\), \\( 256 \\mod 8 = 0 \\), not \\( 2 \\). Fails.\n\n- \\( n = 9 \\): composite.  \n\\( 2^9 = 512 \\), \\( 512 \\mod 9 \\):  \n\\( 512 \\div 9 \\), \\( 9 \\times 56 = 504 \\), remainder \\( 8 \\), not \\( 2 \\). Fails.\n\n- \\( n = 10 \\): composite.  \n\\( 2^{10} = 1024 \\), \\( 1024 \\mod 10 = 4 \\), not \\( 2 \\). Fails.\n\n---\n\n**Step 3: Try n = 15**\n\n\\( n = 15 \\): composite.  \n\\( 2^{15} = 32768 \\).  \nCheck mod 15:  \nWe can use mod arithmetic:  \n\\( 2^4 = 16 \\equiv 1 \\pmod{15} \\)?? Wait, check: \\( 16 \\mod 15 = 1 \\), yes.  \nSo \\( 2^{15} = (2^4)^3 \\times 2^3 = 1^3 \\times 8 = 8 \\pmod{15} \\).  \nSo \\( 2^{15} \\equiv 8 \\not\\equiv 2 \\pmod{15} \\). Fails.\n\n---\n\n**Step 4: Try n = 21**\n\n\\( n = 21 \\): composite.  \nMod 21:  \n\\( 2^6 = 64 \\equiv 1 \\pmod{21} \\)? Check: \\( 64 \\div 21 = 3 \\times 21 = 63 \\), remainder 1, yes.  \nSo \\( 2^{21} = (2^6)^3 \\times 2^3 \\equiv 1^3 \\times 8 = 8 \\pmod{21} \\). Not 2. Fails.\n\n---\n\n**Step 5: Try n = 22**\n\n\\( n = 22 \\): composite.  \nUse mod 22:  \nEuler’s totient not needed, just check order:  \n\\( 2^5 = 32 \\equiv 10 \\pmod{22} \\)  \n\\( 2^{10} = (2^5)^2 \\equiv 10^2 = 100 \\equiv 12 \\pmod{22} \\)  \n\\( 2^{11} \\equiv 24 \\equiv 2 \\pmod{22} \\)  \nSo \\( 2^{22} = (2^{11})^2 \\equiv 2^2 = 4 \\pmod{22} \\). Not 2. Fails.\n\n---\n\n**Step 6: Try n = 25**\n\n\\( n = 25 \\): composite.  \nCheck mod 25:  \n\\( 2^{10} = 1024 \\equiv 24 \\pmod{25} \\) (since \\( 1000 \\equiv 0 \\), \\( 24 \\equiv -1 \\))  \nActually check: \\( 1024 - 1000 = 24 \\), yes \\( 24 \\equiv -1 \\).  \nSo \\( 2^{20} \\equiv 1 \\pmod{25} \\).  \nThus \\( 2^{25} = 2^{20} \\times 2^5 \\equiv 1 \\times 32 = 7 \\pmod{25} \\). Not 2. Fails.\n\n---\n\n**Step 7: Known smallest base-2 Fermat pseudoprime**\n\nFrom known number theory facts:  \nThe smallest Fermat pseudoprime to base 2 is \\( n = 341 \\).  \nBut let's check smaller ones just in case I’m skipping one.\n\n---\n\n**Step 8: Try n = 91**\n\n91 = 7 × 13.  \nCheck mod 7: \\( 2^{91} \\equiv (2^6)^{15} \\times 2^1 \\equiv 1^{15} \\times 2 = 2 \\pmod{7} \\).  \nCheck mod 13: \\( 2^{12} \\equiv 1 \\pmod{13} \\),  \n\\( 91 = 12 \\times 7 + 7 \\), so \\( 2^{91} \\equiv 2^7 = 128 \\equiv 11 \\pmod{13} \\).  \nFor \\( 2^{91} \\equiv 2 \\pmod{91} \\), we need \\( 2^{91} \\equiv 2 \\) mod 7 and mod 13.  \nMod 13 we got 11, not 2. So fails.\n\n---\n\n**Step 9: Try n = 341**\n\n341 = 11 × 31.  \nCheck mod 11: \\( 2^{10} \\equiv 1 \\pmod{11} \\),  \n\\( 341 = 10 \\times 34 + 1 \\), so \\( 2^{341} \\equiv 2^1 = 2 \\pmod{11} \\). Good.  \nCheck mod 31: \\( 2^5 = 32 \\equiv 1 \\pmod{31} \\),  \n\\( 341 = 5 \\times 68 + 1 \\), so \\( 2^{341} \\equiv 2^1 = 2 \\pmod{31} \\).  \nSo \\( 2^{341} \\equiv 2 \\) mod 11 and mod 31, hence mod 341.  \nThus \\( 341 \\) works.\n\n---\n\n**Step 10: Check smaller than 341**\n\nKnown list: 341 is smallest. Let’s check a candidate like 121 = 11^2:  \n\\( 2^{121} \\mod 121 \\):  \nMod 11: \\( 2^{10} \\equiv 1 \\), \\( 121 = 10 \\times 12 + 1 \\), so \\( 2^{121} \\equiv 2 \\pmod{11} \\).  \nBut for modulus 121, need more:  \nActually, \\( 2^{110} \\equiv 1 \\pmod{121} \\) by Euler’s theorem? Not exactly, but known: 121 is not a pseudoprime base 2. Quick check:  \n\\( 2^{121} - 2 \\) divisible by 121? Probably not. Let’s skip detailed check; known fact is 341 is smallest.\n\n---\n\n**Step 11: Verify no smaller composite works**\n\nFrom known sequence A001567:  \nFirst few base-2 Fermat pseudoprimes: 341, 561, 645, 1105, …  \nSo smallest is 341.\n\n---\n\n**Step 12: Conclusion**\n\nThe smallest composite \\( n > 1 \\) with \\( 2^n \\equiv 2 \\pmod{n} \\) is \\( n = 341 \\).\n\n---\n\n**Answer:** 341",
-      "turns": 1,
-      "cacheHitRatio": 0,
-      "costUsd": 0.0018940500000000002,
-      "claudeEquivalentUsd": 0.02577,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "pseudoprime_base2",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The smallest composite integer \\(n > 1\\) satisfying \\(2^n \\equiv 2 \\pmod{n}\\) is \\(n = 341\\). This is composite (\\(341 = 11 \\times 31\\)), and it satisfies the congruence because \\(2^{341} \\equiv 2 \\pmod{11}\\) and \\(2^{341} \\equiv 2 \\pmod{31}\\) by Fermat's little theorem, implying \\(2^{341} \\equiv 2 \\pmod{341}\\). All smaller composite integers (e.g., 4, 6, 8, 9, 10, 15, 21) fail to satisfy the condition.\n\nAnswer: 341",
-      "turns": 1,
-      "cacheHitRatio": 0.7529411764705882,
-      "costUsd": 0.0031281200000000003,
-      "claudeEquivalentUsd": 0.02154,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "pseudoprime_base2",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "fail",
-      "checkNote": "This operation was aborted",
-      "finalAgentMessage": "",
-      "turns": 0,
-      "cacheHitRatio": 0,
-      "costUsd": 0,
-      "claudeEquivalentUsd": 0,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0,
-      "errorMessage": "This operation was aborted"
-    },
-    {
-      "taskId": "derangements_d7",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Alright, let's go step-by-step.\n\n---\n\n**Step 1: Understanding the problem**\n\nWe want the number of permutations of \\(\\{1, 2, 3, 4, 5, 6, 7\\}\\) with **no fixed points**.  \nA fixed point means \\(\\pi(i) = i\\) for some \\(i\\).  \nA permutation with no fixed points is called a **derangement**.  \nThe number of derangements of \\(n\\) elements is denoted \\(D_n\\) or \\(!n\\).\n\n---\n\n**Step 2: Formula for derangements**\n\nThe formula for \\(D_n\\) is:\n\n\\[\nD_n = n! \\sum_{k=0}^{n} \\frac{(-1)^k}{k!}\n\\]\n\nFor \\(n=7\\):\n\n\\[\nD_7 = 7! \\left( \\frac{1}{0!} - \\frac{1}{1!} + \\frac{1}{2!} - \\frac{1}{3!} + \\frac{1}{4!} - \\frac{1}{5!} + \\frac{1}{6!} - \\frac{1}{7!} \\right)\n\\]\n\n---\n\n**Step 3: Compute step-by-step**\n\nFirst, \\(7! = 5040\\).\n\nNow compute the sum:\n\n\\[\n\\sum_{k=0}^{7} \\frac{(-1)^k}{k!} \n= 1 - 1 + \\frac{1}{2} - \\frac{1}{6} + \\frac{1}{24} - \\frac{1}{120} + \\frac{1}{720} - \\frac{1}{5040}\n\\]\n\nLet's compute term-by-term:\n\n- \\(1 - 1 = 0\\)\n- \\(0 + 0.5 = 0.5\\)\n- \\(0.5 - \\frac{1}{6} \\approx 0.5 - 0.1666667 = 0.3333333\\)\n- \\(0.3333333 + \\frac{1}{24} \\approx 0.3333333 + 0.0416667 = 0.375\\)\n- \\(0.375 - \\frac{1}{120} \\approx 0.375 - 0.0083333 = 0.3666667\\)\n- \\(0.3666667 + \\frac{1}{720} \\approx 0.3666667 + 0.0013889 = 0.3680556\\)\n- \\(0.3680556 - \\frac{1}{5040} \\approx 0.3680556 - 0.0001984 = 0.3678572\\)\n\nSo the sum is approximately \\(0.367857142857...\\)\n\n---\n\n**Step 4: Multiply by \\(7!\\)**\n\n\\[\nD_7 = 5040 \\times 0.367857142857\\ldots\n\\]\n\nLet's compute exactly using fractions:\n\n\\[\n\\sum_{k=0}^{7} \\frac{(-1)^k}{k!} \n= \\frac{1}{2!} - \\frac{1}{3!} + \\frac{1}{4!} - \\frac{1}{5!} + \\frac{1}{6!} - \\frac{1}{7!}\n\\]\n\nWrite with common denominator \\(7! = 5040\\):\n\n- \\(\\frac{1}{2!} = \\frac{2520}{5040}\\)\n- \\(\\frac{1}{3!} = \\frac{840}{5040}\\)\n- \\(\\frac{1}{4!} = \\frac{210}{5040}\\)\n- \\(\\frac{1}{5!} = \\frac{42}{5040}\\)\n- \\(\\frac{1}{6!} = \\frac{7}{5040}\\)\n- \\(\\frac{1}{7!} = \\frac{1}{5040}\\)\n\nNow alternate signs:\n\n\\[\n\\left( \\frac{2520}{5040} - \\frac{840}{5040} \\right) \n= \\frac{1680}{5040}\n\\]\n\\[\n\\frac{1680}{5040} + \\frac{210}{5040} = \\frac{1890}{5040}\n\\]\n\\[\n\\frac{1890}{5040} - \\frac{42}{5040} = \\frac{1848}{5040}\n\\]\n\\[\n\\frac{1848}{5040} + \\frac{7}{5040} = \\frac{1855}{5040}\n\\]\n\\[\n\\frac{1855}{5040} - \\frac{1}{5040} = \\frac{1854}{5040}\n\\]\n\nSo the sum is \\(\\frac{1854}{5040}\\).\n\n---\n\n**Step 5: Multiply by \\(7!\\)**\n\n\\[\nD_7 = 5040 \\times \\frac{1854}{5040} = 1854\n\\]\n\n---\n\n**Step 6: Quick check with known derangement numbers**\n\nKnown small derangements:  \n\\(D_1 = 0, D_2 = 1, D_3 = 2, D_4 = 9, D_5 = 44, D_6 = 265, D_7 = 1854\\).  \nYes, matches known sequence.\n\n---\n\n**Final answer:**\n\nAnswer: 1854",
-      "turns": 1,
-      "cacheHitRatio": 0,
-      "costUsd": 0.00127253,
-      "claudeEquivalentUsd": 0.017292,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "derangements_d7",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The number of derangements of \\(n\\) distinct objects, denoted \\(D_n\\), counts permutations with no fixed points. For \\(n = 7\\), this can be computed using the recurrence \\(D_n = (n-1)(D_{n-1} + D_{n-2})\\) with initial values \\(D_1 = 0\\) and \\(D_2 = 1\\).  \nThus:  \n\\(D_3 = 2(1+0) = 2\\),  \n\\(D_4 = 3(2+1) = 9\\),  \n\\(D_5 = 4(9+2) = 44\\),  \n\\(D_6 = 5(44+9) = 265\\),  \n\\(D_7 = 6(265+44) = 1854\\).\n\nAnswer: 1854",
-      "turns": 1,
-      "cacheHitRatio": 0.7191011235955056,
-      "costUsd": 0.00349167,
-      "claudeEquivalentUsd": 0.024027,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "derangements_d7",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The number of derangements of \\(n\\) elements, denoted \\(D_n\\), counts permutations with no fixed points. For \\(n=7\\), this can be computed using the recurrence \\(D_n = (n-1)(D_{n-1} + D_{n-2})\\) with initial values \\(D_0 = 1\\) and \\(D_1 = 0\\). Applying the recurrence yields \\(D_2 = 1\\), \\(D_3 = 2\\), \\(D_4 = 9\\), \\(D_5 = 44\\), \\(D_6 = 265\\), and finally \\(D_7 = 6 \\times (265 + 44) = 1854\\).\n\nAnswer: 1854",
-      "turns": 1,
-      "cacheHitRatio": 0.7191011235955056,
-      "costUsd": 0.0028434299999999997,
-      "claudeEquivalentUsd": 0.019587,
-      "harvestedTurns": 1,
-      "totalSubgoals": 3,
-      "totalUncertainties": 2,
-      "totalHypotheses": 2,
-      "totalRejectedPaths": 2
-    },
-    {
-      "taskId": "euler_quadratic_break",
-      "mode": "baseline",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "Let’s check \\( f(n) = n^2 + n + 41 \\) for small \\( n \\):\n\n- \\( n = 0 \\): \\( f(0) = 41 \\) (prime)  \n- \\( n = 1 \\): \\( f(1) = 1 + 1 + 41 = 43 \\) (prime)  \n- \\( n = 2 \\): \\( f(2) = 4 + 2 + 41 = 47 \\) (prime)  \n- \\( n = 3 \\): \\( f(3) = 9 + 3 + 41 = 53 \\) (prime)  \n- \\( n = 4 \\): \\( f(4) = 16 + 4 + 41 = 61 \\) (prime)  \n- \\( n = 5 \\): \\( f(5) = 25 + 5 + 41 = 71 \\) (prime)  \n- \\( n = 6 \\): \\( f(6) = 36 + 6 + 41 = 83 \\) (prime)  \n- \\( n = 7 \\): \\( f(7) = 49 + 7 + 41 = 97 \\) (prime)  \n- \\( n = 8 \\): \\( f(8) = 64 + 8 + 41 = 113 \\) (prime)  \n- \\( n = 9 \\): \\( f(9) = 81 + 9 + 41 = 131 \\) (prime)  \n- \\( n = 10 \\): \\( f(10) = 100 + 10 + 41 = 151 \\) (prime)  \n\nWe can keep going, but the first known counterexample for this famous Euler polynomial is \\( n = 40 \\), because  \n\\[\nf(40) = 40^2 + 40 + 41 = 1600 + 40 + 41 = 1681.\n\\]  \nCheck if \\( 1681 \\) is prime:  \n\\( 1681 = 41^2 \\) (since \\( 41 \\times 41 = 1681 \\)), so it’s composite.\n\nBut wait — let’s check if there’s a smaller \\( n \\) where \\( f(n) \\) is composite.  \nWe can test \\( n = 41 \\):  \n\\( f(41) = 41^2 + 41 + 41 = 1681 + 82 = 1763 \\), which is \\( 41 \\times 43 \\) (since \\( 1763/41 = 43 \\)), so composite. But \\( n = 41 \\) is larger than \\( n = 40 \\).\n\nCheck \\( n = 39 \\): \\( f(39) = 1521 + 39 + 41 = 1601 \\). Test divisibility: not divisible by primes up to \\( \\sqrt{1601} \\approx 40 \\), seems prime.  \n\nCheck \\( n = 40 \\) is indeed the first composite? Let’s check \\( n = 1 \\) to \\( 39 \\) quickly for a smaller counterexample:  \nWe can note \\( f(n) = n(n+1) + 41 \\). If \\( 41 \\mid f(n) \\), then \\( 41 \\mid n(n+1) \\).  \nThat happens when \\( n \\equiv 0 \\pmod{41} \\) or \\( n \\equiv 40 \\pmod{41} \\).  \nThe smallest positive such \\( n \\) is \\( n = 40 \\) (since \\( n=41 \\) is larger).  \nSo for \\( n = 40 \\), \\( f(40) = 40 \\times 41 + 41 = 41 \\times 41 \\) is composite.  \n\nCheck \\( n = 0 \\) to \\( 39 \\): no multiple of 41 except \\( n=0 \\) (which gives 41, prime). So \\( n=40 \\) is indeed the first nonnegative integer where \\( f(n) \\) is composite.\n\nThus smallest \\( n \\) is \\( 40 \\), \\( f(40) = 1681 = 41^2 \\).\n\n**Answer:** 40",
-      "turns": 1,
-      "cacheHitRatio": 0,
-      "costUsd": 0.00098535,
-      "claudeEquivalentUsd": 0.013365,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "euler_quadratic_break",
-      "mode": "reasoner",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The smallest nonnegative integer \\(n\\) for which \\(f(n) = n^2 + n + 41\\) is composite is \\(n = 40\\). At this value, \\(f(40) = 40^2 + 40 + 41 = 1681 = 41^2\\), which is not prime.\n\nAnswer: 40",
-      "turns": 1,
-      "cacheHitRatio": 0.6095238095238096,
-      "costUsd": 0.0014309199999999998,
-      "claudeEquivalentUsd": 0.0099,
-      "harvestedTurns": 0,
-      "totalSubgoals": 0,
-      "totalUncertainties": 0,
-      "totalHypotheses": 0,
-      "totalRejectedPaths": 0
-    },
-    {
-      "taskId": "euler_quadratic_break",
-      "mode": "reasoner-harvest",
-      "repeat": 1,
-      "verdict": "pass",
-      "finalAgentMessage": "The polynomial \\(f(n) = n^2 + n + 41\\) is known to produce primes for \\(n = 0, 1, 2, \\dots, 39\\). At \\(n = 40\\), we have \\(f(40) = 40^2 + 40 + 41 = 1681\\), and \\(1681 = 41^2\\), which is composite. Therefore, \\(n = 40\\) is the smallest nonnegative integer where \\(f(n)\\) fails to be prime.\n\nAnswer: 40",
-      "turns": 1,
-      "cacheHitRatio": 0.6095238095238096,
-      "costUsd": 0.0025938099999999998,
-      "claudeEquivalentUsd": 0.017865,
-      "harvestedTurns": 1,
-      "totalSubgoals": 3,
-      "totalUncertainties": 2,
-      "totalHypotheses": 3,
-      "totalRejectedPaths": 2
-    }
-  ]
-}

benchmarks/harvest/runner.ts

@@ -1,291 +0,0 @@
-/** Harvest-bench runner — writes results.json. CLI flags + sample invocations in benchmarks/README.md. */
-
-import { type WriteStream, mkdirSync, writeFileSync } from "node:fs";
-import { join } from "node:path";
-import { pathToFileURL } from "node:url";
-import {
-  CacheFirstLoop,
-  DeepSeekClient,
-  ImmutablePrefix,
-  claudeEquivalentCost,
-  loadDotenv,
-} from "../../src/index.js";
-import { openTranscriptFile, recordFromLoopEvent, writeRecord } from "../../src/transcript/log.js";
-import { TASKS } from "./tasks.js";
-import type {
-  HarvestBenchMeta,
-  HarvestBenchReport,
-  HarvestMode,
-  HarvestRunResult,
-  HarvestTask,
-} from "./types.js";
-import { ALL_MODES } from "./types.js";
-
-loadDotenv();
-
-interface CliArgs {
-  taskFilter: string | null;
-  modes: HarvestMode[];
-  repeats: number;
-  outPath: string | null;
-  transcriptsDir: string | null;
-  /** Per-HTTP-call timeout in ms. Default 300_000 (5 min) — reasoner + harvest runs legitimately need this. */
-  timeoutMs: number;
-  dry: boolean;
-  verbose: boolean;
-}
-
-function parseArgs(argv: string[]): CliArgs {
-  const out: CliArgs = {
-    taskFilter: null,
-    modes: [...ALL_MODES],
-    repeats: 1,
-    outPath: null,
-    transcriptsDir: null,
-    timeoutMs: 300_000,
-    dry: false,
-    verbose: false,
-  };
-  for (let i = 0; i < argv.length; i++) {
-    const a = argv[i];
-    if (a === "--task") out.taskFilter = argv[++i] ?? null;
-    else if (a === "--mode") {
-      const v = (argv[++i] ?? "").toLowerCase() as HarvestMode;
-      if ((ALL_MODES as string[]).includes(v)) out.modes = [v];
-    } else if (a === "--repeats") out.repeats = Number.parseInt(argv[++i] ?? "1", 10);
-    else if (a === "--out") out.outPath = argv[++i] ?? null;
-    else if (a === "--transcripts-dir") out.transcriptsDir = argv[++i] ?? null;
-    else if (a === "--timeout") {
-      // Accept seconds for humans, convert to ms internally.
-      const secs = Number.parseInt(argv[++i] ?? "300", 10);
-      if (Number.isFinite(secs) && secs > 0) out.timeoutMs = secs * 1000;
-    } else if (a === "--dry") out.dry = true;
-    else if (a === "--verbose" || a === "-v") out.verbose = true;
-  }
-  return out;
-}
-
-function modelForMode(mode: HarvestMode): string {
-  if (mode === "baseline") return "deepseek-chat";
-  return "deepseek-reasoner";
-}
-
-function harvestForMode(mode: HarvestMode): boolean {
-  return mode === "reasoner-harvest";
-}
-
-async function runOnce(
-  task: HarvestTask,
-  mode: HarvestMode,
-  rep: number,
-  client: DeepSeekClient,
-  transcriptStream: WriteStream | null,
-): Promise<HarvestRunResult> {
-  const model = modelForMode(mode);
-  const harvest = harvestForMode(mode);
-
-  const prefix = new ImmutablePrefix({ system: task.systemPrompt });
-  const loop = new CacheFirstLoop({
-    client,
-    prefix,
-    model,
-    harvest,
-    stream: false,
-  });
-  const prefixHash = prefix.fingerprint;
-
-  let harvestedTurns = 0;
-  let totalSubgoals = 0;
-  let totalUncertainties = 0;
-  let totalHypotheses = 0;
-  let totalRejectedPaths = 0;
-  let finalText = "";
-  let errorMessage: string | undefined;
-
-  if (transcriptStream) {
-    writeRecord(transcriptStream, {
-      ts: new Date().toISOString(),
-      turn: 1,
-      role: "user",
-      content: task.prompt,
-    });
-  }
-
-  try {
-    for await (const ev of loop.step(task.prompt)) {
-      if (transcriptStream && ev.role !== "assistant_delta") {
-        writeRecord(transcriptStream, recordFromLoopEvent(ev, { model, prefixHash }));
-      }
-      if (ev.role === "assistant_final") {
-        finalText = ev.content;
-        if (ev.planState) {
-          if (
-            ev.planState.subgoals.length +
-              ev.planState.hypotheses.length +
-              ev.planState.uncertainties.length +
-              ev.planState.rejectedPaths.length >
-            0
-          ) {
-            harvestedTurns++;
-            totalSubgoals += ev.planState.subgoals.length;
-            totalHypotheses += ev.planState.hypotheses.length;
-            totalUncertainties += ev.planState.uncertainties.length;
-            totalRejectedPaths += ev.planState.rejectedPaths.length;
-          }
-        }
-      } else if (ev.role === "done") {
-        finalText = ev.content || finalText;
-        break;
-      } else if (ev.role === "error") {
-        throw new Error(ev.error ?? "loop error");
-      }
-    }
-  } catch (err) {
-    errorMessage = (err as Error).message;
-  }
-
-  const verdict = errorMessage
-    ? { verdict: "fail" as const, note: errorMessage }
-    : task.check(finalText);
-
-  return {
-    taskId: task.id,
-    mode,
-    repeat: rep + 1,
-    verdict: verdict.verdict,
-    checkNote: verdict.note,
-    finalAgentMessage: finalText,
-    turns: loop.stats.turns.length,
-    cacheHitRatio: loop.stats.aggregateCacheHitRatio,
-    costUsd: loop.stats.totalCost,
-    claudeEquivalentUsd: loop.stats.turns.reduce((s, t) => s + claudeEquivalentCost(t.usage), 0),
-    harvestedTurns,
-    totalSubgoals,
-    totalUncertainties,
-    totalHypotheses,
-    totalRejectedPaths,
-    errorMessage,
-  };
-}
-
-function runDry(args: CliArgs): HarvestBenchReport {
-  const tasks = filterTasks(args.taskFilter);
-  const results: HarvestRunResult[] = [];
-  for (const task of tasks) {
-    // Exercise each checker with a deliberately wrong reply so we know
-    // the wiring is live — not a pass on empty.
-    const sanity = task.check("Answer: this is not the right answer at all");
-    console.log(`[${task.id}] checker exercises ok (sanity verdict=${sanity.verdict})`);
-    for (const mode of args.modes) {
-      results.push({
-        taskId: task.id,
-        mode,
-        repeat: 1,
-        verdict: "inconclusive",
-        checkNote: "dry-run (no model call)",
-        finalAgentMessage: "[dry-run]",
-        turns: 0,
-        cacheHitRatio: 0,
-        costUsd: 0,
-        claudeEquivalentUsd: 0,
-        harvestedTurns: 0,
-        totalSubgoals: 0,
-        totalUncertainties: 0,
-        totalHypotheses: 0,
-        totalRejectedPaths: 0,
-      });
-    }
-  }
-  return { meta: buildMeta(args, tasks.length), results };
-}
-
-function filterTasks(filter: string | null): HarvestTask[] {
-  if (!filter) return TASKS;
-  const t = TASKS.find((x) => x.id === filter);
-  if (!t) throw new Error(`unknown task: ${filter}`);
-  return [t];
-}
-
-function buildMeta(args: CliArgs, taskCount: number): HarvestBenchMeta {
-  return {
-    date: new Date().toISOString(),
-    repeatsPerTask: args.repeats,
-    taskCount,
-    modesRun: args.modes,
-    reasonixVersion: "0.2.2",
-  };
-}
-
-async function main(): Promise<void> {
-  const args = parseArgs(process.argv.slice(2));
-
-  if (args.dry) {
-    const report = runDry(args);
-    writeReport(report, args.outPath);
-    return;
-  }
-
-  const client = new DeepSeekClient({ timeoutMs: args.timeoutMs });
-  const tasks = filterTasks(args.taskFilter);
-
-  if (args.transcriptsDir) mkdirSync(args.transcriptsDir, { recursive: true });
-
-  const results: HarvestRunResult[] = [];
-  for (const task of tasks) {
-    for (let rep = 0; rep < args.repeats; rep++) {
-      for (const mode of args.modes) {
-        let stream: WriteStream | null = null;
-        if (args.transcriptsDir) {
-          const fname = `${task.id}.${mode}.r${rep + 1}.jsonl`;
-          stream = openTranscriptFile(join(args.transcriptsDir, fname), {
-            version: 1,
-            source: `harvest-bench/${mode}`,
-            model: modelForMode(mode),
-            task: task.id,
-            mode,
-            repeat: rep + 1,
-            startedAt: new Date().toISOString(),
-          });
-        }
-        const started = Date.now();
-        let result: HarvestRunResult;
-        try {
-          result = await runOnce(task, mode, rep, client, stream);
-        } finally {
-          stream?.end();
-        }
-        const elapsed = ((Date.now() - started) / 1000).toFixed(1);
-        const note = result.checkNote ? `  ${result.checkNote}` : "";
-        console.log(
-          `[${task.id}/${mode}/r${rep + 1}] verdict=${result.verdict} cache=${(result.cacheHitRatio * 100).toFixed(1)}% cost=$${result.costUsd.toFixed(6)} harvest=${result.harvestedTurns}:${result.totalSubgoals}s/${result.totalUncertainties}u (${elapsed}s)${note}`,
-        );
-        results.push(result);
-      }
-    }
-  }
-
-  const report: HarvestBenchReport = { meta: buildMeta(args, tasks.length), results };
-  writeReport(report, args.outPath);
-}
-
-function writeReport(report: HarvestBenchReport, outPath: string | null): void {
-  const path =
-    outPath ?? `benchmarks/harvest/results-${new Date().toISOString().replace(/[:.]/g, "-")}.json`;
-  writeFileSync(path, `${JSON.stringify(report, null, 2)}\n`, "utf8");
-  console.log(`wrote ${path}`);
-}
-
-function isMain(): boolean {
-  const entry = process.argv[1];
-  if (!entry) return false;
-  return import.meta.url === pathToFileURL(entry).href;
-}
-
-if (isMain()) {
-  main().catch((err) => {
-    console.error(err);
-    process.exit(1);
-  });
-}
-
-export { main as runHarvestBench };

benchmarks/harvest/tasks.ts

@@ -1,143 +0,0 @@
-/** Seed reasoning tasks for harvest-bench — single-turn Q/A with deterministic checkers. */
-
-import type { HarvestTask } from "./types.js";
-
-const CONCISE_SYSTEM = `You are a precise reasoner. Think carefully, then state your final answer on a clearly-labeled line starting with "Answer:".`;
-
-/** Parse non-negative integers; expands "2-4" → 2,3,4; strips LaTeX decorations first. */
-function extractIntegers(text: string): number[] {
-  // Strip LaTeX delimiters so n_1 / n^2 / \{ ... \} don't confuse us.
-  const cleaned = text
-    .replace(/\\[a-zA-Z]+/g, " ") // \frac, \mod etc.
-    .replace(/[{}\\]/g, " ")
-    .replace(/[_^]\d+/g, " "); // subscript/superscript numbers
-  // Expand ranges: "2-4" → "2,3,4" (only for small gaps)
-  const rangeExpanded = cleaned.replace(/(\d+)\s*[-–]\s*(\d+)/g, (_, a, b) => {
-    const lo = Number.parseInt(a, 10);
-    const hi = Number.parseInt(b, 10);
-    if (!Number.isFinite(lo) || !Number.isFinite(hi) || hi - lo > 50) return `${a} ${b}`;
-    const out: number[] = [];
-    for (let i = Math.min(lo, hi); i <= Math.max(lo, hi); i++) out.push(i);
-    return out.join(" ");
-  });
-  const matches = rangeExpanded.match(/-?\d+/g) ?? [];
-  return matches.map((m) => Number.parseInt(m, 10)).filter((n) => Number.isFinite(n));
-}
-
-/** Extract content after the last "Answer:" label (case-insensitive). */
-function extractAnswerLine(text: string): string {
-  const matches = [...text.matchAll(/answer\s*[:：]\s*([^\n]+)/gi)];
-  if (matches.length === 0) return text;
-  return matches[matches.length - 1]![1]!.trim();
-}
-
-/** Positive n ≤ 100 such that n^2 + n + 1 ≡ 0 (mod 7). */
-const MOD7_EXPECTED: number[] = (() => {
-  const out: number[] = [];
-  for (let n = 1; n <= 100; n++) {
-    if ((n * n + n + 1) % 7 === 0) out.push(n);
-  }
-  return out;
-})();
-
-/** Single-integer checker — first plausible int from the last "Answer:" line, exact equality. */
-function checkSingleInt(expected: number) {
-  return (reply: string) => {
-    const answerLine = extractAnswerLine(reply);
-    const nums = extractIntegers(answerLine);
-    if (nums.length === 0) return { verdict: "fail" as const, note: "no number in Answer line" };
-    const n = nums[0]!;
-    if (n === expected) return { verdict: "pass" as const };
-    return { verdict: "fail" as const, note: `got ${n}, expected ${expected}` };
-  };
-}
-
-export const TASKS: HarvestTask[] = [
-  {
-    id: "mod7_list",
-    description:
-      "Number theory — find all positive n ≤ 100 with n^2+n+1 ≡ 0 (mod 7). 29-element set. Classic R1 scratch-work territory.",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'Find all positive integers n with n ≤ 100 such that n^2 + n + 1 is divisible by 7. Briefly justify, then end with the full list on a line starting with "Answer:".',
-    check: (reply) => {
-      const answerLine = extractAnswerLine(reply);
-      const found = new Set(extractIntegers(answerLine).filter((n) => n >= 1 && n <= 100));
-      const expected = new Set(MOD7_EXPECTED);
-      if (found.size === 0) {
-        return { verdict: "fail", note: "no integers extracted from Answer line" };
-      }
-      // Exact set equality — both containment directions.
-      for (const n of expected)
-        if (!found.has(n)) return { verdict: "fail", note: `missing n=${n}` };
-      for (const n of found) if (!expected.has(n)) return { verdict: "fail", note: `extra n=${n}` };
-      return { verdict: "pass" };
-    },
-  },
-
-  {
-    id: "flips_until_3heads",
-    description:
-      "Probability — expected coin flips until 3 consecutive heads. Answer: 14. Tests whether R1 derives via recurrence or via memorized result.",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'A fair coin is flipped repeatedly until you see 3 heads in a row. What is the expected number of total flips? Give a single integer on a line starting with "Answer:".',
-    check: (reply) => {
-      const answerLine = extractAnswerLine(reply);
-      const nums = extractIntegers(answerLine);
-      if (nums.length === 0) return { verdict: "fail", note: "no number in Answer line" };
-      // Accept the first sensible number — sometimes agents write "Answer: 14 flips".
-      const n = nums[0]!;
-      if (n === 14) return { verdict: "pass" };
-      return { verdict: "fail", note: `got ${n}, expected 14` };
-    },
-  },
-
-  {
-    id: "three_hats",
-    description:
-      "Logic — three people wear red/blue hats, at least one red, they're asked in order. First two say no, third says yes. What color? Answer: red (deducible because if third were blue, the first two could see one red + one unknown and wouldn't have said no on their full first round, for most configurations the third can only be certain if red).",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'Three people wear hats that are each red or blue. They are told at least one hat is red. Each can see the other two hats but not their own. They are asked in order: "Do you know your hat color?" The first two answer no. The third answers yes. What color is the third person\'s hat? Give a one-word answer on a line starting with "Answer:".',
-    check: (reply) => {
-      const answerLine = extractAnswerLine(reply).toLowerCase();
-      // Accept "red" as a standalone word (not "redirected" etc).
-      if (/\bred\b/.test(answerLine) && !/\bblue\b/.test(answerLine)) {
-        return { verdict: "pass" };
-      }
-      if (/\bblue\b/.test(answerLine)) return { verdict: "fail", note: "answered blue" };
-      return { verdict: "fail", note: "no clear color in Answer line" };
-    },
-  },
-
-  {
-    id: "pseudoprime_base2",
-    description:
-      "Number theory — smallest composite n > 1 such that 2^n ≡ 2 (mod n). This is the smallest Fermat pseudoprime to base 2. Answer: 341 (= 11·31). V3 frequently answers 561 (Carmichael / wrong base-2 property) or 9 (not actually a pseudoprime to base 2, verifying the model isn't computing). R1 usually gets this right.",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'Find the smallest composite integer n > 1 such that 2^n is congruent to 2 modulo n (i.e., n is a Fermat pseudoprime to base 2). Briefly justify, then give a single integer on a line starting with "Answer:".',
-    check: checkSingleInt(341),
-  },
-
-  {
-    id: "derangements_d7",
-    description:
-      "Combinatorics — number of derangements of 7 elements. Answer: 1854 (D_7 = 7! · (1 − 1/1! + 1/2! − 1/3! + … ± 1/7!) = 5040 · 0.367857… = 1854). V3 sometimes approximates via n!/e and rounds wrong, or confuses with D_6=265 or D_8=14833. R1's explicit recurrence D_n = (n-1)(D_{n-1}+D_{n-2}) avoids the mistake.",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'How many permutations of {1, 2, 3, 4, 5, 6, 7} have no fixed points? (This is D_7, the derangement number.) Briefly justify, then give a single integer on a line starting with "Answer:".',
-    check: checkSingleInt(1854),
-  },
-
-  {
-    id: "euler_quadratic_break",
-    description:
-      "Number theory — smallest nonneg integer n where n^2 + n + 41 is NOT prime. Euler's prime-generating polynomial famously produces primes for n = 0..39 and first fails at n = 40 (40²+40+41 = 1681 = 41²). V3 often confuses this with the related n²-n+41 (first fails at 41) or cites the wrong threshold. R1 checks small n systematically.",
-    systemPrompt: CONCISE_SYSTEM,
-    prompt:
-      'Consider the polynomial f(n) = n^2 + n + 41 for n = 0, 1, 2, …. What is the smallest nonneg integer n at which f(n) fails to be prime? Briefly justify (including the value of f(n) at that point), then give just the integer n on a line starting with "Answer:".',
-    check: checkSingleInt(40),
-  },
-];

benchmarks/harvest/types.ts

@@ -1,61 +0,0 @@
-/** Harvest-bench types — single-turn Q/A with deterministic checkers; isolates Pillar 2 the way τ-bench-lite isolates Pillar 1. */
-
-export type HarvestMode =
-  /** deepseek-chat, no Reasonix tricks. Floor reference for answer quality on reasoning tasks (V3 at its best). */
-  | "baseline"
-  /** deepseek-reasoner, no harvest. Isolates "does reasoner help over chat on these problems?" */
-  | "reasoner"
-  /** deepseek-reasoner + harvest. Isolates "does the extra V3 harvest call add quality / change behavior?" */
-  | "reasoner-harvest";
-
-export const ALL_MODES: HarvestMode[] = ["baseline", "reasoner", "reasoner-harvest"];
-
-/** Checker verdict — `inconclusive` reserves room for "right shape, can't verify without running it". */
-export type CheckVerdict = "pass" | "fail" | "inconclusive";
-
-export interface HarvestTask {
-  id: string;
-  /** One-line human description, not shown to the model. */
-  description: string;
-  /** System prompt. Kept minimal so harvest signal can be attributed to the model, not to prompt engineering. */
-  systemPrompt: string;
-  /** The single question. */
-  prompt: string;
-  /** Pure + deterministic — no LLM calls, no file I/O. Verdict + optional one-line note. */
-  check: (agentReply: string) => { verdict: CheckVerdict; note?: string };
-}
-
-export interface HarvestRunResult {
-  taskId: string;
-  mode: HarvestMode;
-  repeat: number;
-  verdict: CheckVerdict;
-  checkNote?: string;
-  finalAgentMessage: string;
-  /** Turns the loop completed (usually 1 for a single-shot Q/A, > 1 if tool-calls happen). */
-  turns: number;
-  cacheHitRatio: number;
-  costUsd: number;
-  claudeEquivalentUsd: number;
-  /** Harvest-specific — only non-zero on reasoner-harvest mode. */
-  harvestedTurns: number;
-  totalSubgoals: number;
-  totalUncertainties: number;
-  totalHypotheses: number;
-  totalRejectedPaths: number;
-  /** Optional error message if the run crashed (model down, timeout, etc). */
-  errorMessage?: string;
-}
-
-export interface HarvestBenchMeta {
-  date: string;
-  repeatsPerTask: number;
-  taskCount: number;
-  modesRun: HarvestMode[];
-  reasonixVersion: string;
-}
-
-export interface HarvestBenchReport {
-  meta: HarvestBenchMeta;
-  results: HarvestRunResult[];
-}

benchmarks/real-world-cache/README.md

@@ -0,0 +1,76 @@
+# Real-world cache hit — single user, single day
+
+A real Reasonix user shared their DeepSeek dashboard for **2026-05-01**.
+Used with permission, anonymized.
+
+![DeepSeek usage dashboard, 2026-05-01](2026-05-01-deepseek-dashboard.png)
+
+## The numbers
+
+| | Tokens |
+|---|---:|
+| Input — cache hit | 435,033,856 |
+| Input — cache miss | 767,616 |
+| Output | 179,763 |
+| **Day total** | **435,981,235** |
+
+**Cache hit ratio (input):**
+`435,033,856 / (435,033,856 + 767,616)` = **99.82%**
+
+## Cost — using the prices Reasonix bills against (`src/telemetry/stats.ts`)
+
+USD per 1M tokens — `inputCacheHit / inputCacheMiss / output`:
+- `deepseek-v4-flash` — `0.028 / 0.139 / 0.278`
+- `deepseek-v4-pro` — `0.139 / 1.667 / 3.333`
+
+Assuming **v4-flash** (the project default):
+
+| | This user (99.82% hit) | Same workload, **0% cache** |
+|---|---:|---:|
+| Cache-hit input | $12.18 | — |
+| Cache-miss input | $0.11 | $60.58 |
+| Output | $0.05 | $0.05 |
+| **Total / day** | **$12.34** | **$60.63** |
+
+→ Cache saved this user **$48.29**, or **~80%** off the un-cached baseline, on a single day.
+
+On **v4-pro** (5× the prefix-cache discount) the same workload would cost
+**~$62.35** vs **~$727.08** without cache — a **~91% saving**.
+
+## "Isn't that just DeepSeek's prefix cache?"
+
+DeepSeek's API ships prefix caching enabled by default; the *cache* is theirs,
+the *hit rate* is the client's. Same API, different clients, very different
+hit rates:
+
+- DeepSeek's own web chat: 60–80% within a single conversation, drops to 0%
+  on a new session (system prompt may differ).
+- Cherry Studio / Open WebUI / generic OpenAI-shape SDKs: typically 30–60%
+  on long sessions — history gets reordered, tool specs get re-serialized,
+  every drift breaks the prefix.
+- Cline / Continue and other XML-tool-call clients: lower still — every tool
+  result inlines into the conversation, shifting bytes the cache keys on.
+
+99.82% is what falls out of these four design choices in Reasonix:
+
+1. **`ImmutablePrefix`** (`src/memory.ts`) — system prompt + tool specs are
+   frozen at session start. Same byte sequence every turn.
+2. **`AppendOnlyLog`** — turns only append. No reorder, no edit-in-place.
+3. **`VolatileScratch`** — chain-of-thought / per-turn scratch lives outside
+   the cached prefix so it never poisons the next hit.
+4. **Auto-compact** — when context approaches the cap, older turns fold into
+   a summary message *appended* to the prefix; the prefix itself isn't
+   rewritten, so the cache survives the fold.
+
+DeepSeek gave us cacheable bytes. The four mechanisms above are how we keep
+the bytes cacheable.
+
+## Reproduce
+
+The synthetic side of this lives in `benchmarks/tau-bench/` — same task set
+run through `CacheFirstLoop` vs a deliberately cache-hostile baseline. The
+real-world data above is what the synthetic numbers look like once a user
+runs the harness in anger.
+
+Submit your own dashboard screenshot if you want it anonymized and added
+here — open an issue.

benchmarks/spike-tdd-kernel/tdd-eval.mjs

@@ -45,7 +45,7 @@ const PROMPTS = [
   { id: "m3", level: "medium", task: "A function \`debounceAsync<T extends any[], R>(fn: (...args: T) => Promise<R>, ms: number): (...args: T) => Promise<R>\` in src/util/debounce.ts. Resolves only the latest call's promise; earlier callers reject with an AbortError-like." },
 
   // hard (2) — these touch domain types from the repo
-  { id: "h1", level: "hard", task: "A function \`extractTestId(file: string, fullName: string, source: string): { id: string, source: 'native' | 'annotation' }\` in src/repair/test-id.ts. If \`source\` contains a '// @reasonix-test-id: <slug>' comment within 3 lines above an it()/test() whose name matches \`fullName\`, return that slug with source='annotation'. Otherwise return \`\${file}::\${fullName}\` with source='native'." },
+  { id: "h1", level: "hard", task: "A function \`extractTestId(file: string, fullName: string, source: string): { id: string, source: 'native' | 'annotation' }\` in src/repair/test-id.ts. If \`source\` contains a '// @reasonix-test-id: <slug>' comment within 3 lines above an it()/test() whose name matches \`fullName\`, return that slug with source='annotation'. Otherwise return \`${file}::${fullName}\` with source='native'." },
   { id: "h2", level: "hard", task: "A function \`pairRedGreen(events: Array<{type:string, test_id?:string, status?:string, ts:number}>): Array<{ test_id: string, red_ts: number, green_ts: number }>\` in src/events/pair.ts. For each test_id, find the most recent fail→pass transition and return one entry per test_id. Ignore test_ids that never went green." },
 ];
 

dashboard/PARITY.md

@@ -0,0 +1,139 @@
+# Dashboard ↔ TUI parity
+
+Closing audit of #369. Inventories every slash command and how the
+web dashboard handles it.
+
+## Coverage legend
+
+- **panel** — has a dedicated SPA panel or modal beyond just typing the
+  slash in the chat box.
+- **chat-box** — works by typing the slash into the web chat input.
+  Result text shows in the dashboard scrollback as an info row. No
+  dedicated UI; the chat box _is_ the UI.
+- **tui-only** — keyboard binding or process-controlling action that
+  has no useful web equivalent. Stays a TUI affordance by design.
+
+## Counts
+
+Roughly **20 commands have a dedicated panel**, **38 work via the chat
+box**, and **`/exit` plus 4 raw keyboard shortcuts** stay TUI-only.
+Aliases (e.g. `/sessions` and `/resume` share one picker) collapse
+into one row.
+
+## Core / observability
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/help` | command reference | chat-box |
+| `/keys` | keyboard shortcuts + prompt prefixes | chat-box |
+| `/status` | model + flags + ctx + session | chat-box |
+| `/context` | context-window breakdown (stacked bar) | chat-box · `ctxBreakdown` payload |
+| `/cost` | last turn / next turn estimate | chat-box · usage card |
+| `/stats` | cross-session cost dashboard | **panel** · Usage tab |
+| `/think` | last R1 reasoning dump | chat-box |
+| `/tool [N]` | dump full output of Nth tool call | chat-box |
+| `/clear` | wipe visible scrollback | chat-box |
+| `/new` (`/reset`) | wipe context + scrollback | chat-box |
+| `/exit` (`/quit`, `/q`) | quit the TUI | **tui-only** |
+| `/stop` | abort current model turn | chat-box |
+| `/retry` | resend last user message | chat-box |
+| `/compact` | fold older turns into summary | chat-box |
+| `/update` | show current vs latest version | chat-box |
+| `/doctor` | health check card | chat-box · doctor card |
+
+## Model & compute
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/preset` | model bundle (auto / flash / pro) | **panel** · Settings → Defaults |
+| `/effort` | reasoning cap (high / max) | **panel** · Settings → Defaults |
+| `/model` | active model | **panel** · Settings → Runtime (D-4 #437) |
+| `/models` | list available models | chat-box |
+| `/pro` | arm v4-pro for next turn | **panel** · Settings → Compute (D-2 #435) |
+| `/budget` | session USD cap | **panel** · Settings → Budget + cockpit tile (D-3 #436) |
+| `/loop` | auto-resubmit on interval | **panel** · Settings → Loop (D-5 #438) |
+
+## Memory & project
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/memory [list / show / forget / clear]` | manage pinned memory | **panel** · Memory tab |
+| `/init` | synthesize baseline REASONIX.md | chat-box |
+| `/semantic` | semantic-search index status | chat-box |
+| `/search-engine` (`/se`) | switch web search backend | chat-box |
+| `/language` (`/lang`) | runtime language | **panel** · Settings → Language |
+
+## Sessions
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/sessions` | list saved sessions | **panel** · SessionPicker modal (C-2 #423) |
+| `/resume` | open a session | **panel** · same picker |
+| `/rename` | rename current session | chat-box |
+| `/forget` | delete current session | chat-box |
+| `/plans` | active + archived plans | **panel** · Plans tab |
+| `/replay [N]` | read-only plan archive | **panel** · Viewer modal (C-5 #427) |
+
+## MCP
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/mcp` (list) | bridged servers + tools | **panel** · MCP tab |
+| `/mcp browse` | marketplace + install | **panel** · MCP marketplace picker (C-4 #426) |
+| `/mcp disable` / `enable` / `reconnect` / `text` | server admin | chat-box |
+| `/resource [uri]` | browse / read MCP resources | chat-box |
+| `/prompt [name]` | browse / fetch MCP prompts | chat-box |
+
+## Permissions & admin
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/permissions [list / add / remove / clear]` | shell allowlist | **panel** · Permissions tab |
+| `/hooks [reload]` | active hooks | **panel** · Hooks tab |
+| `/dashboard [stop]` | embedded dashboard lifecycle | chat-box · *intentional — admin command for the surface you're typing in* |
+
+## Code-mode only
+
+| Slash | Purpose | Coverage |
+|---|---|---|
+| `/init [force]` | scan + synthesize REASONIX.md | chat-box |
+| `/apply [N]` | commit pending edits | chat-box |
+| `/discard [N]` | drop pending edits | chat-box |
+| `/walk` | step through pending edits | **panel** · edit-review modal already covered web pre-#369 |
+| `/undo` | roll back last edit batch | chat-box |
+| `/history` | edit batch list | chat-box |
+| `/show [id]` | dump stored edit diff | chat-box |
+| `/commit "msg"` | git commit | chat-box |
+| `/checkpoint [name / list / forget]` | snapshot touched files | chat-box |
+| `/restore` | roll back to checkpoint | **panel** · CheckpointPicker modal (C-3 #425) |
+| `/plan [on / off]` | read-only plan mode | chat-box |
+| `/apply-plan` | force-approve pending plan | chat-box |
+| `/mode [review / auto / yolo]` | edit gate | **panel** · Chat header pill |
+| `/jobs` | list background jobs | chat-box |
+| `/kill <id>` | stop background job | chat-box |
+| `/logs <id> [lines]` | tail job output | chat-box |
+| `/skill [list / show / new / <name>]` | skill management | **panel** · Skills tab |
+
+## Keyboard / TTY-native
+
+These don't have slashes. They ride alongside the slash surface and
+stay TUI-only:
+
+- `Esc` — abort current model turn (web equivalent: Abort button in chat)
+- `Shift+Tab` — cycle edit mode (web equivalent: mode pill in chat header)
+- `Ctrl-L`, `Ctrl-O`, `space`, `u` — TTY scroll / undo banner / pause hotkeys
+
+## Done
+
+Buckets covered:
+- **C** (#416, closed): pickers — `/sessions`, `/restore`, `/mcp browse`, `/replay` + `/walk` (already)
+- **D** (#428, closed): settings — `/preset`, `/effort`, `/model`, `/pro`, `/budget`, `/loop`, `/language`
+
+Bucket A (text outputs that stay chat-box) and bucket B (structured
+outputs that already have panels) need no further work — the panels
+shipped in C/D plus the long-standing Memory / Permissions / Hooks /
+Skills / Plans / Usage tabs already cover every command whose output
+warranted a dedicated UI.
+
+This file is the source of truth for the audit. Add a row when a
+new slash lands.

dashboard/app.css

@@ -812,6 +812,9 @@ main { padding: 32px 40px 60px 32px; min-width: 0; }
 }
 .chip-f:hover { background: var(--bg-hover); color: var(--fg-1); }
 .chip-f.active { color: var(--c-brand); border-color: var(--c-brand); background: rgba(121,192,255,.08); }
+.chip-f.static { cursor: default; }
+.chip-f.static:hover { background: var(--bg-elev); color: var(--fg-2); }
+.chip-f.static.active:hover { color: var(--c-brand); background: rgba(121,192,255,.08); }
 .chip-f .ct { color: var(--fg-4); font-size: 10px; }
 .chip-f.active .ct { color: var(--c-brand); }
 .chip-f .x { color: var(--fg-4); padding: 0 2px; }
@@ -2016,6 +2019,155 @@ input:focus, textarea:focus, select:focus { border-color: var(--c-brand); }
   max-height: 200px;
 }
 
+.modal-picker-search {
+  width: 100%;
+  background: var(--bg-elev-2);
+  border: 1px solid var(--bd);
+  border-radius: var(--r);
+  color: var(--fg-1);
+  padding: 6px 10px;
+  font-size: 13px;
+  margin-bottom: 8px;
+}
+
+.modal-picker-list {
+  display: flex;
+  flex-direction: column;
+  gap: 2px;
+  max-height: 320px;
+  overflow-y: auto;
+  border: 1px solid var(--bd);
+  border-radius: var(--r);
+  background: var(--bg-elev-2);
+  padding: 4px;
+  margin-bottom: 8px;
+}
+
+.modal-picker-row {
+  display: grid;
+  grid-template-columns: 1fr auto;
+  gap: 4px 8px;
+  align-items: baseline;
+  padding: 6px 8px;
+  background: transparent;
+  border: 1px solid transparent;
+  border-radius: var(--r-sm, 4px);
+  color: var(--fg-1);
+  text-align: left;
+  cursor: pointer;
+}
+
+.modal-picker-row:hover {
+  background: var(--bg-elev-3, var(--bg-elev-2));
+}
+
+.modal-picker-row.selected {
+  border-color: var(--accent, #fcd34d);
+  background: var(--bg-elev-3, var(--bg-elev-2));
+}
+
+.modal-picker-title {
+  grid-column: 1;
+  font-weight: 500;
+}
+
+.modal-picker-badge {
+  grid-column: 2;
+  font-size: 11px;
+  color: var(--fg-2);
+  padding: 1px 6px;
+  border-radius: 999px;
+  background: var(--bg-elev-1, var(--bg-elev-2));
+}
+
+.modal-picker-subtitle {
+  grid-column: 1 / 3;
+  font-size: 12px;
+  color: var(--fg-2);
+}
+
+.modal-picker-meta {
+  grid-column: 1 / 3;
+  font-size: 11px;
+  color: var(--fg-3, var(--fg-2));
+}
+
+.modal-picker-empty {
+  padding: 16px;
+  text-align: center;
+  color: var(--fg-2);
+  font-size: 12px;
+}
+
+.modal-picker-more {
+  align-self: center;
+  margin: 0 auto 8px;
+  font-size: 12px;
+}
+
+.modal-picker-form {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+  margin-bottom: 8px;
+}
+
+.modal-picker-form input {
+  width: 100%;
+  background: var(--bg-elev-2);
+  border: 1px solid var(--bd);
+  border-radius: var(--r);
+  color: var(--fg-1);
+  padding: 6px 10px;
+  font-size: 13px;
+}
+
+.modal-viewer-steps {
+  list-style: none;
+  padding: 0;
+  margin: 0 0 8px;
+  display: flex;
+  flex-direction: column;
+  gap: 4px;
+}
+
+.modal-viewer-step {
+  display: grid;
+  grid-template-columns: 18px 1fr;
+  gap: 6px;
+  padding: 4px 8px;
+  border-radius: var(--r-sm, 4px);
+  background: var(--bg-elev-2);
+  align-items: baseline;
+}
+
+.modal-viewer-step-mark {
+  text-align: center;
+  font-weight: 600;
+  color: var(--fg-2);
+}
+
+.modal-viewer-step-done .modal-viewer-step-mark {
+  color: #86efac;
+}
+
+.modal-viewer-step-title {
+  color: var(--fg-1);
+  font-size: 13px;
+}
+
+.modal-viewer-step-done .modal-viewer-step-title {
+  color: var(--fg-2);
+}
+
+.modal-viewer-body {
+  margin-top: 4px;
+  font-size: 13px;
+  color: var(--fg-1);
+  max-height: 360px;
+  overflow-y: auto;
+}
+
 /* Side-by-side diff for the edit-review modal — left is "before" (red
  * tint), right is "after" (green tint), context rows render unchanged.
  * Lines hljs-highlight per the file's language. */

dashboard/src/components/chat-internals.ts

@@ -104,6 +104,49 @@ interface RevisionSpec {
   remainingSteps: RevisionStep[];
 }
 
+export type PickerActionName =
+  | "pick"
+  | "delete"
+  | "rename"
+  | "new"
+  | "install"
+  | "uninstall"
+  | "load-more"
+  | "refine"
+  | "cancel";
+
+export interface PickerItemSpec {
+  id: string;
+  title: string;
+  subtitle?: string;
+  badge?: string;
+  meta?: string;
+}
+
+export interface PickerModalSpec {
+  pickerKind: string;
+  title: string;
+  query?: string;
+  items: PickerItemSpec[];
+  actions: PickerActionName[];
+  hasMore?: boolean;
+  hint?: string;
+}
+
+export interface ViewerStep {
+  id: string;
+  title: string;
+  status: "done" | "queued";
+}
+
+export interface ViewerModalSpec {
+  viewerKind: string;
+  title: string;
+  body?: string;
+  steps?: ViewerStep[];
+  meta?: string;
+}
+
 interface DiffEntry {
   kind: "context" | "ins" | "del";
   text: string;
@@ -617,6 +660,220 @@ export function CheckpointModal({ modal, onResolve }: { modal: CheckpointSpec; o
   `;
 }
 
+export function PickerModal({
+  modal,
+  onResolve,
+}: {
+  modal: PickerModalSpec;
+  onResolve: OnResolve;
+}) {
+  useLang();
+  const [selectedId, setSelectedId] = useState<string | null>(modal.items[0]?.id ?? null);
+  const [query, setQuery] = useState(modal.query ?? "");
+  const [renameTarget, setRenameTarget] = useState<string | null>(null);
+  const [renameText, setRenameText] = useState("");
+  const [showNew, setShowNew] = useState(false);
+  const [newText, setNewText] = useState("");
+
+  const has = (a: PickerActionName) => modal.actions.includes(a);
+  const selected = modal.items.find((i) => i.id === selectedId) ?? null;
+
+  const submitRefine = (next: string) => {
+    setQuery(next);
+    if (has("refine")) onResolve("picker", { action: "refine", query: next });
+  };
+
+  const startRename = (id: string) => {
+    const item = modal.items.find((i) => i.id === id);
+    if (!item) return;
+    setRenameTarget(id);
+    setRenameText(item.title);
+  };
+
+  const sendRename = () => {
+    if (!renameTarget || !renameText.trim()) return;
+    onResolve("picker", { action: "rename", id: renameTarget, text: renameText });
+    setRenameTarget(null);
+    setRenameText("");
+  };
+
+  const sendNew = () => {
+    onResolve("picker", newText.trim() ? { action: "new", text: newText } : { action: "new" });
+    setShowNew(false);
+    setNewText("");
+  };
+
+  return html`
+    <${ModalCard}
+      accent="#fcd34d"
+      icon="≡"
+      title=${modal.title}
+      subtitle=${modal.hint}
+    >
+      ${
+        has("refine")
+          ? html`<input
+              class="modal-picker-search"
+              type="search"
+              placeholder=${t("modal.pickerFilter")}
+              value=${query}
+              onInput=${(e: Event) => submitRefine((e.target as HTMLInputElement).value)}
+            />`
+          : null
+      }
+      <div class="modal-picker-list">
+        ${
+          modal.items.length === 0
+            ? html`<div class="modal-picker-empty">${t("modal.pickerEmpty")}</div>`
+            : modal.items.map(
+                (it) => html`
+                  <button
+                    key=${it.id}
+                    class=${`modal-picker-row${it.id === selectedId ? " selected" : ""}`}
+                    onClick=${() => setSelectedId(it.id)}
+                    onDblClick=${() => has("pick") && onResolve("picker", { action: "pick", id: it.id })}
+                  >
+                    <span class="modal-picker-title">${it.title}</span>
+                    ${it.badge ? html`<span class="modal-picker-badge">${it.badge}</span>` : null}
+                    ${it.subtitle ? html`<span class="modal-picker-subtitle">${it.subtitle}</span>` : null}
+                    ${it.meta ? html`<span class="modal-picker-meta">${it.meta}</span>` : null}
+                  </button>
+                `,
+              )
+        }
+      </div>
+      ${
+        modal.hasMore && has("load-more")
+          ? html`<button
+              class="modal-picker-more"
+              onClick=${() => onResolve("picker", { action: "load-more" })}
+            >${t("modal.pickerLoadMore")}</button>`
+          : null
+      }
+      ${
+        renameTarget
+          ? html`
+            <div class="modal-picker-form">
+              <input
+                type="text"
+                value=${renameText}
+                onInput=${(e: Event) => setRenameText((e.target as HTMLInputElement).value)}
+              />
+              <div class="modal-actions">
+                <button class="primary" onClick=${sendRename} disabled=${!renameText.trim()}>${t("common.save")}</button>
+                <button onClick=${() => setRenameTarget(null)}>${t("common.back")}</button>
+              </div>
+            </div>
+          `
+          : showNew
+            ? html`
+              <div class="modal-picker-form">
+                <input
+                  type="text"
+                  placeholder=${t("modal.pickerNewPlaceholder")}
+                  value=${newText}
+                  onInput=${(e: Event) => setNewText((e.target as HTMLInputElement).value)}
+                />
+                <div class="modal-actions">
+                  <button class="primary" onClick=${sendNew}>${t("common.add")}</button>
+                  <button onClick=${() => setShowNew(false)}>${t("common.back")}</button>
+                </div>
+              </div>
+            `
+            : html`
+              <div class="modal-actions">
+                ${
+                  has("pick") && selected
+                    ? html`<button
+                        class="primary"
+                        onClick=${() => onResolve("picker", { action: "pick", id: selected.id })}
+                      >${t("modal.pickerPick")}</button>`
+                    : null
+                }
+                ${
+                  has("install") && selected
+                    ? html`<button
+                        class="primary"
+                        onClick=${() => onResolve("picker", { action: "install", id: selected.id })}
+                      >${t("modal.pickerInstall")}</button>`
+                    : null
+                }
+                ${
+                  has("uninstall") && selected
+                    ? html`<button
+                        onClick=${() => onResolve("picker", { action: "uninstall", id: selected.id })}
+                      >${t("modal.pickerUninstall")}</button>`
+                    : null
+                }
+                ${
+                  has("rename") && selected
+                    ? html`<button onClick=${() => startRename(selected.id)}>${t("modal.pickerRename")}</button>`
+                    : null
+                }
+                ${
+                  has("delete") && selected
+                    ? html`<button
+                        class="danger"
+                        onClick=${() => onResolve("picker", { action: "delete", id: selected.id })}
+                      >${t("common.delete")}</button>`
+                    : null
+                }
+                ${
+                  has("new")
+                    ? html`<button onClick=${() => setShowNew(true)}>${t("modal.pickerNew")}</button>`
+                    : null
+                }
+                <button onClick=${() => onResolve("picker", { action: "cancel" })}>${t("modal.cancel")}</button>
+              </div>
+            `
+      }
+    <//>
+  `;
+}
+
+export function ViewerModal({
+  modal,
+  onResolve,
+}: {
+  modal: ViewerModalSpec;
+  onResolve: OnResolve;
+}) {
+  useLang();
+  return html`
+    <${ModalCard}
+      accent="#67e8f9"
+      icon="◇"
+      title=${modal.title}
+      subtitle=${modal.meta}
+    >
+      ${
+        modal.steps && modal.steps.length > 0
+          ? html`
+            <ol class="modal-viewer-steps">
+              ${modal.steps.map(
+                (s: ViewerStep) => html`
+                  <li key=${s.id} class=${`modal-viewer-step modal-viewer-step-${s.status}`}>
+                    <span class="modal-viewer-step-mark">${s.status === "done" ? "✓" : "·"}</span>
+                    <span class="modal-viewer-step-title">${s.title}</span>
+                  </li>
+                `,
+              )}
+            </ol>
+          `
+          : null
+      }
+      ${
+        modal.body
+          ? html`<div class="md modal-viewer-body" dangerouslySetInnerHTML=${{ __html: marked.parse(modal.body) }}></div>`
+          : null
+      }
+      <div class="modal-actions">
+        <button onClick=${() => onResolve("viewer", { action: "close" })}>${t("modal.viewerClose")}</button>
+      </div>
+    <//>
+  `;
+}
+
 export function RevisionModal({ modal, onResolve }: { modal: RevisionSpec; onResolve: OnResolve }) {
   useLang();
   const riskColor = (r: string | undefined) =>

dashboard/src/i18n/en.ts

@@ -61,8 +61,36 @@ export const en = {
     effortHigh: "high (cheaper / faster)",
     webSearch: "web search",
     webSearchNote: "web_fetch + web_search tools",
+    sectionCompute: "Compute",
+    proNext: "/pro one-shot",
+    proArm: "Arm for next turn",
+    proArmed: "Armed — disarms after next turn",
+    proNextNote: "next turn runs on deepseek-v4-pro, then auto-disarms",
+    sectionBudget: "Budget",
+    budgetOf: "of",
+    budgetSetCap: "set a cap",
+    budgetCustom: "custom",
+    budgetBumpHint: "bump the cap to keep going",
+    budgetClear: "Clear cap",
+    budgetIdleLine: "warns at 80% · refuses past 100%",
+    budgetWarnLine: "approaching cap — loop will refuse past 100%",
+    budgetRefusing: "cap exhausted — next turn refused until bumped or cleared",
+    sectionLoop: "Loop",
+    loopIdleHint: "Auto-resubmit a prompt on a fixed interval.",
+    loopCostHint: "Each iteration costs ~{cost} (last turn).",
+    loopInterval: "interval",
+    loopCustom: "custom",
+    loopRangeError: "interval must fall in 5s..6h",
+    loopPrompt: "prompt",
+    loopPromptPlaceholder: "e.g. check the deploy status and report any errors",
+    loopStart: "Start loop",
+    loopStop: "Stop",
+    loopRunning: "running",
+    loopIter: "iter {iter}",
+    loopFiresIn: "fires in {remaining}",
     sectionRuntime: "Runtime",
     activeModel: "active model",
+    modelPricingLine: "${hit} hit · ${miss} miss · ${out} out  per 1M tok",
     editMode: "edit mode",
     editModeNote: "switch from the Chat tab header",
     sectionLanguage: "Language",
@@ -89,7 +117,8 @@ export const en = {
     clearFailed: "/clear failed: {error}",
     eventStreamError: "event stream interrupted — reconnecting…",
     semanticBanner: "Semantic search isn't enabled for this project.",
-    semanticBannerDesc: "Build the index once and the model can find code by meaning (\"where do we handle auth failures?\") instead of grep on exact strings.",
+    semanticBannerDesc:
+      'Build the index once and the model can find code by meaning ("where do we handle auth failures?") instead of grep on exact strings.',
     semanticBannerBtn: "Build it →",
     semanticBannerDismiss: "dismiss (don't show again)",
     slashCommands: "slash commands",
@@ -98,9 +127,12 @@ export const en = {
     effortMaxTitle: "max (default — best quality)",
     effortHighTitle: "high (cheaper / faster)",
     presetTitle: "preset — model commitment",
-    presetAutoTitle: "auto — flash baseline; auto-escalates to pro on hard turns (NEEDS_PRO / failure threshold)",
-    presetFlashTitle: "flash — always flash; no auto-escalate. /pro still works for one-shot manual",
-    presetProTitle: "pro — always pro; ~3× flash cost (5/31 discount). Locks in on hard architecture work.",
+    presetAutoTitle:
+      "auto — flash baseline; auto-escalates to pro on hard turns (NEEDS_PRO / failure threshold)",
+    presetFlashTitle:
+      "flash — always flash; no auto-escalate. /pro still works for one-shot manual",
+    presetProTitle:
+      "pro — always pro; ~3× flash cost (5/31 discount). Locks in on hard architecture work.",
     editGateTitle: "edit gate — Shift+Tab cycles in TUI",
     editReviewTitle: "review — both edits and non-allowlisted shell ask first",
     editAutoTitle: "auto — edits auto-apply, shell still asks",
@@ -139,12 +171,14 @@ export const en = {
     loading: "loading overview…",
     failed: "overview failed: {error}",
     standaloneTitle: "Standalone mode",
-    standaloneDesc: "Read-only disk view. Start /dashboard from inside reasonix code for live session state, MCP, and tools.",
+    standaloneDesc:
+      "Read-only disk view. Start /dashboard from inside reasonix code for live session state, MCP, and tools.",
     cockpit: "Cockpit",
     balance: "balance",
     tokens7d: "tokens · 7d",
     cacheHit: "cache hit",
     toolCalls24h: "tool calls · 24h",
+    budget: "budget",
     currentSession: "current session",
     noSession: "No live session — /dashboard from inside reasonix code to attach.",
     promptTok: "prompt tok",
@@ -197,7 +231,8 @@ export const en = {
     msgs: "msgs",
     pickHint: "Pick a session on the left to read its transcript.",
     resumeTitle: "Resume in TUI",
-    resumeDesc: "Mid-session swap requires a restart so the message log can rewind cleanly. Quit your current session, then run:",
+    resumeDesc:
+      "Mid-session swap requires a restart so the message log can rewind cleanly. Quit your current session, then run:",
     loadingTranscript: "loading transcript…",
     emptyTranscript: "empty transcript.",
     messages: "{count} message{s}",
@@ -218,24 +253,26 @@ export const en = {
     loading: "loading permissions…",
     failed: "permissions failed: {error}",
     yoloTitle: "YOLO mode",
-    yoloDesc: "Every shell command auto-runs, allowlist bypassed. Switch back with /mode review in the TUI.",
+    yoloDesc:
+      "Every shell command auto-runs, allowlist bypassed. Switch back with /mode review in the TUI.",
     project: "project",
     builtin: "builtin",
     addPrefix: "add a prefix",
-    addPlaceholder: "e.g. \"npm run build\" or \"deploy.sh\"",
+    addPlaceholder: 'e.g. "npm run build" or "deploy.sh"',
     clearAll: "Clear all",
     alreadyIn: "{prefix} already in list",
     added: "added: {prefix}",
     removed: "removed: {prefix}",
     cleared: "cleared {count} entr{y}",
-    removeConfirm: "Remove \"{prefix}\" from this project's allowlist?",
+    removeConfirm: 'Remove "{prefix}" from this project\'s allowlist?',
     clearConfirm: "Wipe every project allowlist entry? Builtin entries are unaffected.",
     projectAllowlist: "Project allowlist · {count}",
     nothingStored: "Nothing stored yet for this project.",
     colNum: "#",
     colPrefix: "prefix",
     builtinTitle: "Builtin · {count} · read-only",
-    standaloneWarning: "Mutations require /dashboard from inside an active reasonix code session — standalone reasonix dashboard can't tell which project's allowlist to edit.",
+    standaloneWarning:
+      "Mutations require /dashboard from inside an active reasonix code session — standalone reasonix dashboard can't tell which project's allowlist to edit.",
   },
   mcp: {
     loading: "loading MCP…",
@@ -255,7 +292,8 @@ export const en = {
     removeBtn: "Remove",
     spec: "spec",
     whyUnbridged: "Why unbridged?",
-    whyUnbridgedDesc: "This spec lives in your config.json but isn't bridged into the live session. MCP servers attach when reasonix code starts; the dashboard alone can't spawn the child process.",
+    whyUnbridgedDesc:
+      "This spec lives in your config.json but isn't bridged into the live session. MCP servers attach when reasonix code starts; the dashboard alone can't spawn the child process.",
     whyUnbridgedHint: "To activate: restart reasonix code, then refresh this dashboard.",
     pickHint: "Pick an MCP server on the left to inspect tools / resources / prompts.",
     toolsTitle: "Tools · {count}",
@@ -283,13 +321,17 @@ export const en = {
     marketplaceAlready: "already installed",
     marketplaceNeedsEnv: "needs env: {names}",
     marketplaceSourceTag: "[{source}]",
-    marketplaceNoInstall: "smithery listing — install metadata not exposed; use `npx -y @smithery/cli install {name}` directly",
-    marketplaceFetchOnInstall: "Smithery listing — install detail fetched on Install. http servers map to streamable-http remotes; stdio servers run via @smithery/cli.",
+    marketplaceNoInstall:
+      "smithery listing — install metadata not exposed; use `npx -y @smithery/cli install {name}` directly",
+    marketplaceFetchOnInstall:
+      "Smithery listing — install detail fetched on Install. http servers map to streamable-http remotes; stdio servers run via @smithery/cli.",
     marketplaceInstalledBadge: "installed",
     marketplaceUninstall: "Uninstall",
     marketplaceEnvTitle: "Required environment variables",
-    marketplaceEnvHint: "Set these in your shell before next `reasonix code` so the bridged server can authenticate.",
-    marketplaceRestartHint: "Spec written to ~/.reasonix/config.json. Restart `reasonix code` to bridge the server (live hot-reload is on the roadmap).",
+    marketplaceEnvHint:
+      "Set these in your shell before next `reasonix code` so the bridged server can authenticate.",
+    marketplaceRestartHint:
+      "Spec written to ~/.reasonix/config.json. Restart `reasonix code` to bridge the server (live hot-reload is on the roadmap).",
   },
   memory: {
     loading: "loading memory…",
@@ -332,7 +374,8 @@ export const en = {
     runs7d: "runs · 7d",
     pickHint: "Pick a skill on the left, or create a new one above.",
     readOnlyBuiltin: "read-only · builtin",
-    builtinDesc: "Built-in skills ship with Reasonix; the model picks them up automatically. To customize, create a project- or global-scoped skill with the same name.",
+    builtinDesc:
+      "Built-in skills ship with Reasonix; the model picks them up automatically. To customize, create a project- or global-scoped skill with the same name.",
     saved: "saved {scope}/{name}",
     deleteConfirm: "Delete skill {scope}/{name}?",
     reloadHint: "re-loaded on next /new or session restart",
@@ -375,6 +418,7 @@ export const en = {
     filterPlaceholder: "filter plans",
     active: "active",
     done: "done",
+    idle: "idle",
     steps: "steps",
     pickHint: "Pick a plan on the left.",
     noTitle: "(no title)",
@@ -388,11 +432,13 @@ export const en = {
     ready: "ready",
     setupNeeded: "setup needed",
     installOllama: "Install Ollama",
-    installOllamaDesc: "Reasonix doesn't run package managers for you. Install Ollama first, then come back:",
+    installOllamaDesc:
+      "Reasonix doesn't run package managers for you. Install Ollama first, then come back:",
     macWindows: "macOS / Windows:",
     download: "download from ollama.com/download",
     linux: "Linux:",
-    refreshHint: "Refresh after install — this panel will offer to start the daemon and pull {model}.",
+    refreshHint:
+      "Refresh after install — this panel will offer to start the daemon and pull {model}.",
     daemon: "Daemon",
     daemonDesc: "ollama is on your PATH but the HTTP daemon isn't reachable.",
     startDaemon: "Start daemon",
@@ -404,16 +450,35 @@ export const en = {
     pullModel: "Pull {model}",
     indexStatus: "index status",
     builtStatus: "● built",
+    incompatibleStatus: "● incompatible",
     chunks: "chunks",
     files: "files",
     dim: "dim",
     size: "size",
     lastBuild: "last build",
+    builtWith: "built with",
+    currentTarget: "current target",
+    incompatibleHint: "This on-disk index was built for a different provider or model. Run Rebuild to replace it.",
     runIndexHint: "Run an index to enable semantic_search.",
     reIndex: "Re-index",
     build: "Build",
     rebuild: "Rebuild",
     stop: "Stop",
+    provider: "provider",
+    providerType: "service type",
+    openaiCompat: "openai-compatible",
+    apiUrl: "API URL",
+    apiKey: "API key",
+    apiKeyStoredNote: "API key is stored in ~/.reasonix/config.json — do not share that file.",
+    customRequestBody: "custom request body",
+    invalidCustomRequestBody: "Custom request body must be valid JSON: {error}",
+    customRequestBodyMustBeObject: "Custom request body must be a JSON object.",
+    saveBeforeIndex: "Save semantic settings before starting an index.",
+    extraBody: "extra body",
+    keepExistingKey: "leave blank to keep existing key",
+    remoteProvider: "Remote embedding provider",
+    remoteProviderDesc:
+      "Configure the full OpenAI-compatible embeddings URL here. Reasonix will send requests exactly to the URL you provide.",
     ollama: "ollama",
     binary: "binary",
     found: "found",
@@ -441,11 +506,14 @@ export const en = {
     nothingSkipped: "nothing skipped — all walked files would be indexed.",
     firstIncluded: "first {count} included file(s)",
     job: "Job",
+    phaseSetup: "preparing",
     phaseScan: "scanning files",
     phaseEmbed: "embedding chunks",
     phaseWrite: "writing index",
     phaseDone: "done",
     phaseError: "error",
+    phaseCancelled: "cancelled",
+    setupFailed: "setup failed",
     stopping: "stopping",
     scanned: "scanned {count}",
     changed: "changed {count}",
@@ -473,7 +541,7 @@ export const en = {
     shellSubtitle: "model wants to run a shell command",
     shellBgSubtitle: "long-running — keeps running after approval",
     runOnce: "Run once",
-    alwaysAllow: "Always allow \"{prefix}\"",
+    alwaysAllow: 'Always allow "{prefix}"',
     deny: "Deny",
     choiceTitle: "model wants you to pick",
     typeOwn: "Type my own answer",
@@ -484,7 +552,8 @@ export const en = {
     cancelSummary: "Drop the question. Model will ask what you actually want.",
     planTitle: "plan submitted",
     planSubtitle: "model proposed a plan; review then pick",
-    approveInstructions: "Optional last instructions / answers to open questions (Enter to send blank)",
+    approveInstructions:
+      "Optional last instructions / answers to open questions (Enter to send blank)",
     refinePlaceholder: "What needs to change? Be specific.",
     approve: "Approve",
     refine: "Refine",
@@ -507,5 +576,15 @@ export const en = {
     reject: "Reject",
     arguments: "arguments",
     revisePlaceholder: "What needs to change before the next step? Leave blank to just continue.",
+    pickerFilter: "Filter…",
+    pickerEmpty: "Nothing to show.",
+    pickerLoadMore: "Load more",
+    pickerPick: "Open",
+    pickerInstall: "Install",
+    pickerUninstall: "Uninstall",
+    pickerRename: "Rename…",
+    pickerNew: "New…",
+    pickerNewPlaceholder: "Name (leave blank for default)",
+    viewerClose: "Close",
   },
 };

dashboard/src/i18n/zh-CN.ts

@@ -61,8 +61,36 @@ export const zhCN = {
     effortHigh: "high（更便宜 / 更快）",
     webSearch: "网页搜索",
     webSearchNote: "web_fetch + web_search 工具",
+    sectionCompute: "计算",
+    proNext: "/pro 单轮",
+    proArm: "为下一轮装备",
+    proArmed: "已装备 — 下一轮后自动解除",
+    proNextNote: "下一轮使用 deepseek-v4-pro，之后自动解除",
+    sectionBudget: "预算",
+    budgetOf: "/",
+    budgetSetCap: "设置上限",
+    budgetCustom: "自定义",
+    budgetBumpHint: "提高上限以继续",
+    budgetClear: "清除上限",
+    budgetIdleLine: "80% 时提醒 · 100% 后拒绝执行",
+    budgetWarnLine: "接近上限 — 超过 100% 将拒绝执行",
+    budgetRefusing: "已超出上限 — 提高或清除后才会继续",
+    sectionLoop: "循环",
+    loopIdleHint: "按固定间隔自动重新提交一段提示词。",
+    loopCostHint: "每次迭代约 {cost}（上一轮成本）。",
+    loopInterval: "间隔",
+    loopCustom: "自定义",
+    loopRangeError: "间隔需在 5s..6h 之间",
+    loopPrompt: "提示词",
+    loopPromptPlaceholder: "例如：检查部署状态并汇报任何错误",
+    loopStart: "启动循环",
+    loopStop: "停止",
+    loopRunning: "运行中",
+    loopIter: "第 {iter} 次",
+    loopFiresIn: "{remaining} 后触发",
     sectionRuntime: "运行时",
     activeModel: "当前模型",
+    modelPricingLine: "${hit} 命中 · ${miss} 未命中 · ${out} 输出  / 100 万 tok",
     editMode: "编辑模式",
     editModeNote: "在对话标签页头部切换",
     sectionLanguage: "语言",
@@ -89,7 +117,8 @@ export const zhCN = {
     clearFailed: "/clear 失败：{error}",
     eventStreamError: "事件流中断 — 正在重连…",
     semanticBanner: "此项目未启用语义搜索。",
-    semanticBannerDesc: "构建一次索引，模型即可按含义查找代码（“哪里处理认证失败？”），而不仅依赖精确字符串的 grep。",
+    semanticBannerDesc:
+      "构建一次索引，模型即可按含义查找代码（“哪里处理认证失败？”），而不仅依赖精确字符串的 grep。",
     semanticBannerBtn: "构建 →",
     semanticBannerDismiss: "关闭（不再显示）",
     slashCommands: "斜杠命令",
@@ -139,12 +168,14 @@ export const zhCN = {
     loading: "加载概览…",
     failed: "概览失败：{error}",
     standaloneTitle: "独立模式",
-    standaloneDesc: "只读磁盘视图。在 reasonix code 内启动 /dashboard 以获取实时会话状态、MCP 和工具。",
+    standaloneDesc:
+      "只读磁盘视图。在 reasonix code 内启动 /dashboard 以获取实时会话状态、MCP 和工具。",
     cockpit: "驾驶舱",
     balance: "余额",
     tokens7d: "tokens · 7 天",
     cacheHit: "缓存命中",
     toolCalls24h: "工具调用 · 24 小时",
+    budget: "预算",
     currentSession: "当前会话",
     noSession: "无活跃会话 — 在 reasonix code 内执行 /dashboard 进行连接。",
     promptTok: "提示 tokens",
@@ -222,20 +253,21 @@ export const zhCN = {
     project: "项目",
     builtin: "内置",
     addPrefix: "添加前缀",
-    addPlaceholder: "例如 \"npm run build\" 或 \"deploy.sh\"",
+    addPlaceholder: '例如 "npm run build" 或 "deploy.sh"',
     clearAll: "清除全部",
     alreadyIn: "{prefix} 已在列表中",
     added: "已添加：{prefix}",
     removed: "已移除：{prefix}",
     cleared: "已清除 {count} 条",
-    removeConfirm: "从此项目的允许列表中移除 \"{prefix}\"？",
+    removeConfirm: '从此项目的允许列表中移除 "{prefix}"？',
     clearConfirm: "清除所有项目允许列表条目？内置条目不受影响。",
     projectAllowlist: "项目允许列表 · {count}",
     nothingStored: "此项目暂无存储的条目。",
     colNum: "#",
     colPrefix: "前缀",
     builtinTitle: "内置 · {count} · 只读",
-    standaloneWarning: "修改操作需要在活跃的 reasonix code 会话内执行 /dashboard — 独立模式的 reasonix dashboard 无法确定要编辑哪个项目的允许列表。",
+    standaloneWarning:
+      "修改操作需要在活跃的 reasonix code 会话内执行 /dashboard — 独立模式的 reasonix dashboard 无法确定要编辑哪个项目的允许列表。",
   },
   mcp: {
     loading: "加载 MCP…",
@@ -255,7 +287,8 @@ export const zhCN = {
     removeBtn: "移除",
     spec: "规格",
     whyUnbridged: "为什么未桥接？",
-    whyUnbridgedDesc: "此规格存在于您的 config.json 中，但未桥接到实时会话。MCP 服务器在 reasonix code 启动时连接；仪表盘本身无法生成子进程。",
+    whyUnbridgedDesc:
+      "此规格存在于您的 config.json 中，但未桥接到实时会话。MCP 服务器在 reasonix code 启动时连接；仪表盘本身无法生成子进程。",
     whyUnbridgedHint: "激活方法：重启 reasonix code，然后刷新此仪表盘。",
     pickHint: "选择左侧的 MCP 服务器以检查工具 / 资源 / 提示。",
     toolsTitle: "工具 · {count}",
@@ -283,13 +316,16 @@ export const zhCN = {
     marketplaceAlready: "已安装过",
     marketplaceNeedsEnv: "需设置环境变量：{names}",
     marketplaceSourceTag: "[{source}]",
-    marketplaceNoInstall: "Smithery 列表项 — 不暴露安装元数据；请直接 `npx -y @smithery/cli install {name}`",
-    marketplaceFetchOnInstall: "Smithery 列表 — 安装时再拉详情。HTTP 服务映射为 streamable-http 远端；stdio 服务通过 @smithery/cli 运行。",
+    marketplaceNoInstall:
+      "Smithery 列表项 — 不暴露安装元数据；请直接 `npx -y @smithery/cli install {name}`",
+    marketplaceFetchOnInstall:
+      "Smithery 列表 — 安装时再拉详情。HTTP 服务映射为 streamable-http 远端；stdio 服务通过 @smithery/cli 运行。",
     marketplaceInstalledBadge: "已安装",
     marketplaceUninstall: "卸载",
     marketplaceEnvTitle: "必需的环境变量",
     marketplaceEnvHint: "下次启动 `reasonix code` 之前在 shell 里设好，桥接的服务器才能正常鉴权。",
-    marketplaceRestartHint: "已写入 ~/.reasonix/config.json。重启 `reasonix code` 后服务器才会真正桥接（热重载在路线图上）。",
+    marketplaceRestartHint:
+      "已写入 ~/.reasonix/config.json。重启 `reasonix code` 后服务器才会真正桥接（热重载在路线图上）。",
   },
   memory: {
     loading: "加载记忆…",
@@ -332,7 +368,8 @@ export const zhCN = {
     runs7d: "次运行 · 7 天",
     pickHint: "选择左侧的技能，或在上方创建新技能。",
     readOnlyBuiltin: "只读 · 内置",
-    builtinDesc: "内置技能随 Reasonix 一起发布；模型会自动识别。如需自定义，请创建同名的项目或全局技能。",
+    builtinDesc:
+      "内置技能随 Reasonix 一起发布；模型会自动识别。如需自定义，请创建同名的项目或全局技能。",
     saved: "已保存 {scope}/{name}",
     deleteConfirm: "删除技能 {scope}/{name}？",
     reloadHint: "在下次 /new 或会话重启时重新加载",
@@ -375,6 +412,7 @@ export const zhCN = {
     filterPlaceholder: "筛选计划",
     active: "进行中",
     done: "已完成",
+    idle: "未开始",
     steps: "步骤",
     pickHint: "选择左侧的计划。",
     noTitle: "（无标题）",
@@ -404,16 +442,34 @@ export const zhCN = {
     pullModel: "拉取 {model}",
     indexStatus: "索引状态",
     builtStatus: "● 已构建",
+    incompatibleStatus: "● 不兼容",
     chunks: "分块",
     files: "文件",
     dim: "维度",
     size: "大小",
     lastBuild: "上次构建",
+    builtWith: "构建来源",
+    currentTarget: "当前目标",
+    incompatibleHint: "磁盘上的这个索引是为不同的 provider 或 model 构建的。运行“完全重建”即可替换。",
     runIndexHint: "运行索引以启用 semantic_search。",
     reIndex: "重建索引",
     build: "构建",
     rebuild: "完全重建",
     stop: "停止",
+    provider: "提供方",
+    providerType: "服务类型",
+    openaiCompat: "OpenAI-Compatible",
+    apiUrl: "API URL",
+    apiKey: "API Key",
+    customRequestBody: "自定义请求体",
+    invalidCustomRequestBody: "自定义请求体必须是合法 JSON：{error}",
+    customRequestBodyMustBeObject: "自定义请求体必须是 JSON 对象。",
+    saveBeforeIndex: "请先保存语义设置，再启动索引。",
+    extraBody: "扩展请求体",
+    keepExistingKey: "留空则保留现有 Key",
+    remoteProvider: "远程向量服务",
+    remoteProviderDesc:
+      "在这里配置 OpenAI-Compatible embeddings 的完整 URL。Reasonix 会严格使用你提供的 URL 发起请求。",
     ollama: "Ollama",
     binary: "二进制",
     found: "已找到",
@@ -441,11 +497,14 @@ export const zhCN = {
     nothingSkipped: "无跳过 — 所有遍历的文件都将被索引。",
     firstIncluded: "前 {count} 个包含的文件",
     job: "任务",
+    phaseSetup: "初始化中",
     phaseScan: "扫描文件",
     phaseEmbed: "嵌入分块",
     phaseWrite: "写入索引",
     phaseDone: "完成",
     phaseError: "错误",
+    phaseCancelled: "已停止",
+    setupFailed: "初始化失败",
     stopping: "停止中",
     scanned: "已扫描 {count}",
     changed: "已变更 {count}",
@@ -473,7 +532,7 @@ export const zhCN = {
     shellSubtitle: "模型想要运行一条 Shell 命令",
     shellBgSubtitle: "长时间运行 — 批准后继续运行",
     runOnce: "运行一次",
-    alwaysAllow: "始终允许 \"{prefix}\"",
+    alwaysAllow: '始终允许 "{prefix}"',
     deny: "拒绝",
     choiceTitle: "模型需要您选择",
     typeOwn: "输入自定义回答",
@@ -507,5 +566,15 @@ export const zhCN = {
     reject: "拒绝",
     arguments: "参数",
     revisePlaceholder: "下一步之前需要更改什么？留空则直接继续。",
+    pickerFilter: "过滤…",
+    pickerEmpty: "暂无内容。",
+    pickerLoadMore: "加载更多",
+    pickerPick: "打开",
+    pickerInstall: "安装",
+    pickerUninstall: "卸载",
+    pickerRename: "重命名…",
+    pickerNew: "新建…",
+    pickerNewPlaceholder: "名称（留空使用默认）",
+    viewerClose: "关闭",
   },
 };

dashboard/src/lib/budget.ts

@@ -0,0 +1,45 @@
+export type BudgetState =
+  | { kind: "off"; spent: number }
+  | { kind: "running"; cap: number; spent: number; pct: number }
+  | { kind: "warn"; cap: number; spent: number; pct: number }
+  | { kind: "exhausted"; cap: number; spent: number; pct: number };
+
+export function deriveBudgetState(
+  cap: number | null | undefined,
+  spent: number | null | undefined,
+): BudgetState {
+  const safeSpent = typeof spent === "number" && spent >= 0 ? spent : 0;
+  if (typeof cap !== "number" || cap <= 0) {
+    return { kind: "off", spent: safeSpent };
+  }
+  const pct = (safeSpent / cap) * 100;
+  if (pct >= 100) return { kind: "exhausted", cap, spent: safeSpent, pct };
+  if (pct >= 80) return { kind: "warn", cap, spent: safeSpent, pct };
+  return { kind: "running", cap, spent: safeSpent, pct };
+}
+
+/** Default quick-cap menu — round dollar amounts users actually pick. */
+export const QUICK_CAPS_USD: ReadonlyArray<number> = [1, 5, 10, 25, 50];
+
+/** 1.5× / 2× / 4× the current cap, snapped to a "nice" round number per bucket. */
+export function bumpSuggestions(currentCap: number): number[] {
+  if (currentCap <= 0) return [];
+  return [niceUp(currentCap * 1.5), niceUp(currentCap * 2), niceUp(currentCap * 4)];
+}
+
+function niceUp(n: number): number {
+  // Subtract a tiny epsilon before ceil so FP noise (0.4 * 1.5 = 0.6000…01)
+  // doesn't bump a value to the next bucket.
+  const eps = 1e-9;
+  if (n < 1) return Math.ceil((n - eps) * 10) / 10;
+  if (n < 10) return Math.ceil((n - eps) * 2) / 2;
+  if (n < 100) return Math.ceil(n - eps);
+  return Math.ceil((n - eps) / 5) * 5;
+}
+
+/** Tone class shared between the cockpit tile and the settings gauge. */
+export function budgetTone(state: BudgetState): "" | "warn" | "err" {
+  if (state.kind === "exhausted") return "err";
+  if (state.kind === "warn") return "warn";
+  return "";
+}

dashboard/src/lib/format.ts

@@ -4,6 +4,24 @@ export function fmtUsd(n: number | null | undefined): string {
   return `$${n.toFixed(n < 0.01 ? 6 : 4)}`;
 }
 
+/** Keep in sync with src/cli/ui/theme/tokens.ts USD_TO_CNY. */
+const USD_TO_CNY = 7.2;
+
+/** USD-internal cost rendered in the wallet's display currency. Undefined currency → CNY (matches CLI default). */
+export function fmtCost(
+  usd: number | null | undefined,
+  currency: string | null | undefined,
+  fractionDigits?: number,
+): string {
+  if (usd === null || usd === undefined) return "—";
+  const cur = currency ?? "CNY";
+  const amount = cur === "CNY" ? usd * USD_TO_CNY : usd;
+  if (amount === 0) return cur === "CNY" ? "¥0" : "$0";
+  const sym = cur === "CNY" ? "¥" : cur === "USD" ? "$" : `${cur} `;
+  const digits = fractionDigits ?? (Math.abs(amount) < 0.01 ? 6 : 4);
+  return `${sym}${amount.toFixed(digits)}`;
+}
+
 export function fmtPct(n: number | null | undefined): string {
   if (n === null || n === undefined) return "—";
   return `${(n * 100).toFixed(1)}%`;

dashboard/src/lib/loop-control.ts

@@ -0,0 +1,48 @@
+export type IntervalUnit = "s" | "m" | "h";
+
+export interface LoopRunStatus {
+  prompt: string;
+  intervalMs: number;
+  iter: number;
+  /** Wall-clock ms until the next fire — server reports a remaining duration, not an absolute. */
+  nextFireMs: number;
+}
+
+/** Quick-pick intervals in ms — covers the 95% of cases users actually run. */
+export const INTERVAL_PRESETS_MS: ReadonlyArray<{ ms: number; label: string }> = [
+  { ms: 30_000, label: "30s" },
+  { ms: 60_000, label: "1m" },
+  { ms: 5 * 60_000, label: "5m" },
+  { ms: 15 * 60_000, label: "15m" },
+  { ms: 60 * 60_000, label: "1h" },
+  { ms: 6 * 60 * 60_000, label: "6h" },
+];
+
+const UNIT_TO_MS: Record<IntervalUnit, number> = {
+  s: 1_000,
+  m: 60_000,
+  h: 60 * 60_000,
+};
+
+const MIN_INTERVAL_MS = 5_000;
+const MAX_INTERVAL_MS = 6 * 60 * 60_000;
+
+/** Convert a "30" + "s" pair to ms, returning null if out of [5s, 6h]. */
+export function parseCustomInterval(value: string, unit: IntervalUnit): number | null {
+  const n = Number.parseFloat(value);
+  if (!Number.isFinite(n) || n <= 0) return null;
+  const ms = Math.round(n * UNIT_TO_MS[unit]);
+  if (ms < MIN_INTERVAL_MS || ms > MAX_INTERVAL_MS) return null;
+  return ms;
+}
+
+/** Human-friendly "5m 12s" / "12s" / "2h 45m" — shows two largest non-zero units. */
+export function formatRemaining(ms: number): string {
+  const safe = Math.max(0, Math.floor(ms / 1000));
+  const h = Math.floor(safe / 3600);
+  const m = Math.floor((safe % 3600) / 60);
+  const s = safe % 60;
+  if (h > 0) return m > 0 ? `${h}h ${m}m` : `${h}h`;
+  if (m > 0) return s > 0 ? `${m}m ${s}s` : `${m}m`;
+  return `${s}s`;
+}

dashboard/src/panels/chat.ts

@@ -6,15 +6,17 @@ import {
   ChoiceModal,
   EditReviewModal,
   type OnResolve,
+  PickerModal,
   PlanModal,
   RevisionModal,
   ShellModal,
+  ViewerModal,
   WorkspaceModal,
   parseToolArgs,
 } from "../components/chat-internals.js";
 import { MODE, TOKEN, api } from "../lib/api.js";
 import { appBus, showToast } from "../lib/bus.js";
-import { fmtUsd } from "../lib/format.js";
+import { fmtCost, fmtUsd } from "../lib/format.js";
 import { html } from "../lib/html.js";
 import { t, useLang } from "../i18n/index.js";
 
@@ -559,7 +561,7 @@ export function ChatPanel() {
     <div class="chat-shell">
       <div style="display:flex;align-items:center;gap:12px;margin-bottom:10px">
         <div class="chips" style="padding:0">
-          <span class="chip-f active">${MODE === "attached" ? t("chat.modeMirror") : t("chat.modeView")}</span>
+          <span class="chip-f static active">${MODE === "attached" ? t("chat.modeMirror") : t("chat.modeView")}</span>
         </div>
         <div class="header-pickers" style="margin-left:auto">
           ${
@@ -684,7 +686,11 @@ export function ChatPanel() {
                       ? html`<${CheckpointModal} modal=${modal} onResolve=${resolveModal} />`
                       : modal.kind === "revision"
                         ? html`<${RevisionModal} modal=${modal} onResolve=${resolveModal} />`
-                        : null
+                        : modal.kind === "picker"
+                          ? html`<${PickerModal} modal=${modal} onResolve=${resolveModal} />`
+                          : modal.kind === "viewer"
+                            ? html`<${ViewerModal} modal=${modal} onResolve=${resolveModal} />`
+                            : null
           : null
       }
 
@@ -791,6 +797,7 @@ function SideRail({ stats, budgetUsd, activePlan }: SideRailProps) {
   const showBudget = stats != null && typeof budgetUsd === "number" && budgetUsd > 0;
   const budgetPct = showBudget ? Math.min(120, (stats.totalCostUsd / budgetUsd) * 100) : 0;
   const budgetTone = budgetPct >= 100 ? "err" : budgetPct >= 80 ? "warn" : "";
+  const walletCurrency = stats?.balance?.[0]?.currency;
   return html`
     <aside class="chat-rail">
       ${activePlan ? html`<${ActivePlanCard} plan=${activePlan} />` : null}
@@ -801,7 +808,7 @@ function SideRail({ stats, budgetUsd, activePlan }: SideRailProps) {
               <div class="rh">${t("chat.railSession")}</div>
               <div class="rail-kv"><span class="k">${t("chat.railTurns")}</span><span class="v">${stats.turns.toLocaleString()}</span></div>
               <div class="rail-kv"><span class="k">${t("chat.railPromptTok")}</span><span class="v">${stats.lastPromptTokens.toLocaleString()}</span></div>
-              <div class="rail-kv"><span class="k">${t("chat.railCost")}</span><span class="v">${fmtUsd(stats.totalCostUsd)}</span></div>
+              <div class="rail-kv"><span class="k">${t("chat.railCost")}</span><span class="v">${fmtCost(stats.totalCostUsd, walletCurrency)}</span></div>
               <div class="progress-row" style="margin-top:8px">
                 <span class="lbl">${t("chat.railCacheHit")}</span>
                 <div class=${`progress ${cacheTone}`}><div class="progress-fill" style=${`width:${cachePct}%`}></div></div>
@@ -819,7 +826,7 @@ function SideRail({ stats, budgetUsd, activePlan }: SideRailProps) {
               <div class="progress-row">
                 <span class="lbl">${t("chat.railSpend")}</span>
                 <div class=${`progress ${budgetTone}`}><div class="progress-fill" style=${`width:${Math.min(100, budgetPct)}%`}></div></div>
-                <span class="v" style=${budgetTone === "err" ? "color:var(--c-err)" : budgetTone === "warn" ? "color:var(--c-warn)" : ""}>${fmtUsd(stats.totalCostUsd)} / ${fmtUsd(budgetUsd)}</span>
+                <span class="v" style=${budgetTone === "err" ? "color:var(--c-err)" : budgetTone === "warn" ? "color:var(--c-warn)" : ""}>${fmtCost(stats.totalCostUsd, walletCurrency)} / ${fmtCost(budgetUsd, walletCurrency)}</span>
               </div>
             </div>
           `
@@ -989,11 +996,11 @@ function ChatStatusBar({ stats, model }: ChatStatusBarProps) {
       </span>
       <span class="status-item">
         <span class="status-label">${t("chat.statusTurn")}</span>
-        <code>${fmtUsd(stats.lastTurnCostUsd)}</code>
+        <code>${fmtCost(stats.lastTurnCostUsd, balance?.currency)}</code>
       </span>
       <span class="status-item">
         <span class="status-label">${t("chat.statusSession")}</span>
-        <code>${fmtUsd(stats.totalCostUsd)}</code>
+        <code>${fmtCost(stats.totalCostUsd, balance?.currency)}</code>
         <span class="muted" style="font-size: 10px;">
           ${t("chat.statusTurns", { count: stats.turns, s: stats.turns === 1 ? "" : "s" })}
         </span>

dashboard/src/panels/hooks.ts

@@ -68,6 +68,7 @@ export function HooksPanel() {
   const [drafts, setDrafts] = useState<Record<string, string>>({});
   const [busy, setBusy] = useState(false);
   const [info, setInfo] = useState<string | null>(null);
+  const [eventFilter, setEventFilter] = useState<string>("all");
 
   const load = useCallback(async () => {
     try {
@@ -129,19 +130,31 @@ export function HooksPanel() {
     data.events.length > 0
       ? data.events
       : Array.from(new Set(matrixRows.flatMap((r) => Object.keys(r.cells))));
+  const visibleRows =
+    eventFilter === "all"
+      ? matrixRows
+      : matrixRows.filter((r) => r.cells[eventFilter]?.on);
   const gridCols = `220px repeat(${Math.max(events.length, 1)}, minmax(0, 1fr))`;
 
   return html`
     <div style="display:flex;flex-direction:column;gap:6px">
       <div class="chips">
-        <span class="chip-f active">${t("hooks.resolved")} <span class="ct">${data.resolved.length}</span></span>
-        ${data.events.map((ev) => html`<span class="chip-f">${ev}</span>`)}
+        <span
+          class=${`chip-f ${eventFilter === "all" ? "active" : ""}`}
+          onClick=${() => setEventFilter("all")}
+        >${t("hooks.resolved")} <span class="ct">${data.resolved.length}</span></span>
+        ${data.events.map(
+          (ev) => html`<span
+            class=${`chip-f ${eventFilter === ev ? "active" : ""}`}
+            onClick=${() => setEventFilter(ev)}
+          >${ev}</span>`,
+        )}
       </div>
       ${info ? html`<div><span class="pill ok">${info}</span></div>` : null}
       ${error ? html`<div class="card accent-err">${error}</div>` : null}
 
       ${sectionH3(t("hooks.eventMatrix"), t("hooks.matrixSub", { scripts: matrixRows.length, s: matrixRows.length === 1 ? "" : "s", events: events.length }))}${
-        matrixRows.length === 0
+        visibleRows.length === 0
           ? html`<div class="card" style="color:var(--fg-3)">
               ${t("hooks.noHooks")}
             </div>`
@@ -152,7 +165,7 @@ export function HooksPanel() {
                   <div>${t("hooks.colScript")}</div>
                   ${events.map((ev) => html`<div>${ev}</div>`)}
                 </div>
-                ${matrixRows.map(
+                ${visibleRows.map(
                   (r) => html`
                     <div class="row" style=${`grid-template-columns:${gridCols}`}>
                       <div class="cell" title=${r.command}>

dashboard/src/panels/overview.ts

@@ -1,4 +1,5 @@
-import { fmtCompactNum, fmtNum, fmtRelativeTime, fmtUsd } from "../lib/format.js";
+import { budgetTone, deriveBudgetState } from "../lib/budget.js";
+import { fmtCompactNum, fmtCost, fmtNum, fmtRelativeTime, fmtUsd } from "../lib/format.js";
 import { html } from "../lib/html.js";
 import { usePoll } from "../lib/use-poll.js";
 import { t, useLang } from "../i18n/index.js";
@@ -65,6 +66,9 @@ interface OverviewData {
   toolCount?: number;
   cwd?: string;
   cockpit?: CockpitData;
+  budgetUsd?: number | null;
+  /** Cumulative session spend in USD — set when a session is attached. */
+  sessionSpendUsd?: number | null;
 }
 
 function kpi(label: string, value: unknown, delta?: unknown, deltaTone?: "up" | "down" | "flat") {
@@ -107,6 +111,25 @@ function balanceKpi(c: CockpitData) {
   return kpi(t("overview.balance"), `${symbol}${c.balance.total}`, c.balance.currency, "flat");
 }
 
+function budgetKpi(o: OverviewData) {
+  const state = deriveBudgetState(o.budgetUsd, o.cockpit?.currentSession?.totalCostUsd ?? null);
+  if (state.kind === "off") return null;
+  const tone = budgetTone(state);
+  const valueColor =
+    tone === "err"
+      ? "color:var(--c-err)"
+      : tone === "warn"
+        ? "color:var(--c-warn)"
+        : "";
+  return html`
+    <div class="kpi cock-w-1">
+      <div class="label">${t("overview.budget")}</div>
+      <div class="value" style=${valueColor}>${fmtUsd(state.spent)} / ${fmtUsd(state.cap)}</div>
+      <div class=${`progress ${tone}`} style="margin-top:4px"><div class="progress-fill" style=${`width:${Math.min(100, state.pct)}%`}></div></div>
+    </div>
+  `;
+}
+
 function tokens7dKpi(c: CockpitData) {
   if (!c.tokens7d) return kpi(t("overview.tokens7d"), "—", t("overview.noUsageYet"), "flat");
   const d = deltaPctText(c.tokens7d.deltaPct);
@@ -144,6 +167,7 @@ function currentSessionBlock(c: CockpitData) {
     `;
   }
   const s = c.currentSession;
+  const currency = c.balance?.currency;
   return html`
     <div class="cock-list cock-w-2">
       <div class="ch"><span class="ttl">${t("overview.currentSession")}</span></div>
@@ -153,7 +177,7 @@ function currentSessionBlock(c: CockpitData) {
       <div style="display:grid;grid-template-columns:repeat(3, 1fr);gap:8px;font-family:var(--font-mono);font-size:11px">
         <div><span style="color:var(--fg-3)">${t("overview.promptTok")}</span><div style="color:var(--fg-0);font-size:13px;font-weight:600">${fmtNum(s.lastPromptTokens)}</div></div>
         <div><span style="color:var(--fg-3)">${t("overview.completionTok")}</span><div style="color:var(--fg-0);font-size:13px;font-weight:600">${fmtNum(s.completionTokens)}</div></div>
-        <div><span style="color:var(--fg-3)">${t("overview.cost")}</span><div style="color:var(--fg-0);font-size:13px;font-weight:600">${fmtUsd(s.totalCostUsd)}</div></div>
+        <div><span style="color:var(--fg-3)">${t("overview.cost")}</span><div style="color:var(--fg-0);font-size:13px;font-weight:600">${fmtCost(s.totalCostUsd, currency)}</div></div>
       </div>
     </div>
   `;
@@ -185,7 +209,7 @@ function costTrendSpark(c: CockpitData) {
   return html`
     <div class="chart cock-w-2">
       <div class="chart-h"><span class="title">${t("overview.costTrend")}</span></div>
-      <div class="chart-v">${fmtUsd(avg)}<span class="unit">${t("overview.dayAvg")}</span></div>
+      <div class="chart-v">${fmtCost(avg, c.balance?.currency)}<span class="unit">${t("overview.dayAvg")}</span></div>
       <div class="chart-spark">
         <svg viewBox=${`0 0 ${w} ${h}`} preserveAspectRatio="none">
           <polyline fill="none" stroke="var(--c-brand)" stroke-width="1.5" points=${points} />
@@ -284,6 +308,7 @@ export function OverviewPanel() {
         ${tokens7dKpi(c)}
         ${cacheHitKpi(c)}
         ${toolCallsKpi(c)}
+        ${budgetKpi(o)}
 
         ${currentSessionBlock(c)}
         ${costTrendSpark(c)}

dashboard/src/panels/permissions.ts

@@ -122,8 +122,8 @@ export function PermissionsPanel() {
       }
 
       <div class="chips">
-        <span class="chip-f active">${t("permissions.project")} <span class="ct">${p.project.length}</span></span>
-        <span class="chip-f">${t("permissions.builtin")} <span class="ct">${p.builtin.length}</span></span>
+        <span class="chip-f static active">${t("permissions.project")} <span class="ct">${p.project.length}</span></span>
+        <span class="chip-f static">${t("permissions.builtin")} <span class="ct">${p.builtin.length}</span></span>
       </div>
 
       ${

dashboard/src/panels/plans.ts

@@ -29,7 +29,7 @@ interface PlansData {
 function statusPill(p: ArchivedPlan) {
   if (p.completionRatio >= 1) return html`<span class="pill ok">${t("plans.done")}</span>`;
   if (p.completionRatio > 0) return html`<span class="pill info">${t("plans.active")}</span>`;
-  return html`<span class="pill">idle</span>`;
+  return html`<span class="pill">${t("plans.idle")}</span>`;
 }
 
 export function PlansPanel() {
@@ -37,6 +37,7 @@ export function PlansPanel() {
   const { data, error, loading } = usePoll<PlansData>("/plans", 8000);
   const [openIdx, setOpenIdx] = useState<number | null>(null);
   const [filter, setFilter] = useState("");
+  const [statusFilter, setStatusFilter] = useState<"all" | "active" | "done">("all");
 
   if (loading && !data)
     return html`<div class="card" style="color:var(--fg-3)">${t("plans.loading")}</div>`;
@@ -48,13 +49,19 @@ export function PlansPanel() {
       ${t("plans.noPlans")}
     </div>`;
 
+  const statusFiltered =
+    statusFilter === "all"
+      ? plans
+      : statusFilter === "active"
+        ? plans.filter((p) => p.completionRatio > 0 && p.completionRatio < 1)
+        : plans.filter((p) => p.completionRatio >= 1);
   const filtered = filter.trim()
-    ? plans.filter(
+    ? statusFiltered.filter(
         (p) =>
           p.session.toLowerCase().includes(filter.toLowerCase()) ||
           (p.summary ?? "").toLowerCase().includes(filter.toLowerCase()),
       )
-    : plans;
+    : statusFiltered;
 
   const open = openIdx !== null ? plans[openIdx] : null;
 
@@ -71,12 +78,21 @@ export function PlansPanel() {
           />
         </div>
         <div class="chips" style="padding:0 12px 8px">
-          <span class="chip-f active">${t("common.all")} <span class="ct">${plans.length}</span></span>
-          <span class="chip-f">
+          <span
+            class=${`chip-f ${statusFilter === "all" ? "active" : ""}`}
+            onClick=${() => setStatusFilter("all")}
+          >${t("common.all")} <span class="ct">${plans.length}</span></span>
+          <span
+            class=${`chip-f ${statusFilter === "active" ? "active" : ""}`}
+            onClick=${() => setStatusFilter("active")}
+          >
             ${t("plans.active")}
             <span class="ct">${plans.filter((p) => p.completionRatio > 0 && p.completionRatio < 1).length}</span>
           </span>
-          <span class="chip-f">
+          <span
+            class=${`chip-f ${statusFilter === "done" ? "active" : ""}`}
+            onClick=${() => setStatusFilter("done")}
+          >
             ${t("plans.done")} <span class="ct">${plans.filter((p) => p.completionRatio >= 1).length}</span>
           </span>
         </div>

dashboard/src/panels/semantic.ts

@@ -1,13 +1,50 @@
-import { useCallback, useEffect, useState } from "preact/hooks";
+import { useCallback, useEffect, useRef, useState } from "preact/hooks";
+import { t, useLang } from "../i18n/index.js";
 import { api } from "../lib/api.js";
 import { fmtBytes, fmtNum, fmtRelativeTime } from "../lib/format.js";
 import { html } from "../lib/html.js";
-import { t, useLang } from "../i18n/index.js";
+
+interface SemanticConfigView {
+  provider: "ollama" | "openai-compat";
+  ollama: {
+    baseUrl: string;
+    model: string;
+  };
+  openaiCompat: {
+    baseUrl: string;
+    apiKey: string;
+    apiKeySet: boolean;
+    model: string;
+    extraBody: Record<string, unknown>;
+  };
+}
 
 interface SemanticData {
   attached?: boolean;
   reason?: string;
   root?: string;
+  provider?: "ollama" | "openai-compat";
+  providerConfig?: SemanticConfigView;
+  providerStatus?:
+    | {
+        kind: "ollama";
+        ready: boolean;
+        baseUrl: string;
+        binaryFound?: boolean;
+        daemonRunning?: boolean;
+        modelPulled?: boolean;
+        modelName?: string;
+        installedModels?: string[];
+        error?: string;
+      }
+    | {
+        kind: "openai-compat";
+        ready: boolean;
+        baseUrl: string;
+        apiKeySet: boolean;
+        model: string;
+        extraBodyKeys: string[];
+      };
   index?: IndexInfo;
   job?: SemanticJob | null;
   pull?: { status: string; startedAt: number; lastLine?: string } | null;
@@ -23,17 +60,25 @@ interface SemanticData {
 
 interface IndexInfo {
   exists: boolean;
+  provider?: "ollama" | "openai-compat";
   chunks?: number;
   files?: number;
   dim?: number;
   sizeBytes?: number;
   lastBuiltMs?: number;
   model?: string;
+  builtWith?: { provider: "ollama" | "openai-compat"; model: string };
+  current?: { provider: "ollama" | "openai-compat"; model: string };
+  compatible?: boolean;
+  mismatch?: "provider" | "model" | null;
 }
 
 interface SemanticJob {
   phase: string;
   startedAt: number;
+  finishedAt?: number | null;
+  cancelledAt?: number | null;
+  lastPhase?: string | null;
   chunksTotal?: number;
   chunksDone?: number;
   filesScanned?: number;
@@ -50,17 +95,44 @@ interface SemanticJob {
   };
 }
 
+interface SemanticConfigDraft {
+  provider: "ollama" | "openai-compat";
+  ollama: {
+    baseUrl: string;
+    model: string;
+  };
+  openaiCompat: {
+    baseUrl: string;
+    apiKey: string;
+    model: string;
+    extraBodyText: string;
+    apiKeySet: boolean;
+  };
+}
+
+export interface SemanticDraftValidation {
+  extraBody: Record<string, unknown>;
+  error: string | null;
+}
+
 export function SemanticPanel() {
   useLang();
   const [data, setData] = useState<SemanticData | null>(null);
+  const [draft, setDraft] = useState<SemanticConfigDraft | null>(null);
+  const [draftDirty, setDraftDirty] = useState(false);
+  const draftDirtyRef = useRef(false);
   const [error, setError] = useState<string | null>(null);
   const [busy, setBusy] = useState(false);
   const [info, setInfo] = useState<string | null>(null);
 
   const load = useCallback(async () => {
     try {
-      const r = await api<SemanticData>("/semantic");
-      setData(r);
+      const [semantic, config] = await Promise.all([
+        api<SemanticData>("/semantic"),
+        api<SemanticConfigView>("/semantic/config"),
+      ]);
+      setData(semantic);
+      setDraft((current) => (current && draftDirtyRef.current ? current : toConfigDraft(config)));
     } catch (err) {
       setError((err as Error).message);
     }
@@ -69,7 +141,7 @@ export function SemanticPanel() {
   useEffect(() => {
     load();
     const phase = data?.job?.phase;
-    const running = phase === "scan" || phase === "embed" || phase === "write";
+    const running = isActiveSemanticPhase(phase);
     const pulling = data?.pull?.status === "pulling";
     const ms = running || pulling ? 1200 : 5000;
     const id = setInterval(load, ms);
@@ -78,10 +150,18 @@ export function SemanticPanel() {
 
   const start = useCallback(
     async (rebuild: boolean) => {
+      if (!draft) return;
       setBusy(true);
       setError(null);
       setInfo(null);
       try {
+        const validation = validateSemanticDraft(draft);
+        if (draftDirty) {
+          throw new Error(t("semantic.saveBeforeIndex"));
+        }
+        if (validation.error) {
+          throw new Error(validation.error);
+        }
         await api("/semantic/start", { method: "POST", body: { rebuild: !!rebuild } });
         setInfo(rebuild ? t("semantic.rebuildStarted") : t("semantic.incrementalStarted"));
         await load();
@@ -91,7 +171,7 @@ export function SemanticPanel() {
         setBusy(false);
       }
     },
-    [load],
+    [draft, draftDirty, load],
   );
 
   const stop = useCallback(async () => {
@@ -113,10 +193,11 @@ export function SemanticPanel() {
     setError(null);
     setInfo(t("semantic.startingDaemon"));
     try {
-      const r = await api<{ ready: boolean }>("/semantic/ollama/start", { method: "POST", body: {} });
-      setInfo(
-        r.ready ? t("semantic.daemonUp") : t("semantic.daemonTimeout"),
-      );
+      const r = await api<{ ready: boolean }>("/semantic/ollama/start", {
+        method: "POST",
+        body: {},
+      });
+      setInfo(r.ready ? t("semantic.daemonUp") : t("semantic.daemonTimeout"));
       await load();
     } catch (err) {
       setError((err as Error).message);
@@ -142,10 +223,45 @@ export function SemanticPanel() {
     [load],
   );
 
-  if (!data && !error)
+  const saveProviderConfig = useCallback(async () => {
+    if (!draft) return;
+    setBusy(true);
+    setError(null);
+    setInfo(null);
+    try {
+      const extraBody = semanticValidation.extraBody;
+      await api("/semantic/config", {
+        method: "POST",
+        body: {
+          provider: draft.provider,
+          ollama: {
+            baseUrl: draft.ollama.baseUrl,
+            model: draft.ollama.model,
+          },
+          openaiCompat: {
+            baseUrl: draft.openaiCompat.baseUrl,
+            apiKey: draft.openaiCompat.apiKey,
+            model: draft.openaiCompat.model,
+            extraBody,
+          },
+        },
+      });
+      setDraftDirty(false);
+      draftDirtyRef.current = false;
+      setInfo(t("semantic.savedConfig", { count: 1 }));
+      await load();
+    } catch (err) {
+      setError((err as Error).message);
+    } finally {
+      setBusy(false);
+    }
+  }, [draft, load]);
+
+  if (!data && !error) {
     return html`<div class="card" style="color:var(--fg-3)">${t("common.loading")}</div>`;
+  }
   if (error && !data) return html`<div class="card accent-err">${error}</div>`;
-  if (!data) return null;
+  if (!data || !draft) return null;
 
   if (!data.attached) {
     return html`
@@ -158,43 +274,185 @@ export function SemanticPanel() {
 
   const job = data.job;
   const phase = job?.phase;
-  const running = phase === "scan" || phase === "embed" || phase === "write";
+  const running = isActiveSemanticPhase(phase);
   const pull = data.pull;
   const pulling = pull?.status === "pulling";
-
-  const o = data.ollama ?? {};
-  const binaryFound = o.binaryFound === true;
-  const daemonRunning = o.daemonRunning === true;
-  const modelPulled = o.modelPulled === true;
-  const modelName = o.modelName ?? "nomic-embed-text";
-  const installedModels = o.installedModels ?? [];
-  const ready = binaryFound && daemonRunning && modelPulled;
+  const provider = data.providerStatus?.kind ?? draft.provider;
+  const ready = data.providerStatus?.ready === true;
+  const isOllama = provider === "ollama";
+  const ollama = data.providerStatus?.kind === "ollama" ? data.providerStatus : null;
+  const remote = data.providerStatus?.kind === "openai-compat" ? data.providerStatus : null;
+  const binaryFound = ollama?.binaryFound === true;
+  const daemonRunning = ollama?.daemonRunning === true;
+  const modelPulled = ollama?.modelPulled === true;
+  const modelName = isOllama
+    ? (ollama?.modelName ?? draft.ollama.model ?? "nomic-embed-text")
+    : draft.openaiCompat.model;
 
   const sectionH3 = (text: string) => html`
     <h3 style="margin:18px 0 8px;font-family:var(--font-mono);font-size:11px;color:var(--fg-3);text-transform:uppercase;letter-spacing:.1em">${text}</h3>
   `;
 
   const idx = data.index;
+  const indexReady = idx?.exists === true && idx.compatible !== false;
+  const indexMismatch = idx?.exists === true && idx.compatible === false;
+  const semanticValidation = validateSemanticDraft(draft);
+  const semanticDraftBlocked = draftDirty || semanticValidation.error !== null;
   return html`
     <div style="display:grid;grid-template-columns:minmax(0,1fr) 280px;gap:14px;align-items:start">
       <div style="display:flex;flex-direction:column;gap:10px;min-width:0">
         <div class="chips">
-          <span class=${`chip-f ${idx?.exists ? "active" : ""}`}>
-            ${idx?.exists ? t("semantic.indexBuilt") : t("semantic.noIndex")}
+          <span class=${`chip-f static ${indexReady ? "active" : ""}`}>
+            ${indexReady ? t("semantic.indexBuilt") : t("semantic.noIndex")}
           </span>
           ${
             ready
-              ? html`<span class="chip-f" style="border-color:var(--c-ok);color:var(--c-ok)">${t("semantic.ready")}</span>`
-              : html`<span class="chip-f" style="border-color:var(--c-warn);color:var(--c-warn)">${t("semantic.setupNeeded")}</span>`
+              ? html`<span class="chip-f static" style="border-color:var(--c-ok);color:var(--c-ok)">${t("semantic.ready")}</span>`
+              : html`<span class="chip-f static" style="border-color:var(--c-warn);color:var(--c-warn)">${t("semantic.setupNeeded")}</span>`
           }
         </div>
-        ${info ? html`<div><span class="pill info">${info}</span></div>` : null}
         ${error ? html`<div class="card accent-err">${error}</div>` : null}
 
-        ${idx?.exists ? html`<${SemanticSearchSection} />` : null}
+        <div class="card">
+          <div class="card-h"><span class="title">${t("semantic.provider")}</span></div>
+          <div class="form-row">
+            <span class="lbl">${t("semantic.providerType")}</span>
+            <select
+              class="input mono"
+              value=${draft.provider}
+              onInput=${(e: Event) => {
+                draftDirtyRef.current = true;
+                setDraftDirty(true);
+                setDraft({
+                  ...draft,
+                  provider: (e.target as HTMLSelectElement).value as "ollama" | "openai-compat",
+                });
+              }}
+            >
+              <option value="ollama">Ollama</option>
+              <option value="openai-compat">OpenAI-Compatible</option>
+            </select>
+          </div>
+          ${
+            draft.provider === "ollama"
+              ? html`
+                <div class="form-row">
+                  <span class="lbl">${t("semantic.model")}</span>
+                  <input
+                    class="input mono"
+                    type="text"
+                    value=${draft.ollama.model}
+                    onInput=${(e: Event) => {
+                      draftDirtyRef.current = true;
+                      setDraftDirty(true);
+                      setDraft({
+                        ...draft,
+                        ollama: { ...draft.ollama, model: (e.target as HTMLInputElement).value },
+                      });
+                    }}
+                  />
+                </div>
+              `
+              : html`
+                <div class="form-row">
+                  <span class="lbl">${t("semantic.apiUrl")}</span>
+                  <input
+                    class="input mono"
+                    type="text"
+                    placeholder="https://api.openai.com/v1/embeddings"
+                    value=${draft.openaiCompat.baseUrl}
+                    onInput=${(e: Event) => {
+                      draftDirtyRef.current = true;
+                      setDraftDirty(true);
+                      setDraft({
+                        ...draft,
+                        openaiCompat: {
+                          ...draft.openaiCompat,
+                          baseUrl: (e.target as HTMLInputElement).value,
+                        },
+                      });
+                    }}
+                  />
+                </div>
+                <div class="form-row">
+                  <span class="lbl">${t("semantic.apiKey")}</span>
+                  <input
+                    class="input mono"
+                    type="password"
+                    placeholder=${draft.openaiCompat.apiKeySet ? t("semantic.keepExistingKey") : "sk-..."}
+                    value=${draft.openaiCompat.apiKey}
+                    onInput=${(e: Event) => {
+                      draftDirtyRef.current = true;
+                      setDraftDirty(true);
+                      setDraft({
+                        ...draft,
+                        openaiCompat: {
+                          ...draft.openaiCompat,
+                          apiKey: (e.target as HTMLInputElement).value,
+                        },
+                      });
+                    }}
+                  />
+                  <div style="color:var(--fg-3);font-size:12px">${t("semantic.apiKeyStoredNote")}</div>
+                </div>
+                <div class="form-row">
+                  <span class="lbl">${t("semantic.model")}</span>
+                  <input
+                    class="input mono"
+                    type="text"
+                    value=${draft.openaiCompat.model}
+                    onInput=${(e: Event) => {
+                      draftDirtyRef.current = true;
+                      setDraftDirty(true);
+                      setDraft({
+                        ...draft,
+                        openaiCompat: {
+                          ...draft.openaiCompat,
+                          model: (e.target as HTMLInputElement).value,
+                        },
+                      });
+                    }}
+                  />
+                </div>
+                <details style="margin-top:10px">
+                  <summary style="cursor:pointer;color:var(--fg-2);font-size:12px">${t("semantic.customRequestBody")}</summary>
+                  <div class="form-row" style="margin-top:10px">
+                    <span class="lbl">${t("semantic.customRequestBody")}</span>
+                    <textarea
+                      class="input mono"
+                      rows="6"
+                      value=${draft.openaiCompat.extraBodyText}
+                      onInput=${(e: Event) => {
+                        draftDirtyRef.current = true;
+                        setDraftDirty(true);
+                        setDraft({
+                          ...draft,
+                          openaiCompat: {
+                            ...draft.openaiCompat,
+                            extraBodyText: (e.target as HTMLTextAreaElement).value,
+                          },
+                        });
+                      }}
+                    ></textarea>
+                  </div>
+                </details>
+                ${
+                  semanticValidation.error
+                    ? html`<div style="color:var(--c-err);font-size:12px;margin-top:-2px">${semanticValidation.error}</div>`
+                    : null
+                }
+              `
+          }
+          <div style="display:flex;gap:6px;margin-top:10px">
+            <button class="btn primary" disabled=${busy || semanticValidation.error !== null} onClick=${saveProviderConfig}>${t("common.save")}</button>
+          </div>
+        </div>
+        ${info ? html`<div><span class="pill info">${info}</span></div>` : null}
+
+        ${indexReady ? html`<${SemanticSearchSection} />` : null}
 
         ${
-          !binaryFound
+          isOllama && !binaryFound
             ? html`
               <div class="card">
                 <div class="card-h"><span class="title">${t("semantic.installOllama")}</span></div>
@@ -211,7 +469,7 @@ export function SemanticPanel() {
             : null
         }
         ${
-          binaryFound && !daemonRunning
+          isOllama && binaryFound && !daemonRunning
             ? html`
               <div class="card">
                 <div class="card-h"><span class="title">${t("semantic.daemon")}</span></div>
@@ -227,7 +485,7 @@ export function SemanticPanel() {
             : null
         }
         ${
-          daemonRunning && !modelPulled
+          isOllama && daemonRunning && !modelPulled
             ? html`
               <div class="card">
                 <div class="card-h"><span class="title">${t("semantic.model")}</span></div>
@@ -254,6 +512,18 @@ export function SemanticPanel() {
             `
             : null
         }
+        ${
+          !isOllama
+            ? html`
+              <div class="card">
+                <div class="card-h"><span class="title">${t("semantic.remoteProvider")}</span></div>
+                <div class="card-b" style="font-size:13px;color:var(--fg-2)">
+                  ${t("semantic.remoteProviderDesc")}
+                </div>
+              </div>
+            `
+            : null
+        }
 
         ${
           job
@@ -272,7 +542,9 @@ export function SemanticPanel() {
             <span class="meta">
               ${
                 idx?.exists
-                  ? html`<span class="pill ok">${t("semantic.builtStatus")}</span>`
+                  ? idx.compatible === false
+                    ? html`<span class="pill warn">${t("semantic.incompatibleStatus")}</span>`
+                    : html`<span class="pill ok">${t("semantic.builtStatus")}</span>`
                   : html`<span class="pill">${t("system.none")}</span>`
               }
             </span>
@@ -280,44 +552,56 @@ export function SemanticPanel() {
           ${
             idx?.exists
               ? html`
+                <div class="rail-kv"><span class="k">${t("semantic.provider")}</span><span class="v">${idx.builtWith?.provider ?? idx.provider ?? provider}</span></div>
                 <div class="rail-kv"><span class="k">${t("semantic.chunks")}</span><span class="v">${fmtNum(idx.chunks)}</span></div>
                 <div class="rail-kv"><span class="k">${t("semantic.files")}</span><span class="v">${fmtNum(idx.files)}</span></div>
-                <div class="rail-kv"><span class="k">${t("semantic.model")}</span><span class="v" style="font-size:11px">${idx.model ?? modelName}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.model")}</span><span class="v" style="font-size:11px">${idx.builtWith?.model ?? idx.model ?? modelName}</span></div>
                 <div class="rail-kv"><span class="k">${t("semantic.dim")}</span><span class="v">${fmtNum(idx.dim)}</span></div>
                 <div class="rail-kv"><span class="k">${t("semantic.size")}</span><span class="v">${fmtBytes(idx.sizeBytes)}</span></div>
                 <div class="rail-kv"><span class="k">${t("semantic.lastBuild")}</span><span class="v">${fmtRelativeTime(idx.lastBuiltMs ?? null)}</span></div>
+                ${
+                  idx.compatible === false
+                    ? html`
+                      <div class="rail-kv"><span class="k">${t("semantic.builtWith")}</span><span class="v" style="font-size:11px">${idx.builtWith?.provider} · ${idx.builtWith?.model}</span></div>
+                      <div class="rail-kv"><span class="k">${t("semantic.currentTarget")}</span><span class="v" style="font-size:11px">${idx.current?.provider} · ${idx.current?.model}</span></div>
+                      <div style="color:var(--c-warn);font-size:12px;padding-top:8px">${t("semantic.incompatibleHint")}</div>
+                    `
+                    : null
+                }
               `
-              : html`
-                <div style="color:var(--fg-3);font-size:12.5px;padding:6px 0">
-                  ${t("semantic.runIndexHint")}
-                </div>
-              `
+              : html`<div style="color:var(--fg-3);font-size:12.5px;padding:6px 0">${t("semantic.runIndexHint")}</div>`
           }
           <div style="display:flex;gap:6px;margin-top:10px;flex-wrap:wrap">
-            <button class="primary" disabled=${busy || running || !ready} onClick=${() => start(false)}>${idx?.exists ? t("semantic.reIndex") : t("semantic.build")}</button>
+            <button class="primary" disabled=${busy || running || !ready || semanticDraftBlocked} onClick=${() => start(false)}>${indexReady ? t("semantic.reIndex") : t("semantic.build")}</button>
             ${
               idx?.exists
-                ? html`<button disabled=${busy || running || !ready} onClick=${() => start(true)}>${t("semantic.rebuild")}</button>`
+                ? html`<button disabled=${busy || running || !ready || semanticDraftBlocked} onClick=${() => start(true)}>${t("semantic.rebuild")}</button>`
+                : null
+            }
+            ${
+              running
+                ? html`<button onClick=${stop} style="border-color:var(--c-err);color:var(--c-err)">${t("semantic.stop")}</button>`
                 : null
             }
-            ${running ? html`<button onClick=${stop} style="border-color:var(--c-err);color:var(--c-err)">${t("semantic.stop")}</button>` : null}
           </div>
         </div>
 
         <div class="card">
-          <div class="card-h"><span class="title">${t("semantic.ollama")}</span></div>
-          <div class="rail-kv">
-            <span class="k">${t("semantic.binary")}</span>
-            <span class="v">${binaryFound ? html`<span class="pill ok">${t("semantic.found")}</span>` : html`<span class="pill err">${t("semantic.missing")}</span>`}</span>
-          </div>
-          <div class="rail-kv">
-            <span class="k">${t("semantic.daemonStatus")}</span>
-            <span class="v">${daemonRunning ? html`<span class="pill ok">${t("semantic.up")}</span>` : html`<span class="pill warn">${t("semantic.down")}</span>`}</span>
-          </div>
-          <div class="rail-kv">
-            <span class="k">${t("semantic.model")}</span>
-            <span class="v">${modelPulled ? html`<span class="pill ok">${t("semantic.pulled")}</span>` : html`<span class="pill warn">${t("semantic.missing")}</span>`}</span>
-          </div>
+          <div class="card-h"><span class="title">${isOllama ? t("semantic.ollama") : t("semantic.openaiCompat")}</span></div>
+          ${
+            isOllama
+              ? html`
+                <div class="rail-kv"><span class="k">${t("semantic.binary")}</span><span class="v">${binaryFound ? html`<span class="pill ok">${t("semantic.found")}</span>` : html`<span class="pill err">${t("semantic.missing")}</span>`}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.daemonStatus")}</span><span class="v">${daemonRunning ? html`<span class="pill ok">${t("semantic.up")}</span>` : html`<span class="pill warn">${t("semantic.down")}</span>`}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.model")}</span><span class="v">${modelPulled ? html`<span class="pill ok">${t("semantic.pulled")}</span>` : html`<span class="pill warn">${t("semantic.missing")}</span>`}</span></div>
+              `
+              : html`
+                <div class="rail-kv"><span class="k">${t("semantic.apiUrl")}</span><span class="v" style="font-size:11px;max-width:160px;overflow-wrap:anywhere;word-break:break-word;text-align:right">${remote?.baseUrl ?? draft.openaiCompat.baseUrl}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.apiKey")}</span><span class="v">${remote?.apiKeySet ? html`<span class="pill ok">${t("semantic.found")}</span>` : html`<span class="pill warn">${t("semantic.missing")}</span>`}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.model")}</span><span class="v" style="font-size:11px">${remote?.model ?? draft.openaiCompat.model}</span></div>
+                <div class="rail-kv"><span class="k">${t("semantic.extraBody")}</span><span class="v">${fmtNum(remote?.extraBodyKeys.length ?? 0)}</span></div>
+              `
+          }
         </div>
 
         <${SemanticExcludesCard} />
@@ -326,6 +610,46 @@ export function SemanticPanel() {
   `;
 }
 
+function toConfigDraft(config: SemanticConfigView): SemanticConfigDraft {
+  return {
+    provider: config.provider,
+    ollama: {
+      baseUrl: config.ollama.baseUrl,
+      model: config.ollama.model,
+    },
+    openaiCompat: {
+      baseUrl: config.openaiCompat.baseUrl,
+      apiKey: "",
+      model: config.openaiCompat.model,
+      extraBodyText: JSON.stringify(config.openaiCompat.extraBody ?? {}, null, 2),
+      apiKeySet: config.openaiCompat.apiKeySet,
+    },
+  };
+}
+
+export function validateSemanticDraft(draft: SemanticConfigDraft): SemanticDraftValidation {
+  if (draft.provider !== "openai-compat") {
+    return { extraBody: {}, error: null };
+  }
+  const raw = draft.openaiCompat.extraBodyText.trim();
+  if (!raw) {
+    return { extraBody: {}, error: null };
+  }
+  let parsed: unknown;
+  try {
+    parsed = JSON.parse(raw);
+  } catch (err) {
+    return {
+      extraBody: {},
+      error: t("semantic.invalidCustomRequestBody", { error: (err as Error).message }),
+    };
+  }
+  if (!isPlainObject(parsed)) {
+    return { extraBody: {}, error: t("semantic.customRequestBodyMustBeObject") };
+  }
+  return { extraBody: parsed, error: null };
+}
+
 interface IndexConfig {
   excludeDirs?: string[];
   excludeFiles?: string[];
@@ -367,6 +691,7 @@ interface SearchHit {
 interface SearchResponse {
   hits: SearchHit[];
   elapsedMs: number;
+  provider?: string;
   model: string;
 }
 
@@ -422,13 +747,15 @@ function SemanticSearchSection() {
         hits || busy || error
           ? html`
             <div style="font-family:var(--font-mono);font-size:11px;color:var(--fg-3);margin:8px 0 6px;display:flex;align-items:center;gap:8px">
-              ${busy
-                ? html`<span>${t("semantic.searching")}</span>`
-                : error
-                ? html`<span style="color:var(--c-err)">${error}</span>`
-                : hits
-                ? html`<span>${t("semantic.results", { count: hits.length, s: hits.length === 1 ? "" : "s", ms: meta?.elapsedMs ?? 0, model: meta?.model ?? "" })}</span>`
-                : null}
+              ${
+                busy
+                  ? html`<span>${t("semantic.searching")}</span>`
+                  : error
+                    ? html`<span style="color:var(--c-err)">${error}</span>`
+                    : hits
+                      ? html`<span>${t("semantic.results", { count: hits.length, s: hits.length === 1 ? "" : "s", ms: meta?.elapsedMs ?? 0, model: meta?.model ?? "" })}</span>`
+                      : null
+              }
             </div>
             ${
               hits && hits.length > 0
@@ -449,8 +776,8 @@ function SemanticSearchSection() {
                   </div>
                 `
                 : hits && hits.length === 0 && !busy
-                ? html`<div class="card" style="color:var(--fg-3);font-size:12px">${t("semantic.noMatches")}</div>`
-                : null
+                  ? html`<div class="card" style="color:var(--fg-3);font-size:12px">${t("semantic.noMatches")}</div>`
+                  : null
             }
           `
           : null
@@ -570,11 +897,7 @@ function SemanticExcludesCard() {
       <div class="card-h">
         <span class="title">${t("semantic.indexConfig")}</span>
         <span class="meta">
-          <a
-            class="mono"
-            style="color:var(--c-brand);text-decoration:none;font-size:11px;cursor:pointer"
-            onClick=${reset}
-          >${t("semantic.reset")}</a>
+          <a class="mono" style="color:var(--c-brand);text-decoration:none;font-size:11px;cursor:pointer" onClick=${reset}>${t("semantic.reset")}</a>
         </span>
       </div>
       ${info ? html`<div style="margin-bottom:8px"><span class="pill ok">${info}</span></div>` : null}
@@ -606,11 +929,7 @@ function SemanticExcludesCard() {
         placeholder="**/*.test.ts"
       />
 
-      <div
-        class="checkbox-row"
-        style="margin-top:8px;cursor:pointer"
-        onClick=${() => setDraft({ ...draft, respectGitignore: !draft.respectGitignore })}
-      >
+      <div class="checkbox-row" style="margin-top:8px;cursor:pointer" onClick=${() => setDraft({ ...draft, respectGitignore: !draft.respectGitignore })}>
         <span class=${`box ${draft.respectGitignore ? "on" : ""}`}>${draft.respectGitignore ? "✓" : ""}</span>
         <span>${t("semantic.respectGitignore")}</span>
       </div>
@@ -623,17 +942,14 @@ function SemanticExcludesCard() {
           min="1024"
           step="1024"
           value=${draft.maxFileBytes}
-          onInput=${(e: Event) =>
-            setDraft({ ...draft, maxFileBytes: Number((e.target as HTMLInputElement).value) || 0 })}
+          onInput=${(e: Event) => setDraft({ ...draft, maxFileBytes: Number((e.target as HTMLInputElement).value) || 0 })}
           style="font-size:12px"
         />
         <span class="help">${t("semantic.skipLarger", { size: (draft.maxFileBytes / 1024 / 1024).toFixed(1) })}</span>
       </div>
 
       <div style="display:flex;gap:6px;margin-top:10px">
-        <button class="btn ghost" style="flex:1" disabled=${busy} onClick=${runPreview}>
-          <span class="g">⊕</span><span>${t("semantic.preview")}</span>
-        </button>
+        <button class="btn ghost" style="flex:1" disabled=${busy} onClick=${runPreview}><span class="g">⊕</span><span>${t("semantic.preview")}</span></button>
         <button class="btn primary" style="flex:1" disabled=${busy} onClick=${save}>${t("common.save")}</button>
       </div>
 
@@ -659,26 +975,24 @@ function ExcludesPreview({ preview }: { preview: PreviewData }) {
   ].filter((k) => (buckets[k] || 0) > 0);
   return html`
     <div class="excludes-preview">
-      <div class="summary">
-        ${t("semantic.previewSummary", { included: preview.filesIncluded, skipped: totalSkipped })}
-      </div>
+      <div class="summary">${t("semantic.previewSummary", { included: preview.filesIncluded, skipped: totalSkipped })}</div>
       ${
         reasons.length === 0
           ? html`<div style="color:var(--fg-3)">${t("semantic.nothingSkipped")}</div>`
           : reasons.map(
               (r) => html`
-                <details>
-                  <summary><strong>${r}: ${buckets[r]}</strong></summary>
-                  <ul>
-                    ${(samples[r] || []).map((p) => html`<li><code>${p}</code></li>`)}
-                    ${
-                      (buckets[r] || 0) > (samples[r] || []).length
-                        ? html`<li style="color:var(--fg-3)">…${(buckets[r] || 0) - (samples[r] || []).length} more</li>`
-                        : null
-                    }
-                  </ul>
-                </details>
-              `,
+              <details>
+                <summary><strong>${r}: ${buckets[r]}</strong></summary>
+                <ul>
+                  ${(samples[r] || []).map((p) => html`<li><code>${p}</code></li>`)}
+                  ${
+                    (buckets[r] || 0) > (samples[r] || []).length
+                      ? html`<li style="color:var(--fg-3)">…${(buckets[r] || 0) - (samples[r] || []).length} more</li>`
+                      : null
+                  }
+                </ul>
+              </details>
+            `,
             )
       }
       ${
@@ -723,14 +1037,11 @@ function ChipFormRow({
   };
   return html`
     <div class="form-row">
-      <span class="lbl">
-        ${label}
-        ${sub ? html`<span style="color:var(--fg-3);font-weight:400;text-transform:none;letter-spacing:0"> · ${sub}</span>` : null}
-      </span>
+      <span class="lbl">${label}${sub ? html`<span style="color:var(--fg-3);font-weight:400;text-transform:none;letter-spacing:0"> · ${sub}</span>` : null}</span>
       <div style="display:flex;flex-wrap:wrap;gap:4px">
         ${value.map(
           (e) => html`
-            <span class="chip-f">
+            <span class="chip-f static">
               <span>${e}</span>
               <span class="x" style="cursor:pointer" onClick=${() => remove(e)} title="remove">×</span>
             </span>
@@ -758,30 +1069,38 @@ function ChipFormRow({
 function SemanticJobView({ job, running }: { job: SemanticJob; running: boolean }) {
   useLang();
   const phaseLabel =
-    ({
-      scan: t("semantic.phaseScan"),
-      embed: t("semantic.phaseEmbed"),
-      write: t("semantic.phaseWrite"),
-      done: t("semantic.phaseDone"),
-      error: t("semantic.phaseError"),
-    } as Record<string, string>)[job.phase] ?? job.phase;
+    (
+      {
+        setup: t("semantic.phaseSetup"),
+        scan: t("semantic.phaseScan"),
+        embed: t("semantic.phaseEmbed"),
+        write: t("semantic.phaseWrite"),
+        done: t("semantic.phaseDone"),
+        error: t("semantic.phaseError"),
+        cancelled: t("semantic.phaseCancelled"),
+      } as Record<string, string>
+    )[job.phase] ?? job.phase;
   const total = job.chunksTotal ?? 0;
   const doneN = job.chunksDone ?? 0;
   const ratio = total > 0 ? Math.min(1, doneN / total) : 0;
-  const elapsed = ((Date.now() - job.startedAt) / 1000).toFixed(1);
+  const elapsedBase = job.finishedAt ?? Date.now();
+  const elapsedSeconds = (elapsedBase - job.startedAt) / 1000;
+  const elapsed = elapsedSeconds < 0.1 ? "<0.1s" : `${elapsedSeconds.toFixed(1)}s`;
+  const phaseSummary =
+    job.phase === "error" && job.lastPhase === "setup"
+      ? t("semantic.setupFailed")
+      : phaseLabel;
 
   return html`
     <div class="kv">
       <div><span class="kv-key">phase</span>
-        <span class=${`pill ${job.phase === "error" ? "pill-err" : running ? "pill-active" : "pill-dim"}`}>${phaseLabel}</span>
-        ${job.aborted ? html`<span class="pill warn" style="margin-left: 6px;">${t("semantic.stopping")}</span>` : null}
-        <span style="color:var(--fg-3);margin-left:8px">${elapsed}s</span>
+        <span class=${`pill ${job.phase === "error" ? "pill-err" : job.phase === "cancelled" ? "warn" : running ? "pill-active" : "pill-dim"}`}>${phaseSummary}</span>
+        ${job.aborted && running ? html`<span class="pill warn" style="margin-left: 6px;">${t("semantic.stopping")}</span>` : null}
+        <span style="color:var(--fg-3);margin-left:8px">${elapsed}</span>
       </div>
       ${
         job.filesScanned !== null && job.filesScanned !== undefined
-          ? html`<div><span class="kv-key">${t("semantic.files")}</span>${t("semantic.scanned", { count: job.filesScanned })}${
-              job.filesChanged != null ? ` · ${t("semantic.changed", { count: job.filesChanged })}` : ""
-            }${job.filesSkipped ? ` · ${t("semantic.skipped", { count: job.filesSkipped })}` : ""}</div>`
+          ? html`<div><span class="kv-key">${t("semantic.files")}</span>${t("semantic.scanned", { count: job.filesScanned })}${job.filesChanged != null ? ` · ${t("semantic.changed", { count: job.filesChanged })}` : ""}${job.filesSkipped ? ` · ${t("semantic.skipped", { count: job.filesSkipped })}` : ""}</div>`
           : null
       }
       ${
@@ -796,23 +1115,13 @@ function SemanticJobView({ job, running }: { job: SemanticJob; running: boolean
           `
           : null
       }
-      ${
-        job.error
-          ? html`<div><span class="kv-key">${t("semantic.phaseError")}</span><span class="err">${job.error}</span></div>`
-          : null
-      }
+      ${job.error ? html`<div><span class="kv-key">${t("semantic.phaseError")}</span><span class="err">${job.error}</span></div>` : null}
       ${
         job.result
-          ? html`<div><span class="kv-key">${t("semantic.result")}</span>${t("semantic.added", { count: job.result.chunksAdded })} · ${t("semantic.removed", { count: job.result.chunksRemoved })}${
-              job.result.chunksSkipped ? ` · ${t("semantic.failed", { count: job.result.chunksSkipped })}` : ""
-            } · ${(job.result.durationMs / 1000).toFixed(1)}s</div>`
-          : null
-      }
-      ${
-        job.result?.skipBuckets
-          ? html`<${SkipBucketsView} buckets=${job.result.skipBuckets} />`
+          ? html`<div><span class="kv-key">${t("semantic.result")}</span>${t("semantic.added", { count: job.result.chunksAdded })} · ${t("semantic.removed", { count: job.result.chunksRemoved })}${job.result.chunksSkipped ? ` · ${t("semantic.failed", { count: job.result.chunksSkipped })}` : ""} · ${(job.result.durationMs / 1000).toFixed(1)}s</div>`
           : null
       }
+      ${job.result?.skipBuckets ? html`<${SkipBucketsView} buckets=${job.result.skipBuckets} />` : null}
     </div>
   `;
 }
@@ -836,3 +1145,13 @@ function SkipBucketsView({ buckets }: { buckets: Record<string, number> }) {
     .map(([k, label]) => `${label}: ${buckets[k]}`);
   return html`<div><span class="kv-key">${t("semantic.skipped")}</span>${t("semantic.skippedFiles", { total, details: parts.join(", ") })}</div>`;
 }
+
+function isActiveSemanticPhase(phase: string | undefined): boolean {
+  return phase === "setup" || phase === "scan" || phase === "embed" || phase === "write";
+}
+
+function isPlainObject(value: unknown): value is Record<string, unknown> {
+  if (value === null || typeof value !== "object" || Array.isArray(value)) return false;
+  const proto = Object.getPrototypeOf(value);
+  return proto === Object.prototype || proto === null;
+}

dashboard/src/panels/sessions.ts

@@ -68,7 +68,7 @@ export function SessionsPanel() {
           />
         </div>
         <div class="chips" style="padding:0 12px 8px">
-          <span class="chip-f active">${t("common.all")} <span class="ct">${sessions.length}</span></span>
+          <span class="chip-f static active">${t("common.all")} <span class="ct">${sessions.length}</span></span>
         </div>
         <div class="ssl-rows">
           ${filtered.map(

dashboard/src/panels/settings.ts

@@ -1,6 +1,20 @@
-import { useCallback, useEffect, useState } from "preact/hooks";
+import { useCallback, useEffect, useRef, useState } from "preact/hooks";
 import { api } from "../lib/api.js";
+import {
+  type BudgetState,
+  QUICK_CAPS_USD,
+  budgetTone,
+  bumpSuggestions,
+  deriveBudgetState,
+} from "../lib/budget.js";
 import { html } from "../lib/html.js";
+import {
+  INTERVAL_PRESETS_MS,
+  type IntervalUnit,
+  type LoopRunStatus,
+  formatRemaining,
+  parseCustomInterval,
+} from "../lib/loop-control.js";
 import { type DashboardLang, getLang, setLang, t, useLang } from "../i18n/index.js";
 
 interface SettingsData {
@@ -11,6 +25,364 @@ interface SettingsData {
   search?: boolean;
   model?: string;
   editMode?: string;
+  proNext?: boolean;
+  budgetUsd?: number | null;
+  /** Cumulative session spend (USD); null when no session is attached. */
+  sessionSpendUsd?: number | null;
+}
+
+function fmtUsd2(n: number): string {
+  return `$${n.toFixed(n < 1 ? 4 : 2)}`;
+}
+
+interface ModelPriceEntry {
+  inputCacheHit: number;
+  inputCacheMiss: number;
+  output: number;
+}
+
+interface ModelCatalog {
+  models: string[] | null;
+  current: string | null;
+  pricing: Record<string, ModelPriceEntry>;
+}
+
+function formatPricing(p: ModelPriceEntry | undefined): string | null {
+  if (!p) return null;
+  return t("settings.modelPricingLine", {
+    hit: p.inputCacheHit.toFixed(3),
+    miss: p.inputCacheMiss.toFixed(3),
+    out: p.output.toFixed(3),
+  });
+}
+
+function ModelRow({
+  current,
+  catalog,
+  saving,
+  onPick,
+}: {
+  current: string;
+  catalog: ModelCatalog | null;
+  saving: boolean;
+  onPick: (model: string) => void;
+}) {
+  const list = catalog?.models ?? null;
+  const ready = list && list.length > 0;
+  if (!ready) {
+    // Fallback: catalog hasn't loaded (or API failed). Read-only — same as before D-4.
+    return html`<code class="mono">${current ?? "—"}</code>`;
+  }
+  // Ensure the live model is selectable even if the catalog hasn't reported it
+  // yet (preset overrides, custom IDs).
+  const options = list.includes(current) ? list : [current, ...list];
+  const price = catalog?.pricing[current];
+  return html`
+    <span style="display:inline-flex;flex-direction:column;gap:4px">
+      <select
+        value=${current}
+        onChange=${(e: Event) => {
+          const next = (e.target as HTMLSelectElement).value;
+          if (next && next !== current) onPick(next);
+        }}
+        disabled=${saving}
+        style="font-family:var(--font-mono);min-width:200px"
+      >
+        ${options.map((m) => html`<option key=${m} value=${m}>${m}</option>`)}
+      </select>
+      ${
+        price
+          ? html`<span style="color:var(--fg-3);font-size:11px;font-family:var(--font-mono)">${formatPricing(price)}</span>`
+          : null
+      }
+    </span>
+  `;
+}
+
+function BudgetGauge({ state }: { state: BudgetState }) {
+  if (state.kind === "off") return null;
+  const tone = budgetTone(state);
+  const fill = Math.min(100, state.pct);
+  const valueColor =
+    tone === "err"
+      ? "color:var(--c-err)"
+      : tone === "warn"
+        ? "color:var(--c-warn)"
+        : "color:var(--fg-1)";
+  return html`
+    <div style="display:flex;flex-direction:column;gap:6px">
+      <div style="display:flex;justify-content:space-between;align-items:baseline;font-size:13px">
+        <span style=${valueColor}>
+          <strong style="font-family:var(--font-mono)">${fmtUsd2(state.spent)}</strong>
+          <span style="color:var(--fg-3)"> ${t("settings.budgetOf")} </span>
+          <strong style="font-family:var(--font-mono)">${fmtUsd2(state.cap)}</strong>
+        </span>
+        <span style=${`font-family:var(--font-mono);font-size:11px;${valueColor}`}>${state.pct.toFixed(1)}%</span>
+      </div>
+      <div class=${`progress ${tone}`}><div class="progress-fill" style=${`width:${fill}%`}></div></div>
+      <span style="color:var(--fg-3);font-size:11px">
+        ${
+          state.kind === "exhausted"
+            ? t("settings.budgetRefusing")
+            : state.kind === "warn"
+              ? t("settings.budgetWarnLine")
+              : t("settings.budgetIdleLine")
+        }
+      </span>
+    </div>
+  `;
+}
+
+interface BudgetSectionProps {
+  state: BudgetState;
+  saving: boolean;
+  onSetCap: (usd: number) => void;
+  onClear: () => void;
+}
+
+function BudgetSection({ state, saving, onSetCap, onClear }: BudgetSectionProps) {
+  const [custom, setCustom] = useState("");
+  const submitCustom = () => {
+    const n = Number.parseFloat(custom);
+    if (Number.isFinite(n) && n > 0) {
+      onSetCap(n);
+      setCustom("");
+    }
+  };
+
+  const quickButtons = (caps: ReadonlyArray<number>) =>
+    caps.map(
+      (c) => html`
+        <button
+          key=${c}
+          class="btn"
+          style="font-family:var(--font-mono)"
+          disabled=${saving}
+          onClick=${() => onSetCap(c)}
+        >$${c}</button>
+      `,
+    );
+
+  const customField = html`
+    <span style="display:inline-flex;align-items:center;gap:4px;margin-left:auto">
+      <span style="color:var(--fg-3);font-size:11px">${t("settings.budgetCustom")}</span>
+      <input
+        type="number"
+        min="0.01"
+        step="0.01"
+        value=${custom}
+        placeholder="0.00"
+        onInput=${(e: Event) => setCustom((e.target as HTMLInputElement).value)}
+        onKeyDown=${(e: KeyboardEvent) => {
+          if (e.key === "Enter") submitCustom();
+        }}
+        style="width:72px;font-family:var(--font-mono)"
+        disabled=${saving}
+      />
+      <button
+        class="btn primary"
+        disabled=${saving || !(Number.parseFloat(custom) > 0)}
+        onClick=${submitCustom}
+      >→</button>
+    </span>
+  `;
+
+  return html`
+    <div class="card" style="display:flex;flex-direction:column;gap:12px">
+      <${BudgetGauge} state=${state} />
+
+      ${
+        state.kind === "off"
+          ? html`
+              <div>
+                <div style="color:var(--fg-3);font-size:11px;margin-bottom:6px">${t("settings.budgetSetCap")}</div>
+                <div style="display:flex;gap:6px;align-items:center;flex-wrap:wrap">
+                  ${quickButtons(QUICK_CAPS_USD)}
+                  ${customField}
+                </div>
+              </div>
+            `
+          : state.kind === "warn" || state.kind === "exhausted"
+            ? html`
+                <div>
+                  <div style="color:var(--fg-3);font-size:11px;margin-bottom:6px">${t("settings.budgetBumpHint")}</div>
+                  <div style="display:flex;gap:6px;align-items:center;flex-wrap:wrap">
+                    ${bumpSuggestions(state.cap).map(
+                      (next) => html`
+                        <button
+                          key=${next}
+                          class="btn primary"
+                          style="font-family:var(--font-mono)"
+                          disabled=${saving}
+                          onClick=${() => onSetCap(next)}
+                        >→ $${next % 1 === 0 ? next : next.toFixed(2)}</button>
+                      `,
+                    )}
+                    ${customField}
+                  </div>
+                  <div style="margin-top:8px">
+                    <button class="btn" disabled=${saving} onClick=${onClear}>${t("settings.budgetClear")}</button>
+                  </div>
+                </div>
+              `
+            : html`
+                <div style="display:flex;gap:6px;align-items:center;flex-wrap:wrap">
+                  ${bumpSuggestions(state.cap).map(
+                    (next) => html`
+                      <button
+                        key=${next}
+                        class="btn"
+                        style="font-family:var(--font-mono)"
+                        disabled=${saving}
+                        onClick=${() => onSetCap(next)}
+                      >→ $${next % 1 === 0 ? next : next.toFixed(2)}</button>
+                    `,
+                  )}
+                  ${customField}
+                  <button
+                    class="btn"
+                    style="margin-left:8px"
+                    disabled=${saving}
+                    onClick=${onClear}
+                  >${t("settings.budgetClear")}</button>
+                </div>
+              `
+      }
+    </div>
+  `;
+}
+
+interface LoopSectionProps {
+  status: LoopRunStatus | null;
+  /** ms remaining until next fire — ticks down client-side between status polls. */
+  remainingMs: number;
+  /** Last-turn cost in USD; used as a hint for "each iteration costs ~". */
+  avgIterCostUsd: number | null;
+  busy: boolean;
+  onStart: (intervalMs: number, prompt: string) => void;
+  onStop: () => void;
+}
+
+function LoopSection({
+  status,
+  remainingMs,
+  avgIterCostUsd,
+  busy,
+  onStart,
+  onStop,
+}: LoopSectionProps) {
+  const [intervalMs, setIntervalMs] = useState<number>(INTERVAL_PRESETS_MS[1]!.ms);
+  const [prompt, setPrompt] = useState("");
+  const [customValue, setCustomValue] = useState("");
+  const [customUnit, setCustomUnit] = useState<IntervalUnit>("m");
+
+  if (status) {
+    return html`
+      <div class="card" style="display:flex;flex-direction:column;gap:10px">
+        <div style="display:flex;justify-content:space-between;align-items:baseline">
+          <span style="color:var(--c-warn);font-family:var(--font-mono);font-size:11px">⟳ ${t("settings.loopRunning")}</span>
+          <span style="color:var(--fg-3);font-size:11px">
+            ${t("settings.loopIter", { iter: status.iter })} · ${t("settings.loopFiresIn", { remaining: formatRemaining(remainingMs) })}
+          </span>
+        </div>
+        <div style="background:var(--bg-elev-2);border:1px solid var(--bd);border-radius:var(--r);padding:8px 10px;font-family:var(--font-mono);font-size:12px;color:var(--fg-1);white-space:pre-wrap;max-height:120px;overflow-y:auto">${status.prompt}</div>
+        <div>
+          <button class="btn danger" disabled=${busy} onClick=${onStop}>${t("settings.loopStop")}</button>
+        </div>
+      </div>
+    `;
+  }
+
+  const customMs = parseCustomInterval(customValue, customUnit);
+  const canStart = !busy && intervalMs > 0 && prompt.trim().length > 0;
+
+  return html`
+    <div class="card" style="display:flex;flex-direction:column;gap:10px">
+      <div style="color:var(--fg-3);font-size:11px">
+        ${t("settings.loopIdleHint")}
+        ${
+          typeof avgIterCostUsd === "number" && avgIterCostUsd > 0
+            ? html` ${t("settings.loopCostHint", { cost: `$${avgIterCostUsd.toFixed(4)}` })}`
+            : null
+        }
+      </div>
+      <div style="display:flex;flex-direction:column;gap:6px">
+        <span style="color:var(--fg-3);font-size:11px">${t("settings.loopInterval")}</span>
+        <div style="display:flex;gap:6px;align-items:center;flex-wrap:wrap">
+          ${INTERVAL_PRESETS_MS.map(
+            (p) => html`
+              <button
+                key=${p.ms}
+                class=${`btn ${intervalMs === p.ms && customValue === "" ? "primary" : ""}`}
+                style="font-family:var(--font-mono)"
+                disabled=${busy}
+                onClick=${() => {
+                  setIntervalMs(p.ms);
+                  setCustomValue("");
+                }}
+              >${p.label}</button>
+            `,
+          )}
+          <span style="display:inline-flex;align-items:center;gap:4px;margin-left:auto">
+            <span style="color:var(--fg-3);font-size:11px">${t("settings.loopCustom")}</span>
+            <input
+              type="number"
+              min="1"
+              step="1"
+              value=${customValue}
+              onInput=${(e: Event) => {
+                const raw = (e.target as HTMLInputElement).value;
+                setCustomValue(raw);
+                const ms = parseCustomInterval(raw, customUnit);
+                if (ms !== null) setIntervalMs(ms);
+              }}
+              style="width:64px;font-family:var(--font-mono)"
+              disabled=${busy}
+            />
+            <select
+              value=${customUnit}
+              onChange=${(e: Event) => {
+                const next = (e.target as HTMLSelectElement).value as IntervalUnit;
+                setCustomUnit(next);
+                if (customValue) {
+                  const ms = parseCustomInterval(customValue, next);
+                  if (ms !== null) setIntervalMs(ms);
+                }
+              }}
+              disabled=${busy}
+            >
+              <option value="s">s</option>
+              <option value="m">m</option>
+              <option value="h">h</option>
+            </select>
+          </span>
+        </div>
+        ${
+          customValue && customMs === null
+            ? html`<span style="color:var(--c-err);font-size:11px">${t("settings.loopRangeError")}</span>`
+            : null
+        }
+      </div>
+      <div style="display:flex;flex-direction:column;gap:6px">
+        <span style="color:var(--fg-3);font-size:11px">${t("settings.loopPrompt")}</span>
+        <textarea
+          rows="3"
+          placeholder=${t("settings.loopPromptPlaceholder")}
+          value=${prompt}
+          onInput=${(e: Event) => setPrompt((e.target as HTMLTextAreaElement).value)}
+          style="width:100%;font-family:var(--font-mono);resize:vertical"
+          disabled=${busy}
+        ></textarea>
+      </div>
+      <div>
+        <button
+          class="btn primary"
+          disabled=${!canStart}
+          onClick=${() => onStart(intervalMs, prompt.trim())}
+        >${t("settings.loopStart")}</button>
+      </div>
+    </div>
+  `;
 }
 
 export function SettingsPanel() {
@@ -20,6 +392,13 @@ export function SettingsPanel() {
   const [saving, setSaving] = useState(false);
   const [saved, setSaved] = useState<string | null>(null);
   const [draft, setDraft] = useState<Partial<SettingsData>>({});
+  const [catalog, setCatalog] = useState<ModelCatalog | null>(null);
+  const [loopStatus, setLoopStatus] = useState<LoopRunStatus | null>(null);
+  const [loopAvgCost, setLoopAvgCost] = useState<number | null>(null);
+  const [loopBusy, setLoopBusy] = useState(false);
+  /** Wall-clock time of the last status sync — used to interpolate the countdown. */
+  const lastStatusSyncRef = useRef<number>(0);
+  const [now, setNow] = useState<number>(() => Date.now());
 
   const load = useCallback(async () => {
     try {
@@ -33,6 +412,73 @@ export function SettingsPanel() {
   useEffect(() => {
     load();
   }, [load]);
+  useEffect(() => {
+    api<ModelCatalog>("/models")
+      .then(setCatalog)
+      .catch(() => undefined);
+  }, []);
+
+  const refreshLoop = useCallback(async () => {
+    try {
+      const r = await api<{ status: LoopRunStatus | null }>("/loop/status");
+      setLoopStatus(r.status);
+      lastStatusSyncRef.current = Date.now();
+    } catch {
+      /* ignore — status is best-effort */
+    }
+    try {
+      const r = await api<{ stats?: { lastTurnCostUsd?: number } }>("/overview");
+      setLoopAvgCost(r.stats?.lastTurnCostUsd ?? null);
+    } catch {
+      /* ignore */
+    }
+  }, []);
+  useEffect(() => {
+    let cancelled = false;
+    refreshLoop();
+    const id = setInterval(() => {
+      if (!cancelled) refreshLoop();
+    }, 5_000);
+    return () => {
+      cancelled = true;
+      clearInterval(id);
+    };
+  }, [refreshLoop]);
+  useEffect(() => {
+    if (!loopStatus) return;
+    const id = setInterval(() => setNow(Date.now()), 1_000);
+    return () => clearInterval(id);
+  }, [loopStatus]);
+
+  const remainingMs = loopStatus
+    ? Math.max(0, loopStatus.nextFireMs - (now - lastStatusSyncRef.current))
+    : 0;
+
+  const startLoop = useCallback(
+    async (intervalMs: number, prompt: string) => {
+      setLoopBusy(true);
+      try {
+        await api("/loop/start", { method: "POST", body: { intervalMs, prompt } });
+        await refreshLoop();
+      } catch (err) {
+        setError((err as Error).message);
+      } finally {
+        setLoopBusy(false);
+      }
+    },
+    [refreshLoop],
+  );
+  const stopLoop = useCallback(async () => {
+    setLoopBusy(true);
+    try {
+      await api("/loop/stop", { method: "POST" });
+      await refreshLoop();
+    } catch (err) {
+      setError((err as Error).message);
+    } finally {
+      setLoopBusy(false);
+    }
+  }, [refreshLoop]);
 
   const save = useCallback(
     async (fields: Partial<SettingsData>) => {
@@ -189,11 +635,50 @@ export function SettingsPanel() {
         )}
       </div>
 
+      ${sectionH3(t("settings.sectionCompute"))}
+      <div class="card">
+        ${fieldRow(
+          t("settings.proNext"),
+          html`
+            <button
+              class=${`btn ${v.proNext ? "primary" : ""}`}
+              onClick=${() => save({ proNext: !v.proNext })}
+              disabled=${saving}
+            >${v.proNext ? t("settings.proArmed") : t("settings.proArm")}</button>
+          `,
+          t("settings.proNextNote"),
+        )}
+      </div>
+
+      ${sectionH3(t("settings.sectionBudget"))}
+      <${BudgetSection}
+        state=${deriveBudgetState(v.budgetUsd, v.sessionSpendUsd)}
+        saving=${saving}
+        onSetCap=${(usd: number) => save({ budgetUsd: usd })}
+        onClear=${() => save({ budgetUsd: null })}
+      />
+
+      ${sectionH3(t("settings.sectionLoop"))}
+      <${LoopSection}
+        status=${loopStatus}
+        remainingMs=${remainingMs}
+        avgIterCostUsd=${loopAvgCost}
+        busy=${loopBusy}
+        onStart=${startLoop}
+        onStop=${stopLoop}
+      />
+
       ${sectionH3(t("settings.sectionRuntime"))}
       <div class="card">
         ${fieldRow(
           t("settings.activeModel"),
-          html`<code class="mono">${v.model ?? "—"}</code>`,
+          html`<${ModelRow}
+            current=${v.model ?? "—"}
+            catalog=${catalog}
+            saving=${saving}
+            onPick=${(m: string) => save({ model: m })}
+          />`,
+          t("settings.appliesNextTurn"),
         )}
         ${fieldRow(
           t("settings.editMode"),

dashboard/src/panels/skills.ts

@@ -29,6 +29,7 @@ export function SkillsPanel() {
   const [newName, setNewName] = useState("");
   const [newScope, setNewScope] = useState<"global" | "project">("global");
   const [filter, setFilter] = useState("");
+  const [scopeFilter, setScopeFilter] = useState<"all" | Scope>("all");
 
   const load = useCallback(async () => {
     try {
@@ -120,13 +121,15 @@ export function SkillsPanel() {
     ...data.global.map((s) => ({ scope: "global" as Scope, ...s })),
     ...data.builtin.map((s) => ({ scope: "builtin" as Scope, ...s })),
   ];
+  const scopeFiltered =
+    scopeFilter === "all" ? allWith : allWith.filter((s) => s.scope === scopeFilter);
   const filtered = filter.trim()
-    ? allWith.filter(
+    ? scopeFiltered.filter(
         (s) =>
           s.name.toLowerCase().includes(filter.toLowerCase()) ||
           (s.description ?? "").toLowerCase().includes(filter.toLowerCase()),
       )
-    : allWith;
+    : scopeFiltered;
 
   return html`
     <div class="sessions-grid">
@@ -141,10 +144,22 @@ export function SkillsPanel() {
           />
         </div>
         <div class="chips" style="padding:0 12px 8px">
-          <span class="chip-f active">${t("common.all")} <span class="ct">${allWith.length}</span></span>
-          <span class="chip-f">${t("skills.project")} <span class="ct">${data.project.length}</span></span>
-          <span class="chip-f">${t("skills.global")} <span class="ct">${data.global.length}</span></span>
-          <span class="chip-f">${t("skills.builtin")} <span class="ct">${data.builtin.length}</span></span>
+          <span
+            class=${`chip-f ${scopeFilter === "all" ? "active" : ""}`}
+            onClick=${() => setScopeFilter("all")}
+          >${t("common.all")} <span class="ct">${allWith.length}</span></span>
+          <span
+            class=${`chip-f ${scopeFilter === "project" ? "active" : ""}`}
+            onClick=${() => setScopeFilter("project")}
+          >${t("skills.project")} <span class="ct">${data.project.length}</span></span>
+          <span
+            class=${`chip-f ${scopeFilter === "global" ? "active" : ""}`}
+            onClick=${() => setScopeFilter("global")}
+          >${t("skills.global")} <span class="ct">${data.global.length}</span></span>
+          <span
+            class=${`chip-f ${scopeFilter === "builtin" ? "active" : ""}`}
+            onClick=${() => setScopeFilter("builtin")}
+          >${t("skills.builtin")} <span class="ct">${data.builtin.length}</span></span>
         </div>
 
         <div style="padding:0 12px 8px;display:flex;gap:6px;flex-wrap:wrap">

dashboard/src/panels/tools.ts

@@ -37,8 +37,8 @@ export function ToolsPanel() {
   return html`
     <div style="display:flex;flex-direction:column;gap:14px">
       <div class="chips">
-        <span class="chip-f active">${t("common.all")} <span class="ct">${d.total}</span></span>
-        ${d.planMode ? html`<span class="chip-f" style="border-color:var(--c-warn);color:var(--c-warn)">${t("tools.planMode")}</span>` : null}
+        <span class="chip-f static active">${t("common.all")} <span class="ct">${d.total}</span></span>
+        ${d.planMode ? html`<span class="chip-f static" style="border-color:var(--c-warn);color:var(--c-warn)">${t("tools.planMode")}</span>` : null}
       </div>
 
       ${

dashboard/src/panels/usage.ts

@@ -174,8 +174,8 @@ export function UsagePanel() {
   return html`
     <div style="display:flex;flex-direction:column;gap:6px">
       <div class="chips">
-        <span class="chip-f active">${t("usage.records", { count: u.recordCount.toLocaleString() })}</span>
-        <span class="chip-f">${u.logSize}</span>
+        <span class="chip-f static active">${t("usage.records", { count: u.recordCount.toLocaleString() })}</span>
+        <span class="chip-f static">${u.logSize}</span>
       </div>
 
       ${

docs/ARCHITECTURE.md

@@ -43,27 +43,30 @@ timestamps each turn — cache hit rate in practice: <20%.
 **Metric.** `prompt_cache_hit_tokens / (hit + miss)` exposed per-turn and
 aggregated per-session. Visible in the TUI's top-bar cache cell.
 
-### Pillar 2 — R1 Thought Harvesting *(opt-in)*
+#### Parallel tool dispatch
 
-**Problem.** R1 / V4-thinking emits extensive `reasoning_content`. DeepSeek's
-own docs recommend *not* feeding it back to the next turn. Most frameworks
-display it to the user and discard. The planning signal inside is lost.
+Each tool declares `parallelSafe?: boolean` (default `false`). The loop
+dispatcher groups consecutive parallel-safe calls into chunks and races
+them via `Promise.allSettled`; the first non-parallel-safe call ends the
+chunk and runs alone (serial barrier — read-after-write order
+preserved). Tool-result yields and history append still land in declared
+order regardless of which call settles first, so the model sees the
+same shape it would under a fully serial dispatch.
 
-**Solution.** A two-stage process:
+| Env var | Default | Effect |
+|---|---|---|
+| `REASONIX_PARALLEL_MAX` | `3` (hard cap `16`) | Max chunk size. |
+| `REASONIX_TOOL_DISPATCH=serial` | unset | Forces serial dispatch — escape hatch. |
 
-```
-R1 output → Harvester (v4-flash, thinking off) → TypedPlanState
-                                                  ├─ subgoals: string[]
-                                                  ├─ hypotheses: string[]
-                                                  ├─ uncertainties: string[]
-                                                  └─ rejectedPaths: string[]
-```
+Built-in opt-ins: read-only filesystem (`read_file`, `list_directory`,
+`directory_tree`, `search_files`, `search_content`, `get_file_info`),
+web (`web_search`, `web_fetch`), `recall_memory`, `semantic_search`,
+isolated child loops (`run_skill`, `spawn_subagent`), in-memory job
+queries (`job_output`, `list_jobs`). Mutating / side-effecting tools
+stay default. MCP-bridged tools default `false` — third-party tools
+opt in only when the server explicitly declares parallel safety.
 
-The harvester is a cheap flash call with a strict JSON schema. Output is
-validated at runtime. **Not enabled in any preset** (see Pillar 4) — the
-additional round-trip rarely pays back, so it's `/harvest on` opt-in.
-
-### Pillar 3 — Tool-Call Repair
+### Pillar 2 — Tool-Call Repair
 
 **Problem.** Empirical DeepSeek failure modes:
 - Tool-call JSON emitted inside `<think>`, missing from the final message.
@@ -83,7 +86,7 @@ additional round-trip rarely pays back, so it's `/harvest on` opt-in.
 4. **`storm`** — identical `(tool, args)` tuple within a sliding window →
    suppress the call, inject a reflection turn.
 
-### Pillar 4 — Cost Control *(v0.6)*
+### Pillar 3 — Cost Control *(v0.6)*
 
 **Problem.** Coding agents that default to the frontier model (v4-pro, ~12×
 flash cost) and accumulate full tool results in context are $150-$250/month
@@ -95,17 +98,13 @@ tuning in the common case:
 
 #### 4.1 Tiered defaults (flash-first)
 
-The three presets now trade only **model tier** and **reasoning effort**:
+The three presets trade **model tier** and **reasoning effort**:
 
-| Preset | Model | Effort | Harvest | Branch | Cost |
-|---|---|---|---|---|---:|
-| `fast` | `v4-flash` | `high` | off | 1 | 1× |
-| `smart` (default) | `v4-flash` | `max` | off | 1 | ~1.5× |
-| `max` | `v4-pro` | `max` | off | 1 | ~12× |
-
-**Branching and harvest are never in a preset.** Both multiply cost and
-rarely pay back; `/branch N` and `/harvest on` are pure opt-ins. A user can
-run for weeks without ever touching them.
+| Preset | Model | Effort | Cost |
+|---|---|---|---:|
+| `flash` | `v4-flash` | `max` | 1× |
+| `auto` (default) | `v4-flash` → `v4-pro` on hard turns | `max` | 1–3× |
+| `pro` | `v4-pro` | `max` | ~12× |
 
 All auxiliary calls — `forceSummaryAfterIterLimit`, subagent spawns,
 truncation repair retries — hard-code `v4-flash + effort=high` regardless
@@ -153,9 +152,8 @@ Per-turn and session cost are colored in the StatsPanel:
 ```
 src/
 ├── client.ts               # DeepSeek client (fetch + SSE)
-├── loop.ts                 # Pillar 1 + 4 — CacheFirstLoop
-├── harvest.ts              # Pillar 2 — flash-backed plan-state extraction
-├── repair/                 # Pillar 3 pipeline
+├── loop.ts                 # Pillar 1 + 3 — CacheFirstLoop
+├── repair/                 # Pillar 2 pipeline
 │   ├── index.ts
 │   ├── scavenge.ts
 │   ├── flatten.ts
@@ -172,7 +170,7 @@ src/
 │   ├── skills.ts           # list + invoke SKILL.md playbooks
 │   ├── subagent.ts         # spawn_subagent — flash+high by default
 │   ├── plan.ts             # submit_plan (review gate)
-│   └── web.ts              # web_search, web_fetch
+│   └── web.ts              # web_search, web_fetch (multi-engine: Mojeek or SearXNG)
 ├── mcp/                    # MCP client + bridge (stdio + SSE)
 ├── memory.ts               # ImmutablePrefix / AppendOnlyLog / VolatileScratch
 ├── project-memory.ts       # REASONIX.md loader
@@ -220,23 +218,24 @@ means editing one handler file and one registry line.
 
 ## Design evolution
 
-- **v0.0.x** — Pillar 1 end-to-end, Pillar 3 complete, Ink TUI scaffold.
+- **v0.0.x** — Pillar 1 end-to-end, repair pipeline complete, Ink TUI scaffold.
 - **v0.1** — τ-bench numbers published, streaming polish, transcript replay.
-- **v0.2** — Self-consistency / branch-budget sampling driven by plan state.
 - **v0.3** — MCP client (stdio + SSE), session persistence.
 - **v0.4.x** — `reasonix code` with SEARCH/REPLACE edits, review/auto
   gate, background jobs, hooks.
 - **v0.5.x** — V4 model support, skills, memory, subagents, actionable
   error messages.
-- **v0.6** *(current)* —
-  - **Pillar 4 cost control** (flash-first defaults, auto-compaction,
-    `/pro` one-shot, failure-triggered escalation, cost badges).
+- **v0.6** —
+  - **Cost control** (flash-first defaults, auto-compaction, `/pro` one-shot,
+    failure-triggered escalation, cost badges).
   - `deepseek-chat` / `deepseek-reasoner` scheduled for deprecation —
     all user-facing surfaces updated to `v4-flash` / `v4-pro`.
-  - `branch` + `harvest` removed from every preset; manual opt-in only.
   - Shared prompt fragments (`TUI_FORMATTING_RULES`, `NEGATIVE_CLAIM_RULE`).
   - UI refactor: App.tsx split into 6 hooks/components, slash.ts split
     into 13 per-topic modules.
+- **v0.31** *(current)* — `branch` + `harvest` features removed entirely
+  (the parallel-sample selector and Pillar 2 plan-state extractor); both
+  rarely paid for themselves and bloated the slash surface.
 
 ## Explicit non-goals
 

docs/index.html

@@ -403,7 +403,7 @@ npx reasonix diff \
         </div>
       </section>
 
-      <section class="cli">
+      <section id="cli" class="cli">
         <div class="container">
           <h2 class="section-title" data-i18n="cli.title">CLI at a glance</h2>
           <pre class="code"><code>npx reasonix code [path]                 <span class="hash"># <span data-i18n="cli.code">coding mode scoped to path</span></span>

package.json

@@ -1,6 +1,6 @@
 {
   "name": "reasonix",
-  "version": "0.27.1",
+  "version": "0.32.0",
   "description": "DeepSeek-native coding agent: cache-first loop, flash-first cost control, tool-call repair.",
   "type": "module",
   "bin": {
@@ -72,11 +72,12 @@
     "commander": "^12.1.0",
     "eventsource-parser": "^3.0.0",
     "ignore": "^7.0.5",
+    "ink": "^7.0.2",
+    "ink-text-input": "^6.0.0",
+    "node-html-parser": "^7.1.0",
     "picomatch": "^4.0.4",
-    "react": "^18.3.1",
-    "react-reconciler": "^0.29.2",
+    "react": "^19.2.6",
     "string-width": "^7.2.0",
-    "yoga-layout": "^3.2.1",
     "zod": "^4.4.1"
   },
   "devDependencies": {
@@ -85,14 +86,12 @@
     "@stryker-mutator/vitest-runner": "^9.6.1",
     "@types/node": "^22.9.0",
     "@types/picomatch": "^4.0.3",
-    "@types/react": "^18.3.12",
-    "@types/react-reconciler": "^0.28.9",
+    "@types/react": "^19.2.14",
     "@vitest/coverage-v8": "^2.1.5",
     "esbuild": "^0.21.5",
     "highlight.js": "^11.10.0",
     "htm": "^3.1.1",
-    "ink": "^5.1.0",
-    "ink-text-input": "^6.0.0",
+    "ink-testing-library": "^4.0.0",
     "marked": "^15.0.12",
     "preact": "^10.22.0",
     "simple-git-hooks": "^2.13.1",

src/at-mentions.ts

@@ -13,6 +13,9 @@ import {
 /** Caps match tool-result dispatch truncation (0.5.2). */
 export const DEFAULT_AT_MENTION_MAX_BYTES = 64 * 1024;
 
+/** Cap on entries returned for a `@<dir>` listing. ~200 paths × ~50 chars ≈ 10 KB — fits inside DEFAULT_AT_MENTION_MAX_BYTES with room for the rest of the prompt. */
+export const DEFAULT_AT_DIR_MAX_ENTRIES = 200;
+
 /** Universally-uninteresting build / VCS dirs. Framework-specific dirs (Pods, target, …) live in .gitignore. */
 export const DEFAULT_PICKER_IGNORE_DIRS: readonly string[] = [
   "node_modules",
@@ -91,6 +94,19 @@ export function listFilesWithStatsSync(root: string, opts: ListFilesOptions = {}
           /* stat failed (permission / EAGAIN) — keep the entry with mtime=0 */
         }
         out.push({ path: relPath, mtimeMs });
+      } else if (ent.isSymbolicLink()) {
+        // Dirent.isFile() returns false for symlinks even when they point at
+        // regular files — stat the target to recover them. Symlinks-to-dirs
+        // are not followed (cycle risk).
+        let target: ReturnType<typeof statSync> | null = null;
+        try {
+          target = statSync(absPath);
+        } catch {
+          continue;
+        }
+        if (!target.isFile()) continue;
+        if (ignoredByLayers(effectiveLayers, absPath, false)) continue;
+        out.push({ path: relPath, mtimeMs: target.mtimeMs });
       }
     }
   };
@@ -146,7 +162,9 @@ export async function listFilesWithStatsAsync(
           if (out.length >= maxResults) return;
         }
         await walk(absPath, relPath, effectiveLayers);
-      } else if (ent.isFile()) {
+      } else if (ent.isFile() || ent.isSymbolicLink()) {
+        // Symlinks land in the same batch — statBatch resolves them and drops
+        // any whose target isn't a regular file (broken or symlink-to-dir).
         fileEnts.push(ent);
       }
     }
@@ -176,15 +194,20 @@ async function statBatch(
   const stats = await Promise.all(
     accepted.map((e) =>
       stat(join(dirAbs, e.name))
-        .then((s) => s.mtimeMs)
-        .catch(() => 0),
+        .then((s) => ({ mtimeMs: s.mtimeMs, isFile: s.isFile() }))
+        .catch(() => null),
     ),
   );
   for (let i = 0; i < accepted.length; i++) {
     const ent = accepted[i]!;
+    const s = stats[i];
+    if (ent.isSymbolicLink()) {
+      // Drop broken symlinks and symlinks-to-dirs (latter would cycle).
+      if (!s || !s.isFile) continue;
+    }
     out.push({
       path: dirRel ? `${dirRel}/${ent.name}` : ent.name,
-      mtimeMs: stats[i] ?? 0,
+      mtimeMs: s?.mtimeMs ?? 0,
     });
   }
 }
@@ -283,8 +306,14 @@ export interface AtMentionExpansion {
   path: string;
   /** True if the content was inlined. False = skipped (reason in `skip`). */
   ok: boolean;
-  /** Bytes read (only for ok=true). */
+  /** Bytes read (only for ok=true and isDirectory=false). */
   bytes?: number;
+  /** True when the mention resolved to a directory (ok=true). Block uses `<directory>` instead of `<file>`. */
+  isDirectory?: boolean;
+  /** Number of files listed when isDirectory=true. */
+  entries?: number;
+  /** True iff the directory listing was clipped at maxDirEntries. */
+  truncated?: boolean;
   /** Why the mention was skipped. Set when ok=false. */
   skip?: "missing" | "not-file" | "too-large" | "escape" | "read-error";
 }
@@ -292,9 +321,19 @@ export interface AtMentionExpansion {
 export interface AtMentionOptions {
   /** Max file size in bytes before a mention is skipped. */
   maxBytes?: number;
+  /** Cap on entries returned for a `@<dir>` listing. Default {@link DEFAULT_AT_DIR_MAX_ENTRIES}. */
+  maxDirEntries?: number;
   fs?: {
     exists: (path: string) => boolean;
     isFile: (path: string) => boolean;
+    /** Optional — when omitted, directories are skipped as `not-file`. */
+    isDir?: (path: string) => boolean;
+    /** Optional — receives the directory's absolute path and the project root, returns relative paths and a truncated flag. */
+    listDir?: (
+      dirAbs: string,
+      root: string,
+      max: number,
+    ) => { files: string[]; truncated: boolean };
     size: (path: string) => number;
     read: (path: string) => string;
   };
@@ -306,22 +345,28 @@ export function expandAtMentions(
   opts: AtMentionOptions = {},
 ): { text: string; expansions: AtMentionExpansion[] } {
   const maxBytes = opts.maxBytes ?? DEFAULT_AT_MENTION_MAX_BYTES;
+  const maxDirEntries = Math.max(1, opts.maxDirEntries ?? DEFAULT_AT_DIR_MAX_ENTRIES);
   const fs = opts.fs ?? defaultFs;
   const root = resolve(rootDir);
   // De-dupe by token so `@file.ts` referenced twice inlines once.
   const seen = new Map<string, AtMentionExpansion>();
   const expansions: AtMentionExpansion[] = [];
+  const dirListings = new Map<string, string[]>();
 
   for (const match of text.matchAll(AT_MENTION_PATTERN)) {
     const rawPath = match[1] ?? "";
     // Strip trailing dot (sentence terminator): `@foo.ts.` → `@foo.ts`.
-    // Keep internal dots intact.
-    const cleaned = rawPath.replace(/\.+$/, "");
+    // Keep internal dots intact. Manual loop instead of `/\.+$/` — the
+    // regex is O(n²) on dot-heavy non-matches per CodeQL js/polynomial-redos.
+    let cleaned = rawPath;
+    while (cleaned.endsWith(".")) cleaned = cleaned.slice(0, -1);
+    // Strip a single trailing slash so `@docs/` and `@docs` resolve identically.
+    if (cleaned.endsWith("/") || cleaned.endsWith("\\")) cleaned = cleaned.slice(0, -1);
     if (!cleaned) continue;
     const token = `@${cleaned}`;
     if (seen.has(token)) continue;
 
-    const expansion = resolveMention(cleaned, root, maxBytes, fs);
+    const expansion = resolveMention(cleaned, root, maxBytes, maxDirEntries, fs, dirListings);
     seen.set(token, expansion);
     expansions.push(expansion);
   }
@@ -333,7 +378,14 @@ export function expandAtMentions(
   // both what's here and what's missing.
   const blocks: string[] = [];
   for (const ex of expansions) {
-    if (ex.ok) {
+    if (ex.ok && ex.isDirectory) {
+      const files = dirListings.get(ex.path) ?? [];
+      const truncAttr = ex.truncated ? ' truncated="true"' : "";
+      const body = files.length > 0 ? `\n${files.join("\n")}\n` : "\n";
+      blocks.push(
+        `<directory path="${ex.path}" entries="${ex.entries ?? files.length}"${truncAttr}>${body}</directory>`,
+      );
+    } else if (ex.ok) {
       const content = readSafe(root, ex.path, fs);
       blocks.push(`<file path="${ex.path}">\n${content}\n</file>`);
     } else {
@@ -348,7 +400,9 @@ function resolveMention(
   rawPath: string,
   root: string,
   maxBytes: number,
+  maxDirEntries: number,
   fs: NonNullable<AtMentionOptions["fs"]>,
+  dirListings: Map<string, string[]>,
 ): AtMentionExpansion {
   // Reject absolute paths — `@/etc/passwd` should not inline.
   if (isAbsolute(rawPath)) {
@@ -363,14 +417,28 @@ function resolveMention(
   if (!fs.exists(resolved)) {
     return { token: `@${rawPath}`, path: rawPath, ok: false, skip: "missing" };
   }
-  if (!fs.isFile(resolved)) {
-    return { token: `@${rawPath}`, path: rawPath, ok: false, skip: "not-file" };
+  if (fs.isFile(resolved)) {
+    const size = fs.size(resolved);
+    if (size > maxBytes) {
+      return { token: `@${rawPath}`, path: rawPath, ok: false, skip: "too-large", bytes: size };
+    }
+    return { token: `@${rawPath}`, path: rawPath, ok: true, bytes: size };
   }
-  const size = fs.size(resolved);
-  if (size > maxBytes) {
-    return { token: `@${rawPath}`, path: rawPath, ok: false, skip: "too-large", bytes: size };
+  // Not a file — try the directory branch. listDir is optional; without it,
+  // fall back to the legacy not-file skip so test fixtures don't break.
+  if (fs.isDir?.(resolved) && fs.listDir) {
+    const { files, truncated } = fs.listDir(resolved, root, maxDirEntries);
+    dirListings.set(rawPath, files);
+    return {
+      token: `@${rawPath}`,
+      path: rawPath,
+      ok: true,
+      isDirectory: true,
+      entries: files.length,
+      truncated,
+    };
   }
-  return { token: `@${rawPath}`, path: rawPath, ok: true, bytes: size };
+  return { token: `@${rawPath}`, path: rawPath, ok: false, skip: "not-file" };
 }
 
 function readSafe(root: string, rawPath: string, fs: NonNullable<AtMentionOptions["fs"]>): string {
@@ -391,6 +459,27 @@ const defaultFs: NonNullable<AtMentionOptions["fs"]> = {
       return false;
     }
   },
+  isDir: (p) => {
+    try {
+      return statSync(p).isDirectory();
+    } catch {
+      return false;
+    }
+  },
+  listDir: (dirAbs, root, max) => {
+    // Walk from project root and filter to entries under dirAbs so the
+    // listing inherits the parent .gitignore layers. Walking dirAbs alone
+    // would miss the project-root rules above it.
+    const dirRel = relative(root, dirAbs).split(/[\\/]/).join("/");
+    const walkCap = Math.max(max * 4, 5000);
+    const all = listFilesSync(root, { maxResults: walkCap });
+    const prefix = dirRel ? `${dirRel}/` : "";
+    const filtered = dirRel ? all.filter((f) => f === dirRel || f.startsWith(prefix)) : all;
+    return {
+      files: filtered.slice(0, max),
+      truncated: filtered.length > max,
+    };
+  },
   size: (p) => {
     try {
       return statSync(p).size;

src/cli/commands/chat.tsx

@@ -1,10 +1,17 @@
 import { render } from "ink";
 import React, { useState } from "react";
-import { loadApiKey, readConfig, searchEnabled } from "../../config.js";
+import {
+  loadApiKey,
+  readConfig,
+  searchEnabled,
+  webSearchEndpoint,
+  webSearchEngine,
+} from "../../config.js";
 import { loadDotenv } from "../../env.js";
 import type { CacheFirstLoop } from "../../loop.js";
 import { McpClient } from "../../mcp/client.js";
 import { type InspectionReport, inspectMcpServer } from "../../mcp/inspect.js";
+import { preflightStdioSpec } from "../../mcp/preflight.js";
 import { type McpClientHost, bridgeMcpTools } from "../../mcp/registry.js";
 import { parseMcpSpec } from "../../mcp/spec.js";
 import { SseTransport } from "../../mcp/sse.js";
@@ -24,6 +31,7 @@ import { registerWebTools } from "../../tools/web.js";
 import { App } from "../ui/App.js";
 import { SessionPicker } from "../ui/SessionPicker.js";
 import { Setup } from "../ui/Setup.js";
+import { drainTtyResponses } from "../ui/drain-tty.js";
 import { KeystrokeProvider } from "../ui/keystroke-context.js";
 import { formatMcpLifecycleEvent } from "../ui/mcp-lifecycle.js";
 import { formatMcpSlowToast } from "../ui/mcp-toast.js";
@@ -101,6 +109,7 @@ function createMcpRuntime(ctx: RuntimeContext): McpRuntime {
         : ctx.getRequestedCount() === 1 && ctx.getMcpPrefix()
           ? (ctx.getMcpPrefix() as string)
           : "";
+      if (spec.transport === "stdio") preflightStdioSpec(spec);
       const transport: McpTransport =
         spec.transport === "sse"
           ? new SseTransport({ url: spec.url })
@@ -254,8 +263,6 @@ export interface ChatOptions {
   model: string;
   system: string;
   transcript?: string;
-  harvest?: boolean;
-  branch?: number;
   /**
    * Soft USD cap on session spend. Undefined → no cap (default).
    * The loop warns once at 80% and refuses to start a new turn at
@@ -300,6 +307,13 @@ export interface ChatOptions {
    * URL is visible in the status bar.
    */
   noDashboard?: boolean;
+  /**
+   * Render into the terminal's alternate screen buffer. Default true —
+   * alt-screen avoids the scrollback-mode resize/wrap ghost class. Pass
+   * false (CLI: `--no-alt-screen`) when the chat output needs to remain
+   * in shell scrollback after exit.
+   */
+  altScreen?: boolean;
 }
 
 interface RootProps extends ChatOptions {
@@ -387,8 +401,6 @@ function Root({
         model={appProps.model}
         system={appProps.system}
         transcript={appProps.transcript}
-        harvest={appProps.harvest}
-        branch={appProps.branch}
         budgetUsd={appProps.budgetUsd}
         session={activeSession}
         tools={tools}
@@ -436,7 +448,9 @@ export async function chatCommand(opts: ChatOptions): Promise<void> {
   if (runtime.size() === 0 && !opts.seedTools) {
     tools = undefined;
   }
-  const mcpSpecs = runtime.specs();
+  // Preserve configured specs even when startup leaves them unbridged
+  // so `/mcp` can still surface the configured-but-offline set.
+  const mcpSpecs = [...requestedSpecs];
   const mcpServers = runtime.summaries();
 
   // Register web search/fetch tools unless explicitly disabled. DDG
@@ -444,7 +458,10 @@ export async function chatCommand(opts: ChatOptions): Promise<void> {
   // a question needs info fresher than its training data.
   if (searchEnabled()) {
     if (!tools) tools = new ToolRegistry();
-    registerWebTools(tools);
+    registerWebTools(tools, {
+      webSearchEngine: webSearchEngine(),
+      webSearchEndpoint: webSearchEndpoint(),
+    });
   }
 
   // Memory tools — available in every session, not just code mode.
@@ -475,17 +492,6 @@ export async function chatCommand(opts: ChatOptions): Promise<void> {
   const showPicker =
     !opts.session && !opts.forceResume && listSessionsForWorkspace(launchWorkspace).length > 0;
 
-  // No startup clear, no resize listener. Earlier attempts wrote
-  // various combinations of \x1b[2J / \x1b[3J / cursor-home to
-  // present a 'clean canvas' on launch and to neutralize Ink's
-  // eraseLines miscount on resize, but on xterm.js-based terminals
-  // (VSCode integrated terminal in particular) those sequences
-  // interfere with scrollback in ways that make wheel-up scroll
-  // dead. Letting Ink mount directly leaves the user's previous
-  // shell prompt visible above (scrolling up just works) and
-  // accepts the resize-ghost / launch-noise tradeoffs as known
-  // limitations — `/clear` is the manual reset.
-
   const { waitUntilExit } = render(
     <Root
       initialKey={initialKey}
@@ -498,12 +504,23 @@ export async function chatCommand(opts: ChatOptions): Promise<void> {
       {...opts}
       session={resolvedSession}
     />,
-    // patchConsole:false — winpty/MINTTY redraw-glitch source.
-    { exitOnCtrlC: true, patchConsole: false },
+    {
+      exitOnCtrlC: true,
+      // patchConsole:false — winpty/MINTTY redraw-glitch source.
+      patchConsole: false,
+      incrementalRendering: true,
+      // Default true — alt-screen is the only mode without scrollback-
+      // reflow ghosting. `--no-alt-screen` opts back into scrollback mode
+      // for users who need chat output preserved in shell history on exit.
+      alternateScreen: opts.altScreen !== false,
+    },
   );
   try {
     await waitUntilExit();
   } finally {
     await runtime.closeAll();
+    // Eat any pending terminal-feature-detection responses (#365) so the
+    // parent shell doesn't print them as junk after exit.
+    await drainTtyResponses();
   }
 }

src/cli/commands/code.tsx

@@ -45,15 +45,6 @@ export interface CodeOptions {
   forceResume?: boolean;
   /** Skip the session picker — always wipe prior messages and start fresh. */
   forceNew?: boolean;
-  /**
-   * Opt into Pillar 2 harvesting (extracts a typed plan state from R1
-   * reasoning via an extra V3 call). Default OFF in code mode — the
-   * displayed subgoals/hypotheses/rejectedPaths have no programmatic
-   * consumer, only `uncertainties.length` feeds branching, and the
-   * extra V3 call adds ~10-15% per-turn cost. Users who want the
-   * reasoning surfaced explicitly can pass `--harvest`.
-   */
-  harvest?: boolean;
   /**
    * Soft USD spend cap. Off by default. Same semantics as `chat`:
    * warns at 80%, refuses next turn at 100%. Mid-session adjustable
@@ -66,6 +57,8 @@ export interface CodeOptions {
   systemAppend?: string;
   /** Path to a UTF-8 text file whose contents are appended to the code system prompt. */
   systemAppendFile?: string;
+  /** Default true. Pass false (CLI: `--no-alt-screen`) to keep chat output in shell scrollback. */
+  altScreen?: boolean;
 }
 
 export async function codeCommand(opts: CodeOptions = {}): Promise<void> {
@@ -175,7 +168,6 @@ export async function codeCommand(opts: CodeOptions = {}): Promise<void> {
 
   await chatCommand({
     model: opts.model ?? "deepseek-v4-flash",
-    harvest: opts.harvest ?? false,
     budgetUsd: opts.budgetUsd,
     system: codeSystemPrompt(rootDir, {
       hasSemanticSearch: semantic.enabled,
@@ -190,5 +182,6 @@ export async function codeCommand(opts: CodeOptions = {}): Promise<void> {
     forceResume: opts.forceResume,
     forceNew: opts.forceNew,
     noDashboard: opts.noDashboard,
+    altScreen: opts.altScreen,
   });
 }

src/cli/commands/doctor.ts

@@ -1,10 +1,10 @@
 /** Plain-text (not Ink) — must work when everything else is broken. fail → exit 1; warn → exit 0. */
 
-import { existsSync, statSync } from "node:fs";
+import { existsSync, readFileSync, statSync } from "node:fs";
 import { homedir } from "node:os";
 import { dirname, join, resolve } from "node:path";
 import { DeepSeekClient } from "../../client.js";
-import { defaultConfigPath, readConfig } from "../../config.js";
+import { defaultConfigPath, readConfig, resolveSemanticEmbeddingConfig } from "../../config.js";
 import { loadDotenv } from "../../env.js";
 import { loadHooks } from "../../hooks.js";
 import { indexExists } from "../../index/semantic/builder.js";
@@ -259,9 +259,6 @@ async function checkHooks(projectRoot: string): Promise<Check> {
 }
 
 async function checkOllama(projectRoot: string): Promise<Check> {
-  // Ollama is only meaningful when an index has been built — otherwise
-  // the daemon being down is just noise. Skip the check entirely on
-  // projects that don't use semantic search.
   let exists = false;
   try {
     exists = await indexExists(projectRoot);
@@ -270,48 +267,82 @@ async function checkOllama(projectRoot: string): Promise<Check> {
   }
   if (!exists) {
     return {
-      label: "ollama       ",
+      label: "semantic     ",
       level: "ok",
       detail: "not in use (no semantic index built; `reasonix index` to enable)",
     };
   }
+  const meta = readSemanticMeta(projectRoot);
+  if (meta?.provider === "openai-compat") {
+    const resolved = resolveSemanticEmbeddingConfig();
+    if (resolved.provider !== "openai-compat") {
+      return {
+        label: "semantic     ",
+        level: "warn",
+        detail: `index uses openai-compat/${meta.model} but current config resolves to ${resolved.provider}/${resolved.model} — rebuild before searching`,
+      };
+    }
+    return {
+      label: "semantic     ",
+      level: "ok",
+      detail: `openai-compat · ${resolved.baseUrl} · model ${resolved.model} · api key configured`,
+    };
+  }
   try {
-    const status = await checkOllamaStatus(process.env.REASONIX_EMBED_MODEL ?? "nomic-embed-text");
+    const model = meta?.model || process.env.REASONIX_EMBED_MODEL || "nomic-embed-text";
+    const status = await checkOllamaStatus(model);
     if (!status.binaryFound) {
       return {
-        label: "ollama       ",
+        label: "semantic     ",
         level: "warn",
-        detail: "binary not on PATH — semantic_search will fail; install from https://ollama.com",
+        detail:
+          "ollama binary not on PATH — semantic_search will fail; install from https://ollama.com",
       };
     }
     if (!status.daemonRunning) {
       return {
-        label: "ollama       ",
+        label: "semantic     ",
         level: "warn",
-        detail: "daemon not running — `ollama serve` (or just call /semantic in TUI to auto-start)",
+        detail:
+          "ollama daemon not running — `ollama serve` (or call /semantic in TUI to auto-start)",
       };
     }
     if (!status.modelPulled) {
       return {
-        label: "ollama       ",
+        label: "semantic     ",
         level: "warn",
         detail: `model ${status.modelName} not pulled — \`ollama pull ${status.modelName}\``,
       };
     }
     return {
-      label: "ollama       ",
+      label: "semantic     ",
       level: "ok",
-      detail: `daemon up · model ${status.modelName} ready`,
+      detail: `ollama daemon up · model ${status.modelName} ready`,
     };
   } catch (err) {
     return {
-      label: "ollama       ",
+      label: "semantic     ",
       level: "warn",
       detail: `probe failed — ${(err as Error).message}`,
     };
   }
 }
 
+function readSemanticMeta(
+  projectRoot: string,
+): { provider: "ollama" | "openai-compat"; model: string } | null {
+  try {
+    const raw = readFileSync(join(projectRoot, ".reasonix", "semantic", "index.meta.json"), "utf8");
+    const parsed = JSON.parse(raw) as { provider?: string; model?: string };
+    return {
+      provider: parsed.provider === "openai-compat" ? "openai-compat" : "ollama",
+      model: typeof parsed.model === "string" ? parsed.model : "",
+    };
+  } catch {
+    return null;
+  }
+}
+
 async function checkProject(projectRoot: string): Promise<Check> {
   // Heuristic: a "real" project has either .git, REASONIX.md, or
   // package.json. Lacking all three, `reasonix code` still works but

src/cli/commands/events.ts

@@ -80,6 +80,14 @@ function detailsFor(e: Event): string {
       return `${e.callId} reason=${e.reason}`;
     case "tool.result":
       return `${e.callId} ${e.ok ? "ok" : "err"} ${e.output.length}B${e.truncated ? " [trunc]" : ""}`;
+    case "tool.call":
+      return `${e.name} args=${truncate(JSON.stringify(e.args), 60)}`;
+    case "tool.confirm.allow":
+      return `${e.kind} ${quote(e.payload.command, 60)}`;
+    case "tool.confirm.deny":
+      return `${e.kind} ${quote(e.payload.command, 60)}${e.denyContext ? ` — ${truncate(e.denyContext, 40)}` : ""}`;
+    case "tool.confirm.always_allow":
+      return `${e.kind} ${quote(e.payload.command, 60)} prefix=${truncate(e.prefix, 30)}`;
     case "effect.file.touched":
       return `${e.mode} ${e.path} (${e.bytes}B)`;
     case "effect.memory.written":

src/cli/commands/index.ts

@@ -1,34 +1,37 @@
 /** `reasonix index` — progress writes go to stderr so stdout stays pipeable. */
 
 import { resolve } from "node:path";
-import { loadIndexConfig } from "../../config.js";
+import { loadIndexConfig, resolveSemanticEmbeddingConfig } from "../../config.js";
 import { buildIndex } from "../../index/semantic/builder.js";
 import type { BuildProgress, BuildResult, SkipBuckets } from "../../index/semantic/builder.js";
 import { t } from "../../index/semantic/i18n.js";
-import { ollamaPreflight } from "../../index/semantic/preflight.js";
+import { semanticPreflight } from "../../index/semantic/preflight.js";
 
 export interface IndexCommandOptions {
   rebuild?: boolean;
   model?: string;
   dir?: string;
-  /** Ollama base URL override. */
   ollamaUrl?: string;
-  /** Skip preflight prompts (yes to all). For scripts that already
-   *  know Ollama is set up. Default false. */
   yes?: boolean;
 }
 
 export async function indexCommand(opts: IndexCommandOptions = {}): Promise<void> {
   const root = resolve(opts.dir ?? process.cwd());
   const tty = process.stderr.isTTY === true && process.stdin.isTTY === true;
-  const model = opts.model ?? process.env.REASONIX_EMBED_MODEL ?? "nomic-embed-text";
+  const resolved = resolveSemanticEmbeddingConfig();
+  const embedding =
+    resolved.provider === "ollama"
+      ? {
+          ...resolved,
+          model: opts.model ?? resolved.model,
+          baseUrl: opts.ollamaUrl ?? resolved.baseUrl,
+        }
+      : {
+          ...resolved,
+          model: opts.model ?? resolved.model,
+        };
 
-  // Preflight: detect Ollama state and offer to fix what's missing.
-  // Runs before chunking so the user doesn't watch a scan finish only
-  // to hit "daemon not reachable" on the first embed call.
-  const preflightOk = await ollamaPreflight({
-    model,
-    baseUrl: opts.ollamaUrl,
+  const preflightOk = await semanticPreflight(embedding, {
     interactive: tty && !opts.yes,
     yesToAll: opts.yes ?? false,
   });
@@ -40,9 +43,8 @@ export async function indexCommand(opts: IndexCommandOptions = {}): Promise<void
   let result: BuildResult;
   try {
     result = await buildIndex(root, {
+      ...embedding,
       rebuild: opts.rebuild,
-      model,
-      baseUrl: opts.ollamaUrl,
       indexConfig: loadIndexConfig(),
       onProgress: (p) => writer.update(p),
     });
@@ -135,7 +137,6 @@ function makeNonTtyWriter(): ProgressWriter {
   };
 }
 
-/** Spinner ticks on its own setInterval so a hung embedder still shows liveness. */
 function makeTtyWriter(): ProgressWriter {
   let status = t("progressStarting");
   let lastLineLen = 0;
@@ -152,8 +153,6 @@ function makeTtyWriter(): ProgressWriter {
     lastLineLen = line.length;
   };
 
-  // Paint the initial frame immediately so the user sees text on
-  // line one — don't wait for the first interval tick.
   repaint();
   const interval = setInterval(repaint, SPINNER_INTERVAL_MS);
 
@@ -167,9 +166,6 @@ function makeTtyWriter(): ProgressWriter {
         const pct = total > 0 ? ((done / total) * 100).toFixed(0) : "0";
         status = t("progressEmbed", { done, total, pct });
       }
-      // Repaint immediately on event arrival so the new counter
-      // shows up before the next interval tick (avoids a 120ms lag
-      // on each progress update).
       repaint();
     },
     clear() {

src/cli/commands/mcp-browse.tsx

@@ -1,7 +1,6 @@
 /** `reasonix mcp browse` — Ink TUI for the MCP marketplace. Lazy-loads pages on scroll. */
 
 import { Box, Text, render, useApp, useInput } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useCallback, useEffect, useMemo, useState } from "react";
 import { readConfig, writeConfig } from "../../config.js";
 import { loadDotenv } from "../../env.js";

src/cli/commands/mcp-inspect.ts

@@ -1,6 +1,7 @@
 import { McpClient } from "../../mcp/client.js";
 import { inspectMcpServer } from "../../mcp/inspect.js";
 import type { InspectionReport } from "../../mcp/inspect.js";
+import { preflightStdioSpec } from "../../mcp/preflight.js";
 import { parseMcpSpec } from "../../mcp/spec.js";
 import { SseTransport } from "../../mcp/sse.js";
 import { type McpTransport, StdioTransport } from "../../mcp/stdio.js";
@@ -15,6 +16,7 @@ export interface McpInspectOptions {
 
 export async function mcpInspectCommand(opts: McpInspectOptions): Promise<void> {
   const spec = parseMcpSpec(opts.spec);
+  if (spec.transport === "stdio") preflightStdioSpec(spec);
   const transport: McpTransport =
     spec.transport === "sse"
       ? new SseTransport({ url: spec.url })
@@ -35,6 +37,32 @@ export async function mcpInspectCommand(opts: McpInspectOptions): Promise<void>
   }
 }
 
+export function formatMcpInspectFailure(err: unknown): string {
+  const error = err instanceof Error ? err : new Error(String(err));
+  const message = error.message;
+  const code = (error as NodeJS.ErrnoException).code;
+
+  if (code === "ENOENT") {
+    const command = message.match(/^spawn\s+([^\s]+)\s+ENOENT$/)?.[1] ?? "the command";
+    return `${message} — try: install or verify \`${command}\`, then check the MCP spec's command spelling`;
+  }
+
+  if (code === "ECONNREFUSED") {
+    const target = message.match(/\b(https?:\/\/\S+|\d+\.\d+\.\d+\.\d+:\d+|localhost:\d+)\b/i)?.[1];
+    return `${message} — try: confirm ${target ?? "the MCP server"} is running and the host/port match the spec`;
+  }
+
+  if (/^MCP request initialize \(id=\d+\) timed out after \d+ms$/.test(message)) {
+    return `${message} — try: confirm the target speaks MCP and completes the handshake before the request timeout`;
+  }
+
+  if (/^(empty MCP spec|MCP spec ".*" has name but no command)/.test(message)) {
+    return `${message} — try: pass \`name=command args\` or an http(s):// URL`;
+  }
+
+  return message;
+}
+
 function formatReport(nsName: string, r: InspectionReport): string {
   const lines: string[] = [];
   lines.push(`MCP server [${nsName}]`);

src/cli/commands/replay.ts

@@ -67,11 +67,6 @@ function printReplay(opts: ReplayOptions): void {
   } else if (stats.prefixHashes.length > 1) {
     console.log("  (prefix churned — cache-hostile session)");
   }
-  if (stats.harvestedTurns > 0) {
-    console.log(`harvest (Pillar 2): ${stats.harvestedTurns} turn(s) produced plan state`);
-    console.log(`  subgoals:          ${stats.totalSubgoals}`);
-    console.log(`  uncertainties:     ${stats.totalUncertainties}`);
-  }
 }
 
 function sliceRecords(records: TranscriptRecord[], opts: ReplayOptions): TranscriptRecord[] {
@@ -98,21 +93,6 @@ function renderRecord(rec: TranscriptRecord): void {
           })()
         : "";
     console.log(`${turn} AGENT:${cost}${cache} ${oneLine(rec.content)}`);
-    if (rec.planState) {
-      const ps = rec.planState;
-      if (ps.subgoals.length)
-        console.log(`   ‹ subgoals     (${ps.subgoals.length}): ${ps.subgoals.join(" · ")}`);
-      if (ps.hypotheses.length)
-        console.log(`   ‹ hypotheses   (${ps.hypotheses.length}): ${ps.hypotheses.join(" · ")}`);
-      if (ps.uncertainties.length)
-        console.log(
-          `   ‹ uncertainties(${ps.uncertainties.length}): ${ps.uncertainties.join(" · ")}`,
-        );
-      if (ps.rejectedPaths.length)
-        console.log(
-          `   ‹ rejected     (${ps.rejectedPaths.length}): ${ps.rejectedPaths.join(" · ")}`,
-        );
-    }
   } else if (rec.role === "tool") {
     const args = rec.args ? ` args=${oneLine(rec.args, 80)}` : "";
     console.log(`${turn} TOOL ${rec.tool ?? "?"}:${args} → ${oneLine(rec.content, 120)}`);

src/cli/commands/run.ts

@@ -11,6 +11,7 @@ import {
 import { loadDotenv } from "../../env.js";
 import { CacheFirstLoop, DeepSeekClient, ImmutablePrefix } from "../../index.js";
 import { McpClient } from "../../mcp/client.js";
+import { preflightStdioSpec } from "../../mcp/preflight.js";
 import { bridgeMcpTools } from "../../mcp/registry.js";
 import { parseMcpSpec } from "../../mcp/spec.js";
 import { SseTransport } from "../../mcp/sse.js";
@@ -26,10 +27,6 @@ export interface RunOptions {
   task: string;
   model: string;
   system: string;
-  /** Opt into Pillar 2 plan-state harvesting. Adds a cheap V3 call per turn. */
-  harvest?: boolean;
-  /** Self-consistency budget. > 1 runs N parallel samples and picks the best. */
-  branch?: number;
   budgetUsd?: number;
   /** JSONL transcript path — lets `reasonix replay` / `diff` audit this run. */
   transcript?: string;
@@ -103,6 +100,7 @@ export async function runCommand(opts: RunOptions): Promise<void> {
           : requestedSpecs.length === 1 && opts.mcpPrefix
             ? opts.mcpPrefix
             : "";
+        if (spec.transport === "stdio") preflightStdioSpec(spec);
         const transport: McpTransport =
           spec.transport === "sse"
             ? new SseTransport({ url: spec.url })
@@ -154,8 +152,6 @@ export async function runCommand(opts: RunOptions): Promise<void> {
     prefix,
     tools,
     model: opts.model,
-    harvest: opts.harvest,
-    branch: opts.branch,
     budgetUsd: opts.budgetUsd,
   });
   const prefixHash = prefix.fingerprint;

src/cli/index.ts

@@ -35,7 +35,7 @@ import { doctorCommand } from "./commands/doctor.js";
 import { eventsCommand } from "./commands/events.js";
 import { indexCommand } from "./commands/index.js";
 import { mcpBrowseCommand } from "./commands/mcp-browse.js";
-import { mcpInspectCommand } from "./commands/mcp-inspect.js";
+import { formatMcpInspectFailure, mcpInspectCommand } from "./commands/mcp-inspect.js";
 import { mcpInstallCommand, mcpListCommand, mcpSearchCommand } from "./commands/mcp.js";
 import { pruneSessionsCommand } from "./commands/prune-sessions.js";
 import { replayCommand } from "./commands/replay.js";
@@ -87,8 +87,6 @@ program.action(async (opts: { continue?: boolean }) => {
   await chatCommand({
     model: defaults.model,
     system: applyMemoryStack(DEFAULT_SYSTEM, process.cwd()),
-    harvest: defaults.harvest,
-    branch: defaults.branch,
     session: continueOpts.session,
     mcp: defaults.mcp,
     forceResume: continueOpts.forceResume,
@@ -110,9 +108,9 @@ program
   .option("-r, --resume", t("ui.resumeHint"))
   .option("-n, --new", t("ui.newHint"))
   .option("--transcript <path>", t("ui.transcriptHint"))
-  .option("--harvest", t("ui.harvestHint"))
   .option("--budget <usd>", t("ui.budgetHint"), (v) => Number.parseFloat(v))
   .option("--no-dashboard", t("ui.noDashboard"))
+  .option("--no-alt-screen", "keep chat output in shell scrollback (legacy mode, ghost-prone)")
   .option("--system-append <prompt>", t("ui.systemAppendHint"))
   .option("--system-append-file <path>", t("ui.systemAppendFileHint"))
   .action(async (dir: string | undefined, opts) => {
@@ -123,11 +121,11 @@ program
       transcript: opts.transcript,
       forceResume: !!opts.resume,
       forceNew: !!opts.new,
-      harvest: !!opts.harvest,
       budgetUsd: parseBudgetFlag(opts.budget),
       noDashboard: opts.dashboard === false,
       systemAppend: opts.systemAppend,
       systemAppendFile: opts.systemAppendFile,
+      altScreen: opts.altScreen !== false,
     });
   });
 
@@ -138,8 +136,6 @@ program
   .option("-s, --system <prompt>", t("ui.systemPromptHint"), DEFAULT_SYSTEM)
   .option("--transcript <path>", t("ui.transcriptHint"))
   .option("--preset <name>", t("ui.presetHint"))
-  .option("--harvest", t("ui.harvestOptInHint"))
-  .option("--branch <n>", t("ui.branchHint"), (v) => Number.parseInt(v, 10))
   .option("--budget <usd>", t("ui.budgetHint"), (v) => Number.parseFloat(v))
   .option("--session <name>", t("ui.sessionNameHint"))
   .option("--no-session", t("ui.ephemeralHint"))
@@ -155,11 +151,10 @@ program
   .option("--mcp-prefix <str>", t("ui.mcpPrefixHint"))
   .option("--no-config", t("ui.noConfigHint"))
   .option("--no-dashboard", t("ui.noDashboard"))
+  .option("--no-alt-screen", "keep chat output in shell scrollback (legacy mode, ghost-prone)")
   .action(async (opts) => {
     const defaults = resolveDefaults({
       model: opts.model,
-      harvest: opts.harvest,
-      branch: opts.branch,
       mcp: opts.mcp as string[],
       session: opts.session,
       preset: opts.preset,
@@ -181,8 +176,6 @@ program
       model: defaults.model,
       system: applyMemoryStack(opts.system, process.cwd()),
       transcript: opts.transcript,
-      harvest: defaults.harvest,
-      branch: defaults.branch,
       budgetUsd: parseBudgetFlag(opts.budget),
       session: continueOpts.session,
       mcp: defaults.mcp,
@@ -190,6 +183,7 @@ program
       forceResume: continueOpts.forceResume,
       forceNew: !!opts.new,
       noDashboard: opts.dashboard === false,
+      altScreen: opts.altScreen !== false,
     });
   });
 
@@ -199,8 +193,6 @@ program
   .option("-m, --model <id>", t("ui.modelIdHint"))
   .option("-s, --system <prompt>", t("ui.systemPromptHint"), DEFAULT_SYSTEM)
   .option("--preset <name>", t("ui.presetHintShort"))
-  .option("--harvest", t("ui.harvestHintShort"))
-  .option("--branch <n>", t("ui.branchHintShort"), (v) => Number.parseInt(v, 10))
   .option("--budget <usd>", t("ui.budgetHintShort"), (v) => Number.parseFloat(v))
   .option("--transcript <path>", t("ui.transcriptHintShort"))
   .option(
@@ -214,8 +206,6 @@ program
   .action(async (task: string, opts) => {
     const defaults = resolveDefaults({
       model: opts.model,
-      harvest: opts.harvest,
-      branch: opts.branch,
       mcp: opts.mcp as string[],
       preset: opts.preset,
       noConfig: opts.config === false,
@@ -224,8 +214,6 @@ program
       task,
       model: defaults.model,
       system: applyMemoryStack(opts.system, process.cwd()),
-      harvest: defaults.harvest,
-      branch: defaults.branch,
       budgetUsd: parseBudgetFlag(opts.budget),
       transcript: opts.transcript,
       mcp: defaults.mcp,
@@ -414,7 +402,7 @@ mcp
     try {
       await mcpInspectCommand({ spec, json: !!opts.json });
     } catch (err) {
-      process.stderr.write(`mcp inspect failed: ${(err as Error).message}\n`);
+      process.stderr.write(`mcp inspect failed: ${formatMcpInspectFailure(err)}\n`);
       process.exit(1);
     }
   });

src/cli/resolve.ts

@@ -6,17 +6,12 @@ import { resolvePreset } from "./ui/presets.js";
 export interface ResolvedDefaults {
   model: string;
   reasoningEffort: "high" | "max";
-  harvest: boolean;
-  branch: number | undefined;
   mcp: string[];
   session: string | undefined;
 }
 
 export interface RawCliFlags {
   model?: string;
-  harvest?: boolean;
-  /** From `commander`; already parseInt'd. */
-  branch?: number;
   mcp?: string[];
   /** Commander's `--no-session` surfaces as `false`; `--session X` as a string. */
   session?: string | false;
@@ -33,9 +28,6 @@ export function resolveDefaults(flags: RawCliFlags): ResolvedDefaults {
 
   const model = flags.model ?? presetSettings.model;
   const reasoningEffort = presetSettings.reasoningEffort;
-  const harvest = flags.harvest === true ? true : presetSettings.harvest;
-  const branchFromFlag = normalizeBranch(flags.branch);
-  const branch = branchFromFlag ?? (presetSettings.branch > 1 ? presetSettings.branch : undefined);
 
   // `--mcp` accumulator is [] when absent. Treat empty from flags as
   // "user didn't pass" → fall through to config. Users who explicitly
@@ -44,7 +36,7 @@ export function resolveDefaults(flags: RawCliFlags): ResolvedDefaults {
 
   const session = resolveSession(flags.session, cfg.session);
 
-  return { model, reasoningEffort, harvest, branch, mcp, session };
+  return { model, reasoningEffort, mcp, session };
 }
 
 function pickPreset(
@@ -69,12 +61,6 @@ function isPresetName(s: string): s is PresetName {
   );
 }
 
-function normalizeBranch(raw: number | undefined): number | undefined {
-  if (raw === undefined) return undefined;
-  if (!Number.isFinite(raw) || raw <= 1) return undefined;
-  return Math.min(raw, 8);
-}
-
 function resolveSession(
   flag: string | false | undefined,
   configSession: string | null | undefined,

src/cli/ui/CheckpointPicker.tsx

@@ -0,0 +1,156 @@
+import { Box, Text, useStdout } from "ink";
+// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
+import React, { useMemo, useState } from "react";
+import type { CheckpointMeta } from "../../code/checkpoints.js";
+import { fmtAgo } from "../../code/checkpoints.js";
+import { type PickerBroadcastPorts, usePickerBroadcast } from "./dashboard/use-picker-broadcast.js";
+import { useKeystroke } from "./keystroke-context.js";
+import { FG, TONE } from "./theme/tokens.js";
+
+export type CheckpointPickerOutcome =
+  | { kind: "restore"; id: string }
+  | { kind: "delete"; id: string }
+  | { kind: "quit" };
+
+export interface CheckpointPickerProps {
+  checkpoints: ReadonlyArray<CheckpointMeta>;
+  workspace: string;
+  onChoose: (outcome: CheckpointPickerOutcome) => void;
+  pickerPorts?: PickerBroadcastPorts;
+}
+
+const PAGE_MARGIN = 6;
+
+export function CheckpointPicker({
+  checkpoints,
+  workspace,
+  onChoose,
+  pickerPorts,
+}: CheckpointPickerProps): React.ReactElement {
+  const [focus, setFocus] = useState(0);
+  const { stdout } = useStdout();
+  const rows = stdout?.rows ?? 40;
+  const visibleCount = Math.max(3, rows - PAGE_MARGIN);
+
+  const snapshot = useMemo(
+    () => ({
+      pickerKind: "checkpoints",
+      title: `restore a checkpoint — ${workspace}`,
+      items: checkpoints.map((c) => {
+        const sizeKb = (c.bytes / 1024).toFixed(1);
+        const tag = c.source === "manual" ? "" : ` (${c.source})`;
+        return {
+          id: c.id,
+          title: `${c.name}${tag}`,
+          subtitle: `${c.fileCount} file${c.fileCount === 1 ? "" : "s"} · ${sizeKb} KB`,
+          badge: c.id.slice(0, 7),
+          meta: fmtAgo(c.createdAt),
+        };
+      }),
+      actions: ["pick", "delete", "cancel"] as const,
+      hint: "↑↓ pick · ⏎ restore · [d] forget · esc quit",
+    }),
+    [checkpoints, workspace],
+  );
+
+  usePickerBroadcast(
+    !!pickerPorts,
+    {
+      ...snapshot,
+      actions: [...snapshot.actions],
+    },
+    (res) => {
+      if (res.action === "pick") return onChoose({ kind: "restore", id: res.id });
+      if (res.action === "delete") return onChoose({ kind: "delete", id: res.id });
+      if (res.action === "cancel") return onChoose({ kind: "quit" });
+    },
+    pickerPorts ?? {
+      broadcast: () => undefined,
+      resolverRef: { current: null },
+      snapshotRef: { current: null },
+    },
+  );
+
+  useKeystroke((ev) => {
+    if (ev.escape) return onChoose({ kind: "quit" });
+    if (ev.upArrow) return setFocus((f) => Math.max(0, f - 1));
+    if (ev.downArrow) return setFocus((f) => Math.min(checkpoints.length - 1, f + 1));
+    if (checkpoints.length === 0) {
+      if (ev.return) return onChoose({ kind: "quit" });
+      return;
+    }
+    const target = checkpoints[focus];
+    if (!target) return;
+    if (ev.return) return onChoose({ kind: "restore", id: target.id });
+    if (ev.input === "q") return onChoose({ kind: "quit" });
+    if (ev.input === "d") return onChoose({ kind: "delete", id: target.id });
+  });
+
+  const start = Math.max(
+    0,
+    Math.min(focus - Math.floor(visibleCount / 2), checkpoints.length - visibleCount),
+  );
+  const end = Math.min(checkpoints.length, start + visibleCount);
+  const shown = checkpoints.slice(start, end);
+  const hiddenBelow = checkpoints.length - end;
+
+  return (
+    <Box flexDirection="column" marginY={1}>
+      <Box>
+        <Text bold color={TONE.brand}>
+          {" ◈ REASONIX · pick a checkpoint "}
+        </Text>
+        <Text color={FG.meta}>{`  ·  ${workspace}`}</Text>
+      </Box>
+      <Box height={1} />
+      {checkpoints.length === 0 ? (
+        <Box>
+          <Text color={FG.faint}>
+            {"  no checkpoints in this workspace yet — see /checkpoint to make one"}
+          </Text>
+        </Box>
+      ) : (
+        shown.map((c, i) => <CheckpointRow key={c.id} info={c} focused={start + i === focus} />)
+      )}
+      {hiddenBelow > 0 ? (
+        <Box>
+          <Text color={FG.faint}>{`     … ${hiddenBelow} more`}</Text>
+        </Box>
+      ) : null}
+      <Box marginTop={1}>
+        <Text color={FG.faint}>
+          {checkpoints.length === 0
+            ? "  esc quit"
+            : "  ↑↓ pick  ·  ⏎ restore  ·  [d] forget  ·  esc quit"}
+        </Text>
+      </Box>
+    </Box>
+  );
+}
+
+function CheckpointRow({
+  info,
+  focused,
+}: {
+  info: CheckpointMeta;
+  focused: boolean;
+}): React.ReactElement {
+  const tag = info.source === "manual" ? "" : ` (${info.source})`;
+  const sizeKb = (info.bytes / 1024).toFixed(1);
+  const time = fmtAgo(info.createdAt);
+  return (
+    <Box>
+      <Text color={focused ? TONE.brand : FG.faint}>{focused ? "  ▸ " : "    "}</Text>
+      <Text color={FG.meta}>{info.id.slice(0, 7).padEnd(8)}</Text>
+      <Text bold={focused} color={focused ? FG.strong : FG.sub}>
+        {info.name}
+      </Text>
+      <Text color={FG.faint}>{tag}</Text>
+      <Box flexGrow={1} />
+      <Text color={FG.faint}>{`${time.padStart(8)}  ·  `}</Text>
+      <Text
+        color={FG.faint}
+      >{`${info.fileCount} file${info.fileCount === 1 ? "" : "s"}, ${sizeKb} KB`}</Text>
+    </Box>
+  );
+}

src/cli/ui/ChoiceConfirm.tsx

@@ -1,6 +1,5 @@
 /** Modal picker for `ask_choice` — options + optional "type my own" escape hatch. */
 
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
 import type { ChoiceOption } from "../../tools/choice.js";
 import { SingleSelect } from "./Select.js";

src/cli/ui/ChromeBar.tsx

@@ -1,259 +0,0 @@
-import { Box, Text, useStdout } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React from "react";
-import stringWidth from "string-width";
-import type { SessionSummary } from "../../telemetry/stats.js";
-import { ChromeRule } from "./primitives.js";
-import { COLOR, GRADIENT } from "./theme.js";
-
-const COLD_START_TURNS = 3;
-const CACHE_BAR_CELLS = 10;
-
-export type Preset = "auto" | "flash" | "pro";
-
-export interface ChromeBarProps {
-  summary: SessionSummary;
-  rootDir?: string;
-  sessionName?: string | null;
-  preset?: Preset;
-  planMode?: boolean;
-  proArmed?: boolean;
-  escalated?: boolean;
-  updateAvailable?: string | null;
-  balance?: { currency: string; total: number } | null;
-  budgetUsd?: number | null;
-  /** Scroll progress: 0 = at latest, 1 = at oldest. Hides the pill at 0. */
-  scrollRatio?: number;
-}
-
-export function ChromeBar(props: ChromeBarProps): React.ReactElement {
-  const { stdout } = useStdout();
-  const cols = (stdout?.columns ?? 80) - 2; // subtract paddingX={1} on both sides
-  const cold = props.summary.turns <= COLD_START_TURNS;
-  const projectName = props.rootDir ? basename(props.rootDir) : null;
-  const mode = pickModePill(props.planMode, props.preset);
-  const proPill = props.escalated
-    ? { label: "⇧ pro", color: COLOR.err }
-    : props.proArmed
-      ? { label: "⇧ pro", color: COLOR.warn }
-      : null;
-
-  // Greedy width-aware fit. Layout (each gap is exactly ONE space — single
-  // suffix on update/mode/pro/scroll, single prefix on balance/cache):
-  //   [brand][·project][›session]<spacer>[scroll][update][mode][pro]$cost  w bal cache █▓▓▓
-  // Always-shown carve out the budget first; optional pieces are dropped
-  // by priority — balance > cache > session > update — until the row fits.
-  // (`flexGrow` spacer can shrink to 0, so no minimum reserve.) Cache is
-  // shown unconditionally now: pre-turn-4 it renders dim with a "—" so
-  // it's "default on" instead of materializing only after a few turns.
-  const SEP = 3; // " · " / " › "
-  const GAP = 1;
-  const scrollLabel =
-    props.scrollRatio !== undefined && props.scrollRatio > 0
-      ? `[↑ ${Math.round(props.scrollRatio * 100)}%]`
-      : "";
-  const updateLabel = props.updateAvailable ? `↑ ${props.updateAvailable}` : "";
-  const balanceLabel = props.balance ? formatBalanceLabel(props.balance) : "";
-  const cachePct = Math.round(props.summary.cacheHitRatio * 100);
-  // Use the worst-case rendered cache string (3-digit pct, full bar) so the
-  // shed decision doesn't oscillate as the percentage changes turn-to-turn.
-  const cacheLabel = `cache ${"█".repeat(CACHE_BAR_CELLS)} 100%`;
-  const costLabel = `$${props.summary.totalCostUsd.toFixed(4)}${
-    props.budgetUsd && props.budgetUsd > 0 ? ` / $${props.budgetUsd.toFixed(2)}` : ""
-  }`;
-
-  const brandW = stringWidth("◈ reasonix");
-  const projectW = projectName ? SEP + stringWidth(projectName) : 0;
-  const fixedLeft = brandW + projectW;
-  const scrollW = scrollLabel ? stringWidth(scrollLabel) + GAP : 0;
-  const modeW = mode ? stringWidth(`[${mode.label}]`) + GAP : 0;
-  const proW = proPill ? stringWidth(`[${proPill.label}]`) + GAP : 0;
-  const costW = stringWidth(costLabel);
-  const fixedRight = scrollW + modeW + proW + costW;
-  let budget = cols - fixedLeft - fixedRight;
-
-  const balanceOptW = balanceLabel ? GAP + stringWidth(balanceLabel) : 0;
-  const cacheOptW = GAP + stringWidth(cacheLabel);
-  const sessionOptW = props.sessionName ? SEP + stringWidth(props.sessionName) : 0;
-  const updateOptW = updateLabel ? stringWidth(updateLabel) + GAP : 0;
-
-  const showBalance = balanceOptW > 0 && budget >= balanceOptW;
-  if (showBalance) budget -= balanceOptW;
-  const showCache = budget >= cacheOptW;
-  if (showCache) budget -= cacheOptW;
-  const showSession = sessionOptW > 0 && budget >= sessionOptW;
-  if (showSession) budget -= sessionOptW;
-  const showUpdate = updateOptW > 0 && budget >= updateOptW;
-  if (showUpdate) budget -= updateOptW;
-
-  return (
-    <Box flexDirection="column" paddingX={1}>
-      <Box>
-        <Text bold color={GRADIENT[0]}>
-          {"◈ "}
-        </Text>
-        <Text color={COLOR.brand} bold>
-          reasonix
-        </Text>
-        {projectName ? (
-          <>
-            <Text color={COLOR.info} dimColor>
-              {" · "}
-            </Text>
-            <Text>{projectName}</Text>
-            {showSession && props.sessionName ? (
-              <>
-                <Text color={COLOR.info} dimColor>
-                  {" › "}
-                </Text>
-                <Text color={COLOR.info}>{props.sessionName}</Text>
-              </>
-            ) : null}
-          </>
-        ) : null}
-
-        <Box flexGrow={1} />
-
-        {scrollLabel ? (
-          <>
-            <Text color={COLOR.accent} bold>
-              {scrollLabel}
-            </Text>
-            <Text> </Text>
-          </>
-        ) : null}
-        {showUpdate ? (
-          <>
-            <Text color={COLOR.warn} bold>
-              {updateLabel}
-            </Text>
-            <Text> </Text>
-          </>
-        ) : null}
-        {mode ? (
-          <>
-            <Text color={mode.color} bold>
-              {`[${mode.label}]`}
-            </Text>
-            <Text> </Text>
-          </>
-        ) : null}
-        {proPill ? (
-          <>
-            <Text color={proPill.color} bold>
-              {`[${proPill.label}]`}
-            </Text>
-            <Text> </Text>
-          </>
-        ) : null}
-        <CostPill summary={props.summary} cold={cold} budgetUsd={props.budgetUsd} />
-        {showBalance && props.balance ? (
-          <>
-            <Text> </Text>
-            <BalancePill balance={props.balance} />
-          </>
-        ) : null}
-        {showCache ? (
-          <>
-            <Text> </Text>
-            <CachePill ratio={props.summary.cacheHitRatio} cold={cold} pct={cachePct} />
-          </>
-        ) : null}
-      </Box>
-      <ChromeRule />
-    </Box>
-  );
-}
-
-function formatBalanceLabel(balance: { currency: string; total: number }): string {
-  const sym = balance.currency === "USD" ? "$" : balance.currency === "CNY" ? "¥" : "";
-  const suf = sym ? "" : ` ${balance.currency}`;
-  return `w ${sym}${balance.total.toFixed(2)}${suf}`;
-}
-
-function CostPill({
-  summary,
-  cold,
-  budgetUsd,
-}: {
-  summary: SessionSummary;
-  cold: boolean;
-  budgetUsd?: number | null;
-}): React.ReactElement {
-  const cost = summary.totalCostUsd;
-  const color = cold ? COLOR.info : sessionCostColor(cost);
-  return (
-    <>
-      <Text color={color} bold={!cold} dimColor={cold}>
-        {`$${cost.toFixed(4)}`}
-      </Text>
-      {budgetUsd && budgetUsd > 0 ? (
-        <Text color={COLOR.info} dimColor>
-          {` / $${budgetUsd.toFixed(2)}`}
-        </Text>
-      ) : null}
-    </>
-  );
-}
-
-function BalancePill({
-  balance,
-}: {
-  balance: { currency: string; total: number };
-}): React.ReactElement {
-  const { total } = balance;
-  const color = total < 1 ? COLOR.err : total < 5 ? COLOR.warn : COLOR.ok;
-  return <Text color={color}>{formatBalanceLabel(balance)}</Text>;
-}
-
-function CachePill({
-  ratio,
-  cold,
-  pct,
-}: {
-  ratio: number;
-  cold: boolean;
-  pct: number;
-}): React.ReactElement {
-  const color = ratio >= 0.7 ? COLOR.ok : ratio >= 0.4 ? COLOR.warn : COLOR.err;
-  const filled = Math.max(0, Math.min(CACHE_BAR_CELLS, Math.round(ratio * CACHE_BAR_CELLS)));
-  const empty = CACHE_BAR_CELLS - filled;
-  return (
-    <>
-      <Text color={COLOR.info} dimColor>
-        {"cache "}
-      </Text>
-      <Text color={cold ? COLOR.info : color} dimColor={cold}>
-        {"█".repeat(filled)}
-      </Text>
-      <Text dimColor>{"░".repeat(empty)}</Text>
-      <Text color={cold ? undefined : color} bold={!cold} dimColor={cold}>
-        {cold && pct === 0 ? " —" : ` ${pct}%`}
-      </Text>
-    </>
-  );
-}
-
-function pickModePill(
-  planMode: boolean | undefined,
-  preset: Preset | undefined,
-): { label: string; color: string } | null {
-  if (planMode) return { label: "PLAN", color: COLOR.err };
-  if (!preset) return null;
-  if (preset === "pro") return { label: "pro", color: COLOR.accent };
-  if (preset === "flash") return { label: "flash", color: COLOR.info };
-  return { label: "auto", color: COLOR.primary };
-}
-
-function sessionCostColor(cost: number): string | undefined {
-  if (cost <= 0) return undefined;
-  if (cost >= 5) return COLOR.err;
-  if (cost >= 0.5) return COLOR.warn;
-  return COLOR.ok;
-}
-
-function basename(path: string): string {
-  const norm = path.replace(/\\/g, "/");
-  const i = norm.lastIndexOf("/");
-  return i === -1 ? norm : norm.slice(i + 1);
-}

src/cli/ui/McpBrowser.tsx

@@ -1,7 +1,6 @@
 /** `/mcp` browser modal — keyboard-driven server list per design §24. */
 
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useState } from "react";
 import { useKeystroke } from "./keystroke-context.js";
 import { toggleMcpDisabled } from "./mcp-disable.js";

src/cli/ui/McpHub.tsx

@@ -1,10 +1,10 @@
 /** `/mcp` slash modal — single hub with two tabs: Live (attached servers) + Marketplace (registry). */
 
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useState } from "react";
 import { McpBrowser } from "./McpBrowser.js";
 import { McpMarketplace } from "./McpMarketplace.js";
+import type { PickerBroadcastPorts } from "./dashboard/use-picker-broadcast.js";
 import { useKeystroke } from "./keystroke-context.js";
 import type { ApplyAppend } from "./mcp-reconnect-kickoff.js";
 import type { McpServerSummary } from "./slash/types.js";
@@ -24,6 +24,8 @@ export interface McpHubProps {
     removed: string[];
     failed: Array<{ spec: string; reason: string }>;
   }>;
+  /** Forwarded to the marketplace tab so the web dashboard can drive install / uninstall / refine / load-more. */
+  pickerPorts?: PickerBroadcastPorts;
 }
 
 export function McpHub({
@@ -34,6 +36,7 @@ export function McpHub({
   postInfo,
   applyAppend,
   reloadMcp,
+  pickerPorts,
 }: McpHubProps) {
   const [tab, setTab] = useState<McpHubTab>(initialTab);
 
@@ -65,7 +68,12 @@ export function McpHub({
           applyAppend={applyAppend}
         />
       ) : (
-        <McpMarketplace onClose={onClose} postInfo={postInfo} reloadMcp={reloadMcp} />
+        <McpMarketplace
+          onClose={onClose}
+          postInfo={postInfo}
+          reloadMcp={reloadMcp}
+          pickerPorts={pickerPorts}
+        />
       )}
     </Box>
   );

src/cli/ui/McpMarketplace.tsx

@@ -1,7 +1,6 @@
 /** `/mcp browse` modal — registry marketplace inside the chat session. */
 
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useCallback, useEffect, useMemo, useState } from "react";
 import { readConfig, writeConfig } from "../../config.js";
 import {
@@ -12,6 +11,7 @@ import {
   specStringFor,
 } from "../../mcp/registry-fetch.js";
 import type { RegistryEntry } from "../../mcp/registry-types.js";
+import { type PickerBroadcastPorts, usePickerBroadcast } from "./dashboard/use-picker-broadcast.js";
 import { useKeystroke } from "./keystroke-context.js";
 import { COLOR } from "./theme.js";
 
@@ -27,6 +27,7 @@ export interface McpMarketplaceProps {
     removed: string[];
     failed: Array<{ spec: string; reason: string }>;
   }>;
+  pickerPorts?: PickerBroadcastPorts;
 }
 
 interface State {
@@ -39,6 +40,39 @@ interface State {
   installedSpecs: string[];
 }
 
+export function buildMarketplacePickerSnapshot(args: {
+  filtered: RegistryEntry[];
+  installedSpecs: string[];
+  query: string;
+  status: string;
+  hasMore: boolean;
+}) {
+  return {
+    pickerKind: "mcp-marketplace" as const,
+    title: `MCP marketplace · ${args.status}`,
+    query: args.query,
+    items: args.filtered.map((e) => {
+      const installedSpec = isInstalled(args.installedSpecs, e);
+      return {
+        id: e.name,
+        title: e.title || e.name,
+        subtitle: e.description?.slice(0, 200) ?? undefined,
+        badge: installedSpec
+          ? "installed"
+          : e.source === "official"
+            ? "official"
+            : e.source === "smithery"
+              ? "smithery"
+              : "local",
+        meta: e.popularity !== undefined ? `★ ${e.popularity.toLocaleString()}` : undefined,
+      };
+    }),
+    actions: ["install", "uninstall", "refine", "load-more", "cancel"] as const,
+    hasMore: args.hasMore,
+    hint: "type filter · ↑↓ pick · ⏎ install/toggle · PgDn load more · esc close",
+  };
+}
+
 function rankAndFilter(entries: RegistryEntry[], query: string): RegistryEntry[] {
   const q = query.trim().toLowerCase();
   const list = q
@@ -66,7 +100,7 @@ function isInstalled(installedSpecs: string[], entry: RegistryEntry): string | n
   }
 }
 
-export function McpMarketplace({ onClose, postInfo, reloadMcp }: McpMarketplaceProps) {
+export function McpMarketplace({ onClose, postInfo, reloadMcp, pickerPorts }: McpMarketplaceProps) {
   const [state, setState] = useState<State>({
     handle: null,
     loading: true,
@@ -128,33 +162,30 @@ export function McpMarketplace({ onClose, postInfo, reloadMcp }: McpMarketplaceP
     }
   }, [state.handle, state.loading]);
 
-  const installOrToggle = useCallback(
-    async (entry: RegistryEntry) => {
-      const installed = isInstalled(state.installedSpecs, entry);
-      if (installed) {
-        const cfg = readConfig();
-        const next = (cfg.mcp ?? []).filter((s) => s !== installed);
-        writeConfig({ ...cfg, mcp: next });
-        setState((s) => ({
-          ...s,
-          installedSpecs: next,
-          status: `uninstalled ${entry.name}`,
-        }));
-        if (reloadMcp) {
-          try {
-            await reloadMcp();
-            postInfo(`✓ uninstalled ${entry.name} — bridge dropped`);
-          } catch (err) {
-            postInfo(
-              `✓ uninstalled ${entry.name} — restart \`reasonix code\` to drop the bridge (reload failed: ${(err as Error).message})`,
-            );
-          }
-        } else {
-          postInfo(`✓ uninstalled ${entry.name} — restart \`reasonix code\` to drop the bridge`);
+  const doUninstall = useCallback(
+    async (entry: RegistryEntry, installed: string) => {
+      const cfg = readConfig();
+      const next = (cfg.mcp ?? []).filter((s) => s !== installed);
+      writeConfig({ ...cfg, mcp: next });
+      setState((s) => ({ ...s, installedSpecs: next, status: `uninstalled ${entry.name}` }));
+      if (reloadMcp) {
+        try {
+          await reloadMcp();
+          postInfo(`✓ uninstalled ${entry.name} — bridge dropped`);
+        } catch (err) {
+          postInfo(
+            `✓ uninstalled ${entry.name} — restart \`reasonix code\` to drop the bridge (reload failed: ${(err as Error).message})`,
+          );
         }
-        return;
+      } else {
+        postInfo(`✓ uninstalled ${entry.name} — restart \`reasonix code\` to drop the bridge`);
       }
+    },
+    [postInfo, reloadMcp],
+  );
 
+  const doInstall = useCallback(
+    async (entry: RegistryEntry) => {
       let install = entry.install;
       if (!install && entry.source === "smithery") {
         setState((s) => ({ ...s, loading: true, status: "fetching smithery detail…" }));
@@ -176,7 +207,6 @@ export function McpMarketplace({ onClose, postInfo, reloadMcp }: McpMarketplaceP
         }));
         return;
       }
-
       try {
         const spec = specStringFor(entry.name, install);
         const cfg = readConfig();
@@ -216,7 +246,69 @@ export function McpMarketplace({ onClose, postInfo, reloadMcp }: McpMarketplaceP
         setState((s) => ({ ...s, status: `install failed: ${(err as Error).message}` }));
       }
     },
-    [state.installedSpecs, postInfo, reloadMcp],
+    [postInfo, reloadMcp],
+  );
+
+  const installOrToggle = useCallback(
+    async (entry: RegistryEntry) => {
+      const installed = isInstalled(state.installedSpecs, entry);
+      if (installed) await doUninstall(entry, installed);
+      else await doInstall(entry);
+    },
+    [state.installedSpecs, doInstall, doUninstall],
+  );
+
+  const pickerSnapshot = useMemo(
+    () =>
+      buildMarketplacePickerSnapshot({
+        filtered,
+        installedSpecs: state.installedSpecs,
+        query: state.query,
+        status: state.status,
+        hasMore: state.handle?.cache.pagination.nextCursor != null,
+      }),
+    [filtered, state.installedSpecs, state.handle, state.query, state.status],
+  );
+
+  usePickerBroadcast(
+    !!pickerPorts,
+    { ...pickerSnapshot, actions: [...pickerSnapshot.actions] },
+    (res) => {
+      if (res.action === "cancel") return onClose();
+      if (res.action === "refine") {
+        setState((s) => ({ ...s, query: res.query, selected: 0 }));
+        return;
+      }
+      if (res.action === "load-more") {
+        void fetchMore();
+        return;
+      }
+      if (res.action === "install") {
+        const entry = state.handle?.cache.entries.find((e) => e.name === res.id);
+        if (!entry) return;
+        if (isInstalled(state.installedSpecs, entry)) {
+          setState((s) => ({ ...s, status: `already installed: ${entry.name}` }));
+          return;
+        }
+        void doInstall(entry);
+        return;
+      }
+      if (res.action === "uninstall") {
+        const entry = state.handle?.cache.entries.find((e) => e.name === res.id);
+        if (!entry) return;
+        const installed = isInstalled(state.installedSpecs, entry);
+        if (!installed) {
+          setState((s) => ({ ...s, status: `not installed: ${entry.name}` }));
+          return;
+        }
+        void doUninstall(entry, installed);
+      }
+    },
+    pickerPorts ?? {
+      broadcast: () => undefined,
+      resolverRef: { current: null },
+      snapshotRef: { current: null },
+    },
   );
 
   useKeystroke((ev) => {

src/cli/ui/ModelPicker.tsx

@@ -0,0 +1,219 @@
+import { Box, Text, useStdout } from "ink";
+// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
+import React, { useState } from "react";
+import { useKeystroke } from "./keystroke-context.js";
+import { PRESETS, PRESET_DESCRIPTIONS } from "./presets.js";
+import { PILL_MODEL, Pill, modelBadgeFor } from "./primitives/Pill.js";
+import { FG, TONE } from "./theme/tokens.js";
+
+export type ModelPickerOutcome =
+  | { kind: "select"; id: string }
+  | { kind: "preset"; name: "auto" | "flash" | "pro" }
+  | { kind: "quit" };
+
+export interface ModelPickerProps {
+  /** API-fetched ids; null means "still loading / offline". */
+  models: ReadonlyArray<string> | null;
+  /** Model id currently active in the loop — marked with the cursor on open. */
+  current: string;
+  /** Used to detect which preset (if any) the loop currently matches. */
+  currentEffort: "high" | "max";
+  currentAutoEscalate: boolean;
+  onChoose: (outcome: ModelPickerOutcome) => void;
+  /** Triggers a refetch when the catalog is null/empty and the user presses [r]. */
+  onRefresh?: () => void;
+}
+
+const PAGE_MARGIN = 8;
+const PRESET_NAMES = ["auto", "flash", "pro"] as const;
+type PresetName = (typeof PRESET_NAMES)[number];
+
+type Row = { kind: "preset"; name: PresetName } | { kind: "model"; id: string };
+
+export function ModelPicker({
+  models,
+  current,
+  currentEffort,
+  currentAutoEscalate,
+  onChoose,
+  onRefresh,
+}: ModelPickerProps): React.ReactElement {
+  const modelList = (models && models.length > 0 ? models : FALLBACK_MODELS).slice();
+  if (!modelList.includes(current)) modelList.unshift(current);
+  const presetRows: Row[] = PRESET_NAMES.map((name) => ({ kind: "preset", name }));
+  const modelRows: Row[] = modelList.map((id) => ({ kind: "model", id }));
+  const rows: Row[] = [...presetRows, ...modelRows];
+
+  const activePreset = detectActivePreset(current, currentEffort, currentAutoEscalate);
+  const initialIndex = activePreset
+    ? PRESET_NAMES.indexOf(activePreset)
+    : presetRows.length + Math.max(0, modelList.indexOf(current));
+  const [focus, setFocus] = useState(initialIndex);
+  const { stdout } = useStdout();
+  const termRows = stdout?.rows ?? 40;
+  const visibleCount = Math.max(6, termRows - PAGE_MARGIN);
+
+  useKeystroke((ev) => {
+    if (ev.escape) return onChoose({ kind: "quit" });
+    if (ev.upArrow) return setFocus((f) => Math.max(0, f - 1));
+    if (ev.downArrow) return setFocus((f) => Math.min(rows.length - 1, f + 1));
+    if (ev.return) {
+      const target = rows[focus];
+      if (!target) return;
+      if (target.kind === "preset") return onChoose({ kind: "preset", name: target.name });
+      return onChoose({ kind: "select", id: target.id });
+    }
+    if (!ev.input) return;
+    if (ev.input === "q") return onChoose({ kind: "quit" });
+    if (ev.input === "r") onRefresh?.();
+  });
+
+  const start = Math.max(
+    0,
+    Math.min(focus - Math.floor(visibleCount / 2), rows.length - visibleCount),
+  );
+  const end = Math.min(rows.length, start + visibleCount);
+  const shown = rows.slice(start, end);
+  const hiddenAbove = start;
+  const hiddenBelow = rows.length - end;
+  const loading = models === null;
+  const empty = models !== null && models.length === 0;
+
+  let lastSection: "preset" | "model" | null = null;
+
+  return (
+    <Box flexDirection="column" marginY={1}>
+      <Box>
+        <Text bold color={TONE.brand}>
+          {" ◈ REASONIX · pick a setup "}
+        </Text>
+        <Text color={FG.meta}>
+          {loading
+            ? "  ·  loading catalog…"
+            : empty
+              ? "  ·  catalog empty — using known fallbacks"
+              : `  ·  ${modelList.length} models available`}
+        </Text>
+      </Box>
+      <Box height={1} />
+      {hiddenAbove > 0 ? (
+        <Box>
+          <Text color={FG.faint}>{`     … ${hiddenAbove} earlier`}</Text>
+        </Box>
+      ) : null}
+      {shown.map((row, i) => {
+        const idx = start + i;
+        const focused = idx === focus;
+        const showHeader = row.kind !== lastSection;
+        lastSection = row.kind;
+        const header = showHeader ? (
+          <Box key={`hdr-${row.kind}`} marginTop={idx === 0 ? 0 : 1}>
+            <Text color={FG.meta}>
+              {row.kind === "preset"
+                ? "    PRESETS  ·  recommended — model + effort + auto-escalate"
+                : "    MODELS  ·  raw pick — auto-escalate stays as-is"}
+            </Text>
+          </Box>
+        ) : null;
+        const body =
+          row.kind === "preset" ? (
+            <PresetRow
+              key={`p-${row.name}`}
+              name={row.name}
+              focused={focused}
+              active={activePreset === row.name}
+            />
+          ) : (
+            <ModelRow
+              key={`m-${row.id}`}
+              id={row.id}
+              focused={focused}
+              active={!activePreset && row.id === current}
+            />
+          );
+        return (
+          <React.Fragment key={`row-${idx}`}>
+            {header}
+            {body}
+          </React.Fragment>
+        );
+      })}
+      {hiddenBelow > 0 ? (
+        <Box>
+          <Text color={FG.faint}>{`     … ${hiddenBelow} more`}</Text>
+        </Box>
+      ) : null}
+      <Box marginTop={1}>
+        <Text color={FG.faint}>{"  ↑↓ pick  ·  ⏎ confirm  ·  [r] refresh  ·  esc cancel"}</Text>
+      </Box>
+    </Box>
+  );
+}
+
+function PresetRow({
+  name,
+  focused,
+  active,
+}: {
+  name: PresetName;
+  focused: boolean;
+  active: boolean;
+}): React.ReactElement {
+  const desc = PRESET_DESCRIPTIONS[name];
+  return (
+    <Box>
+      <Text color={focused ? TONE.brand : FG.faint}>{focused ? "  ▸ " : "    "}</Text>
+      <Text bold={focused} color={focused ? FG.strong : FG.sub}>
+        {name.padEnd(8)}
+      </Text>
+      <Text color={focused ? FG.body : FG.meta}>{desc.headline.padEnd(28)}</Text>
+      <Text color={FG.meta}>{`  ${desc.cost}`}</Text>
+      {active ? <Text color={TONE.brand}>{"  · current"}</Text> : null}
+    </Box>
+  );
+}
+
+function ModelRow({
+  id,
+  focused,
+  active,
+}: {
+  id: string;
+  focused: boolean;
+  active: boolean;
+}): React.ReactElement {
+  const badge = modelBadgeFor(id);
+  return (
+    <Box>
+      <Text color={focused ? TONE.brand : FG.faint}>{focused ? "  ▸ " : "    "}</Text>
+      <Text bold={focused} color={focused ? FG.strong : FG.sub}>
+        {id.padEnd(24)}
+      </Text>
+      <Text> </Text>
+      <Pill label={badge.label} {...PILL_MODEL[badge.kind]} bold={false} />
+      {active ? <Text color={TONE.brand}>{"  · current"}</Text> : null}
+    </Box>
+  );
+}
+
+function detectActivePreset(
+  model: string,
+  effort: "high" | "max",
+  autoEscalate: boolean,
+): PresetName | null {
+  for (const name of PRESET_NAMES) {
+    const p = PRESETS[name];
+    if (p.model === model && p.reasoningEffort === effort && p.autoEscalate === autoEscalate) {
+      return name;
+    }
+  }
+  return null;
+}
+
+/** Hard-coded known DeepSeek ids — used when the API catalog hasn't loaded yet so the picker isn't empty on first open. */
+const FALLBACK_MODELS: ReadonlyArray<string> = [
+  "deepseek-v4-flash",
+  "deepseek-v4-pro",
+  "deepseek-chat",
+  "deepseek-reasoner",
+];

src/cli/ui/PlanCheckpointConfirm.tsx

@@ -1,7 +1,6 @@
 /** Modal picker for `PlanCheckpointError`: continue / revise / stop. */
 
 import { Box } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
 import type { PlanStep } from "../../tools/plan.js";
 import { PlanStepList, type StepStatus } from "./PlanStepList.js";

src/cli/ui/PlanConfirm.tsx

@@ -1,14 +1,14 @@
 /** Modal-style picker for `submit_plan`: accept / refine / cancel. */
 
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
 import type { PlanStep } from "../../tools/plan.js";
 import { PlanStepList } from "./PlanStepList.js";
 import { SingleSelect } from "./Select.js";
 import { ApprovalCard } from "./cards/ApprovalCard.js";
 import { useReserveRows } from "./layout/viewport-budget.js";
-import { CARD, TONE } from "./theme/tokens.js";
+import { MarkdownView } from "./markdown-view.js";
+import { CARD, FG, TONE } from "./theme/tokens.js";
 
 export type PlanConfirmChoice = "approve" | "refine" | "revise" | "cancel";
 
@@ -21,9 +21,19 @@ export interface PlanConfirmProps {
   projectRoot?: string;
 }
 
+const PLAN_BODY_PREVIEW_LINES = 24;
+
 function PlanConfirmInner({ plan, steps, onChoose }: PlanConfirmProps) {
   const stepRows = steps?.length ?? 0;
-  useReserveRows("modal", { min: 10, max: Math.max(16, stepRows + 14) });
+  const hasSteps = stepRows > 0;
+  const planLines = plan.split("\n");
+  const truncatedBody = planLines.length > PLAN_BODY_PREVIEW_LINES;
+  const previewBody = truncatedBody ? planLines.slice(0, PLAN_BODY_PREVIEW_LINES).join("\n") : plan;
+  const previewRows = truncatedBody
+    ? PLAN_BODY_PREVIEW_LINES
+    : Math.min(planLines.length, PLAN_BODY_PREVIEW_LINES);
+  const reservedFor = hasSteps ? stepRows : previewRows;
+  useReserveRows("modal", { min: 10, max: Math.max(16, reservedFor + 14) });
 
   const hasOpenQuestions =
     /^#{1,6}\s*(open[-\s]?questions?|risks?|unknowns?|assumptions?|unclear)/im.test(plan) ||
@@ -45,9 +55,20 @@ function PlanConfirmInner({ plan, steps, onChoose }: PlanConfirmProps) {
           </Text>
         </Box>
       ) : null}
-      {steps && steps.length > 0 ? (
+      {hasSteps ? (
         <Box marginBottom={1} flexDirection="column">
-          <PlanStepList steps={steps} />
+          <PlanStepList steps={steps!} />
+        </Box>
+      ) : plan.trim().length > 0 ? (
+        <Box marginBottom={1} flexDirection="column">
+          <MarkdownView text={previewBody} />
+          {truncatedBody ? (
+            <Text color={FG.faint}>
+              {`… ${planLines.length - PLAN_BODY_PREVIEW_LINES} more line${
+                planLines.length - PLAN_BODY_PREVIEW_LINES === 1 ? "" : "s"
+              } above in scrollback`}
+            </Text>
+          ) : null}
         </Box>
       ) : null}
       <SingleSelect

src/cli/ui/PlanStateBlock.tsx

@@ -1,52 +0,0 @@
-/**
- * Shared Ink block that renders a TypedPlanState. Used by RecordView in
- * replay/diff TUIs.
- *
- * Colors are semantic (not decorative):
- *   - subgoals:       cyan         — structure / plan
- *   - hypotheses:     green        — current beliefs (like assistant)
- *   - uncertainties:  yellow       — attention required (like tool)
- *   - rejected paths: red dim      — ruled out (muted, like error-but-resolved)
- *
- * Visual: colored bold label + count, then dim items. No solid-bg pill
- * — every assistant turn surfaces this block, so the bracket-text style
- * keeps it from shouting four colored bg blocks at the user every turn.
- * Color alone (the bold label) is enough to identify the field type.
- */
-
-import { Box, Text } from "ink";
-import React from "react";
-import type { TypedPlanState } from "../../harvest.js";
-import { COLOR } from "./theme.js";
-
-type FieldColor = string;
-
-export function PlanStateBlock({ planState }: { planState: TypedPlanState }) {
-  const fields: Array<[string, string[], FieldColor, boolean]> = [];
-  if (planState.subgoals.length)
-    fields.push(["subgoals", planState.subgoals, COLOR.primary, false]);
-  if (planState.hypotheses.length)
-    fields.push(["hypotheses", planState.hypotheses, COLOR.assistant, false]);
-  if (planState.uncertainties.length)
-    fields.push(["uncertainties", planState.uncertainties, COLOR.warn, false]);
-  if (planState.rejectedPaths.length)
-    fields.push(["rejected", planState.rejectedPaths, COLOR.info, true]);
-  if (fields.length === 0) return null;
-
-  return (
-    <Box flexDirection="column" marginBottom={1}>
-      {fields.map(([label, items, color, dim]) => (
-        <Box key={label}>
-          <Text color={color} bold dimColor={dim}>
-            {label}
-          </Text>
-          <Text dimColor>{` (${items.length})`}</Text>
-          <Text dimColor>{"  · "}</Text>
-          <Text dimColor={dim} color={dim ? undefined : COLOR.info}>
-            {items.join(" · ")}
-          </Text>
-        </Box>
-      ))}
-    </Box>
-  );
-}

src/cli/ui/PromptInput.tsx

@@ -1,7 +1,6 @@
 import { Box, Text, useStdout } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React as a runtime value (classic transform compiles JSX to React.createElement)
 import React, { useRef, useState } from "react";
-import { useCursor } from "../../renderer/index.js";
 import { useKeystroke } from "./keystroke-context.js";
 import { useReserveRows } from "./layout/viewport-budget.js";
 import { type MultilineKey, lineAndColumn, processMultilineKey } from "./multiline-keys.js";
@@ -20,20 +19,25 @@ import { FG, SURFACE, TONE } from "./theme/tokens.js";
 
 /** Raw-stdin keystroke bus → multiline reducer; one logical line per Box row, viewport-clipped. */
 
+/** Pastes shorter than this AND single-line render verbatim; longer ones become a `[paste #N · …]` sentinel chip (#397). */
+export const INLINE_PASTE_THRESHOLD = 200;
+
+export function shouldInlinePaste(content: string): boolean {
+  return !content.includes("\n") && content.length <= INLINE_PASTE_THRESHOLD;
+}
+
 export interface PromptInputProps {
   value: string;
   onChange: (v: string) => void;
   onSubmit: (v: string) => void;
   disabled?: boolean;
   placeholder?: string;
-  /**
-   * History recall. Fired when ↑/↓ hits a buffer boundary (empty
-   * buffer, or cursor already at the first/last line of a multi-line
-   * buffer). Parent walks its prompt history and swaps the value via
-   * `onChange`. Absent → keys are consumed silently.
-   */
+  /** Ctrl+P / Ctrl+N — parent walks its prompt history and swaps `value` via `onChange`. */
   onHistoryPrev?: () => void;
   onHistoryNext?: () => void;
+  /** ↑/↓ on an empty buffer — parent scrolls chat history. */
+  onChatScrollUp?: () => void;
+  onChatScrollDown?: () => void;
 }
 
 export function PromptInput({
@@ -44,10 +48,16 @@ export function PromptInput({
   placeholder,
   onHistoryPrev,
   onHistoryNext,
+  onChatScrollUp,
+  onChatScrollDown,
 }: PromptInputProps) {
   // Cap at 24 — collapseLinesForDisplay hides content past ~20 logical lines.
+  // Quantize spec.max to 4-row buckets so per-keystroke line-count changes
+  // don't churn viewport-budget; without this every single character that
+  // adds/removes a newline re-dispatches the allocator and reflows layout.
   const inputLineCount = value.length > 0 ? value.split("\n").length : 1;
-  useReserveRows("input", { min: 1, max: Math.min(inputLineCount + 3, 24) });
+  const reserveMax = Math.min(Math.ceil(inputLineCount / 4) * 4 + 3, 24);
+  useReserveRows("input", { min: 1, max: reserveMax });
 
   const [cursor, setCursor] = useState(value.length);
 
@@ -71,15 +81,19 @@ export function PromptInput({
   const registerPaste = (content: string) => {
     const v = lastLocalValueRef.current;
     const c = cursorRef.current;
-    const id = nextPasteIdRef.current % PASTE_SENTINEL_RANGE;
-    nextPasteIdRef.current = id + 1;
-    pastesRef.current.set(id, makePasteEntry(id, content));
-    const sentinel = encodePasteSentinel(id);
-    const next = v.slice(0, c) + sentinel + v.slice(c);
+    const insertion = shouldInlinePaste(content)
+      ? content
+      : (() => {
+          const id = nextPasteIdRef.current % PASTE_SENTINEL_RANGE;
+          nextPasteIdRef.current = id + 1;
+          pastesRef.current.set(id, makePasteEntry(id, content));
+          return encodePasteSentinel(id);
+        })();
+    const next = v.slice(0, c) + insertion + v.slice(c);
     lastLocalValueRef.current = next;
-    cursorRef.current = c + 1;
+    cursorRef.current = c + insertion.length;
     onChange(next);
-    setCursor(c + 1);
+    setCursor(c + insertion.length);
   };
 
   useKeystroke((ev) => {
@@ -132,6 +146,8 @@ export function PromptInput({
     }
     if (action.historyHandoff === "prev") onHistoryPrev?.();
     if (action.historyHandoff === "next") onHistoryNext?.();
+    if (action.chatScrollHandoff === "up") onChatScrollUp?.();
+    if (action.chatScrollHandoff === "down") onChatScrollDown?.();
   }, !disabled);
 
   // ── Render ──────────────────────────────────────────────────────
@@ -153,8 +169,6 @@ export function PromptInput({
   const accentColor = disabled ? FG.faint : TONE.brand;
   const cursorVisible = true;
   const { line: cursorLine, col: cursorCol } = lineAndColumn(value, cursor);
-  // PromptInput paints its own cursor (`▌` / inverse char) — hide the real terminal cursor so it doesn't double up at the bottom of the screen.
-  useCursor({ col: 0, rowFromBottom: 0, visible: false });
 
   const renderItems = collapseLinesForDisplay(lines, cursorLine);
   const showHugeBufferHints = lines.length > 20;

src/cli/ui/RecordView.tsx

@@ -1,17 +1,8 @@
-/**
- * Shared renderer for a single TranscriptRecord. Used by ReplayApp and
- * DiffApp — both need the same visual grammar (user cyan, assistant green,
- * tool yellow, error red, cache badge colored by threshold) so transcripts
- * look consistent wherever they're displayed.
- *
- * Kept small on purpose: no streaming/branch/planState paths (those are
- * live-chat concerns and never appear in replayed transcripts).
- */
+/** Shared renderer for a single TranscriptRecord — used by ReplayApp and DiffApp. */
 
 import { Box, Text } from "ink";
 import React from "react";
 import type { TranscriptRecord } from "../../transcript/log.js";
-import { PlanStateBlock } from "./PlanStateBlock.js";
 
 export interface RecordViewProps {
   rec: TranscriptRecord;
@@ -58,7 +49,6 @@ export function RecordView({ rec, compact = false }: RecordViewProps) {
           ) : null}
           {rec.usage ? <CacheBadge usage={rec.usage} /> : null}
         </Box>
-        {rec.planState ? <PlanStateBlock planState={rec.planState} /> : null}
         {rec.content ? (
           <Text>{rec.content}</Text>
         ) : (

src/cli/ui/SessionPicker.tsx

@@ -1,9 +1,10 @@
 import { Box, Text, useStdout } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React, { useState } from "react";
+import React, { useMemo, useState } from "react";
 import type { SessionInfo } from "../../memory/session.js";
+import { type PickerBroadcastPorts, usePickerBroadcast } from "./dashboard/use-picker-broadcast.js";
 import { useKeystroke } from "./keystroke-context.js";
-import { FG, TONE, USD_TO_CNY } from "./theme/tokens.js";
+import { FG, TONE, formatCost } from "./theme/tokens.js";
 
 export type SessionPickerOutcome =
   | { kind: "open"; name: string }
@@ -16,6 +17,10 @@ export interface SessionPickerProps {
   sessions: ReadonlyArray<SessionInfo>;
   workspace: string;
   onChoose: (outcome: SessionPickerOutcome) => void;
+  /** Live wallet currency from App.tsx; falls back to each session's stored `meta.balanceCurrency` per row. */
+  walletCurrency?: string;
+  /** When provided, broadcasts to the web dashboard so it can resolve via `/api/modal/resolve`. */
+  pickerPorts?: PickerBroadcastPorts;
 }
 
 const PAGE_MARGIN = 6;
@@ -24,6 +29,8 @@ export function SessionPicker({
   sessions,
   workspace,
   onChoose,
+  walletCurrency,
+  pickerPorts,
 }: SessionPickerProps): React.ReactElement {
   const [focus, setFocus] = useState(0);
   const [renaming, setRenaming] = useState<{ from: string; buf: string } | null>(null);
@@ -31,6 +38,53 @@ export function SessionPicker({
   const rows = stdout?.rows ?? 40;
   const visibleCount = Math.max(3, rows - PAGE_MARGIN);
 
+  const snapshot = useMemo(
+    () => ({
+      pickerKind: "sessions",
+      title: `pick a session — ${workspace}`,
+      items: sessions.map((s) => {
+        const branch = s.meta.branch ?? "main";
+        const summary =
+          s.meta.summary ?? `${s.messageCount} message${s.messageCount === 1 ? "" : "s"}`;
+        const turns = s.meta.turnCount ?? Math.ceil(s.messageCount / 2);
+        const currency = walletCurrency ?? s.meta.balanceCurrency;
+        const costLabel =
+          s.meta.totalCostUsd !== undefined ? formatCost(s.meta.totalCostUsd, currency, 2) : "";
+        return {
+          id: s.name,
+          title: s.name,
+          subtitle: summary,
+          badge: branch,
+          meta: costLabel ? `${turns} turns · ${costLabel}` : `${turns} turns`,
+        };
+      }),
+      actions: ["pick", "delete", "rename", "new", "cancel"] as const,
+      hint: "↑↓ pick · ⏎ open · [n] new · [d] delete · [r] rename · esc quit",
+    }),
+    [sessions, workspace, walletCurrency],
+  );
+
+  usePickerBroadcast(
+    !!pickerPorts,
+    {
+      ...snapshot,
+      actions: [...snapshot.actions],
+    },
+    (res) => {
+      if (res.action === "pick") return onChoose({ kind: "open", name: res.id });
+      if (res.action === "delete") return onChoose({ kind: "delete", name: res.id });
+      if (res.action === "rename")
+        return onChoose({ kind: "rename", name: res.id, newName: res.text });
+      if (res.action === "new") return onChoose({ kind: "new" });
+      if (res.action === "cancel") return onChoose({ kind: "quit" });
+    },
+    pickerPorts ?? {
+      broadcast: () => undefined,
+      resolverRef: { current: null },
+      snapshotRef: { current: null },
+    },
+  );
+
   useKeystroke((ev) => {
     if (ev.paste) {
       if (renaming) setRenaming({ ...renaming, buf: renaming.buf + ev.input });
@@ -101,7 +155,14 @@ export function SessionPicker({
           <Text color={FG.faint}>{" to start a new one"}</Text>
         </Box>
       ) : (
-        shown.map((s, i) => <SessionRow key={s.name} info={s} focused={start + i === focus} />)
+        shown.map((s, i) => (
+          <SessionRow
+            key={s.name}
+            info={s}
+            focused={start + i === focus}
+            walletCurrency={walletCurrency}
+          />
+        ))
       )}
       {hiddenBelow > 0 ? (
         <Box>
@@ -135,15 +196,19 @@ export function SessionPicker({
 function SessionRow({
   info,
   focused,
-}: { info: SessionInfo; focused: boolean }): React.ReactElement {
+  walletCurrency,
+}: {
+  info: SessionInfo;
+  focused: boolean;
+  walletCurrency?: string;
+}): React.ReactElement {
   const branch = info.meta.branch ?? "main";
   const summary =
     info.meta.summary ?? `${info.messageCount} message${info.messageCount === 1 ? "" : "s"}`;
   const turns = info.meta.turnCount ?? Math.ceil(info.messageCount / 2);
-  const costCny =
-    info.meta.totalCostUsd !== undefined
-      ? `¥${(info.meta.totalCostUsd * USD_TO_CNY).toFixed(2)}`
-      : "";
+  const currency = walletCurrency ?? info.meta.balanceCurrency;
+  const costLabel =
+    info.meta.totalCostUsd !== undefined ? formatCost(info.meta.totalCostUsd, currency, 2) : "";
   const time = relativeTime(info.mtime);
   return (
     <Box>
@@ -156,7 +221,7 @@ function SessionRow({
       <Box flexGrow={1} />
       <Text color={FG.faint}>{`${time.padStart(11)}   `}</Text>
       <Text color={FG.faint}>{`${turns} turns`}</Text>
-      {costCny ? <Text color={FG.faint}>{` · ${costCny}`}</Text> : null}
+      {costLabel ? <Text color={FG.faint}>{` · ${costLabel}`}</Text> : null}
     </Box>
   );
 }

src/cli/ui/ShellConfirm.tsx

@@ -1,5 +1,4 @@
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useState } from "react";
 import { DenyContextInput } from "./DenyContextInput.js";
 import { SingleSelect } from "./Select.js";

src/cli/ui/SlashSuggestions.tsx

@@ -2,40 +2,34 @@ import { Box, Text } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig.jsx = "react" needs React in value scope for JSX compilation
 import React from "react";
 import { t } from "../../i18n/index.js";
-import type { SlashCommandSpec } from "./slash.js";
+import type { SlashCommandSpec, SlashGroup } from "./slash.js";
 import { COLOR, GLYPH } from "./theme.js";
 
 export interface SlashSuggestionsProps {
-  /**
-   * Current matching suggestions, computed by the parent. `null` means
-   * "not in slash-prefix mode" — render nothing. Empty array means "in
-   * slash mode but no matches" — render the "no matches" hint.
-   */
   matches: SlashCommandSpec[] | null;
-  /** Index (within `matches`) of the currently highlighted row. */
   selectedIndex: number;
+  /** True when input is a bare `/` — render section headers + advanced footer. */
+  groupMode?: boolean;
+  /** Count of hidden `advanced` commands; rendered as a footer hint when groupMode is true. */
+  advancedHidden?: number;
 }
 
-/**
- * Slash-command palette. Rendered below the input box while the user
- * is typing a `/…` prefix. Visual grammar matches the design doc's
- * picker style:
- *
- *      / suggestions                                   esc cancel
- *      ▸ /checkpoint   [name]    snapshot the workspace
- *        /restore      <name>    roll back files to a checkpoint
- *        /diff                   cumulative diff vs session start
- *      [↑↓] navigate · [Tab/⏎] pick
- *
- * Cmd tokens render in accent violet (one consistent color for "this
- * is a command"), descriptions in dim info color. The selected row
- * gets a leading ▸ + bold cmd; non-selected rows get a 2-space pad in
- * the same column so the cmd column stays aligned. No solid-bg pill —
- * the bracket-text pattern from the chrome carries through here.
- */
+const GROUP_LABEL: Record<SlashGroup, string> = {
+  chat: "CHAT",
+  setup: "SETUP",
+  info: "INFO",
+  session: "SESSION",
+  extend: "EXTEND",
+  code: "CODE",
+  jobs: "JOBS",
+  advanced: "ADVANCED",
+};
+
 export function SlashSuggestions({
   matches,
   selectedIndex,
+  groupMode,
+  advancedHidden,
 }: SlashSuggestionsProps): React.ReactElement | null {
   if (matches === null) return null;
   if (matches.length === 0) {
@@ -50,16 +44,15 @@ export function SlashSuggestions({
       </Box>
     );
   }
-  // Limit rows so the suggestion list never dwarfs the rest of the
-  // UI. Keep the currently-selected row in view by sliding the
-  // window when the selection is near either edge.
-  const MAX = 8;
+  const MAX = groupMode ? 24 : 8;
   const total = matches.length;
   const windowStart =
     total <= MAX ? 0 : Math.max(0, Math.min(selectedIndex - Math.floor(MAX / 2), total - MAX));
   const shown = matches.slice(windowStart, windowStart + MAX);
   const hiddenAbove = windowStart;
   const hiddenBelow = total - windowStart - shown.length;
+  let lastGroup: SlashGroup | null = null;
+  if (windowStart > 0) lastGroup = matches[windowStart - 1]?.group ?? null;
   return (
     <Box flexDirection="column" paddingX={1} marginTop={1}>
       <Box>
@@ -69,10 +62,27 @@ export function SlashSuggestions({
         <Text dimColor>{`${total} command${total === 1 ? "" : "s"}`}</Text>
         {hiddenAbove > 0 ? <Text dimColor>{`   ↑ ${hiddenAbove} above`}</Text> : null}
       </Box>
-      {shown.map((spec, i) => (
-        <SuggestionRow key={spec.cmd} spec={spec} isSelected={windowStart + i === selectedIndex} />
-      ))}
+      {shown.map((spec, i) => {
+        const idx = windowStart + i;
+        const showHeader = groupMode && spec.group !== lastGroup;
+        lastGroup = spec.group;
+        return (
+          <React.Fragment key={spec.cmd}>
+            {showHeader ? (
+              <Box marginTop={idx === 0 ? 0 : 1}>
+                <Text dimColor>{`  ${GROUP_LABEL[spec.group]}`}</Text>
+              </Box>
+            ) : null}
+            <SuggestionRow spec={spec} isSelected={idx === selectedIndex} />
+          </React.Fragment>
+        );
+      })}
       {hiddenBelow > 0 ? <Text dimColor>{`   ↓ ${hiddenBelow} below`}</Text> : null}
+      {groupMode && advancedHidden && advancedHidden > 0 ? (
+        <Box marginTop={1}>
+          <Text dimColor>{`  + ${advancedHidden} advanced  ·  type a letter to search`}</Text>
+        </Box>
+      ) : null}
       <Box marginTop={0}>
         <Text dimColor>{"  ↑↓ navigate · Tab / ⏎ pick · esc cancel"}</Text>
       </Box>
@@ -86,6 +96,7 @@ function SuggestionRow({ spec, isSelected }: { spec: SlashCommandSpec; isSelecte
   const key = `slash.${spec.cmd}.description`;
   const translated = t(key);
   const summary = translated === key ? spec.summary : translated;
+  const aliasHint = spec.aliases?.length ? ` · /${spec.aliases.join(" /")}` : "";
   return (
     <Box>
       <Text color={isSelected ? COLOR.primary : COLOR.info} bold={isSelected}>
@@ -99,6 +110,7 @@ function SuggestionRow({ spec, isSelected }: { spec: SlashCommandSpec; isSelecte
       <Text color={isSelected ? COLOR.user : COLOR.info} dimColor={!isSelected}>
         {summary}
       </Text>
+      {aliasHint ? <Text dimColor>{aliasHint}</Text> : null}
     </Box>
   );
 }

src/cli/ui/StatsPanel.tsx

@@ -6,6 +6,7 @@ import type { EditMode } from "../../config.js";
 import type { SessionSummary } from "../../telemetry/stats.js";
 import { Bar, ChromeRule } from "./primitives.js";
 import { COLOR, GRADIENT } from "./theme.js";
+import { formatBalance, formatCost } from "./theme/tokens.js";
 
 const COLD_START_TURNS = 3;
 
@@ -91,9 +92,9 @@ function ChromeRow({
   const cacheColor =
     summary.cacheHitRatio >= 0.7 ? COLOR.ok : summary.cacheHitRatio >= 0.4 ? COLOR.warn : COLOR.err;
   const balanceLabel = balance
-    ? `[w ${balance.currency === "USD" ? "$" : ""}${balance.total.toFixed(2)}${balance.currency !== "USD" ? ` ${balance.currency}` : ""}]`
+    ? `[${formatBalance(balance.total, balance.currency, { label: true })}]`
     : "";
-  const costLabel = `[$${summary.totalCostUsd.toFixed(4)}]`;
+  const costLabel = `[${formatCost(summary.totalCostUsd, balance?.currency)}]`;
   const cacheLabel = "[c ▰▰▰▰▰▰ 100%]";
   const updateLabel = updateAvailable ? `↑ ${updateAvailable}` : "";
 

src/cli/ui/WelcomeBanner.tsx

@@ -9,6 +9,8 @@ import { FG, TONE } from "./theme/tokens.js";
 export interface WelcomeBannerProps {
   /** True when running `reasonix code`. Surfaces code-mode hints. */
   inCodeMode?: boolean;
+  /** Pinned workspace root — only meaningful in code mode. Surfaced so first-time users see they can pass --dir at next launch. */
+  workspaceRoot?: string;
   /** Live URL of the embedded dashboard, or null when it isn't running. */
   dashboardUrl?: string | null;
   /** Bumped on language change; forces re-render so t() picks up new locale. */
@@ -19,6 +21,7 @@ const HINTS = ["/help", "/init", "/memory", "/cost"] as const;
 
 export function WelcomeBanner({
   inCodeMode,
+  workspaceRoot,
   dashboardUrl,
 }: WelcomeBannerProps): React.ReactElement {
   const tagline = inCodeMode ? t("ui.taglineCode") : t("ui.taglineChat");
@@ -66,6 +69,15 @@ export function WelcomeBanner({
         ))}
       </Box>
 
+      {inCodeMode && workspaceRoot ? (
+        <Box marginTop={1} flexDirection="row" gap={1}>
+          <Text color={TONE.brand}>{"▸ workspace"}</Text>
+          <Text color={FG.faint}>{"·"}</Text>
+          <Text color={FG.body}>{workspaceRoot}</Text>
+          <Text color={FG.faint}>{"  (relaunch with --dir <path> to switch)"}</Text>
+        </Box>
+      ) : null}
+
       {dashboardUrl ? (
         <Box marginTop={1} flexDirection="row" gap={1}>
           <Text color={TONE.brand} bold>

src/cli/ui/Wizard.tsx

@@ -1,20 +1,17 @@
 /**
  * First-run / re-configure wizard.
  *
- * Walks a new user through: API key → preset pick → MCP server pick →
- * per-server args → save. Saved output lives in `~/.reasonix/config.json`
- * so the next `reasonix chat` starts with everything already wired.
- *
- * The wizard is the antidote to "too many CLI flags" — a new user should
- * never have to read `--help` to get MCP + a sensible model combo
- * working. Everything a user could set via `--mcp`, `--harvest`,
- * `--branch`, etc. can be picked here in a few arrow-key presses.
+ * Walks a new user through: language → API key → preset pick → MCP server
+ * pick → per-server args → save. Saved output lives in
+ * `~/.reasonix/config.json` so the next `reasonix chat` starts with
+ * everything already wired.
  */
 
+import { mkdirSync, statSync } from "node:fs";
 import { Box, Text, useApp, useInput } from "ink";
 import TextInput from "ink-text-input";
 // biome-ignore lint/style/useImportType: JSX (jsx: "react") needs React as a value at runtime
-import React, { useState } from "react";
+import React, { useEffect, useState } from "react";
 import {
   type PresetName,
   type ReasonixConfig,
@@ -24,6 +21,16 @@ import {
   redactKey,
   writeConfig,
 } from "../../config.js";
+import {
+  detectSystemLanguage,
+  getLanguage,
+  getSupportedLanguages,
+  notifyLanguageChange,
+  onLanguageChange,
+  setLanguage,
+  t,
+} from "../../i18n/index.js";
+import type { LanguageCode } from "../../i18n/types.js";
 import { type CatalogEntry, MCP_CATALOG } from "../../mcp/catalog.js";
 import { MultiSelect, type SelectItem, SingleSelect } from "./Select.js";
 import { PRESET_DESCRIPTIONS } from "./presets.js";
@@ -42,22 +49,31 @@ export interface WizardProps {
   };
 }
 
-type Step = "apiKey" | "preset" | "mcp" | "mcpArgs" | "review" | "saved";
+type Step = "language" | "apiKey" | "preset" | "mcp" | "mcpArgs" | "review" | "saved";
 
 interface WizardData {
+  language: LanguageCode;
   apiKey: string;
   preset: PresetName;
-  selectedCatalog: string[]; // entries from MCP_CATALOG by `name`
-  /** Captured user inputs per catalog entry that has `userArgs` (e.g. fs dir). */
+  selectedCatalog: string[];
   catalogArgs: Record<string, string>;
 }
 
 const CATALOG_BY_NAME = new Map(MCP_CATALOG.map((e) => [e.name, e]));
 
+const LANGUAGE_LABELS: Record<LanguageCode, string> = {
+  EN: "English",
+  "zh-CN": "简体中文",
+};
+
 export function Wizard({ onComplete, onCancel, existingApiKey, initial }: WizardProps) {
   const { exit } = useApp();
-  const [step, setStep] = useState<Step>(existingApiKey ? "preset" : "apiKey");
+  const [, setLanguageVersion] = useState(0);
+  useEffect(() => onLanguageChange(() => setLanguageVersion((v) => v + 1)), []);
+
+  const [step, setStep] = useState<Step>("language");
   const [data, setData] = useState<WizardData>({
+    language: getLanguage(),
     apiKey: existingApiKey ?? "",
     preset: initial?.preset ?? "auto",
     selectedCatalog: deriveInitialCatalog(initial?.mcp ?? []),
@@ -65,12 +81,24 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
   });
   const [error, setError] = useState<string | null>(null);
 
-  // Global Esc → cancel. Disabled once we've started saving to avoid
-  // ejecting out of a half-written state.
   useInput((_input, key) => {
     if (key.escape && step !== "saved" && onCancel) onCancel();
   });
 
+  if (step === "language") {
+    return (
+      <LanguageStep
+        initialValue={data.language}
+        onSubmit={(lang) => {
+          setLanguage(lang);
+          notifyLanguageChange();
+          setData((d) => ({ ...d, language: lang }));
+          setStep(existingApiKey ? "preset" : "apiKey");
+        }}
+      />
+    );
+  }
+
   if (step === "apiKey") {
     return (
       <ApiKeyStep
@@ -87,7 +115,7 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
 
   if (step === "preset") {
     return (
-      <StepFrame title="Pick a preset" step={1} total={3}>
+      <StepFrame title={t("wizard.presetTitle")} step={1} total={3}>
         <SingleSelect<PresetName>
           items={presetItems()}
           initialValue={data.preset}
@@ -97,7 +125,7 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
           }}
         />
         <Box marginTop={1}>
-          <Text dimColor>[↑↓] navigate · [Enter] confirm · [Esc] cancel</Text>
+          <Text dimColor>{t("wizard.selectFooter")}</Text>
         </Box>
       </StepFrame>
     );
@@ -105,17 +133,16 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
 
   if (step === "mcp") {
     return (
-      <StepFrame title="Which MCP servers should Reasonix wire up for you?" step={2} total={3}>
+      <StepFrame title={t("wizard.mcpTitle")} step={2} total={3}>
         <MultiSelect
           items={mcpItems()}
           initialSelected={data.selectedCatalog}
           onSubmit={(selected) => {
             setData((d) => ({ ...d, selectedCatalog: selected }));
-            // Only advance to the args step if any selected entry needs args.
             const needsArgs = selected.some((name) => CATALOG_BY_NAME.get(name)?.userArgs);
             setStep(needsArgs ? "mcpArgs" : "review");
           }}
-          footer="[↑↓] navigate  ·  [Space] toggle  ·  [Enter] confirm  ·  [Esc] cancel  ·  empty = skip"
+          footer={t("wizard.mcpFooterMulti")}
         />
       </StepFrame>
     );
@@ -151,13 +178,21 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
   if (step === "review") {
     const specs = data.selectedCatalog.map((name) => buildSpec(name, data.catalogArgs));
     return (
-      <StepFrame title="Ready to save" step={3} total={3}>
+      <StepFrame title={t("wizard.reviewTitle")} step={3} total={3}>
         <Box flexDirection="column">
-          <SummaryLine label="API key" value={redactKey(data.apiKey)} />
-          <SummaryLine label="Preset" value={data.preset} />
           <SummaryLine
-            label="MCP"
-            value={specs.length === 0 ? "(none)" : `${specs.length} server(s)`}
+            label={t("wizard.reviewLabelLanguage")}
+            value={LANGUAGE_LABELS[data.language]}
+          />
+          <SummaryLine label={t("wizard.reviewLabelApiKey")} value={redactKey(data.apiKey)} />
+          <SummaryLine label={t("wizard.reviewLabelPreset")} value={data.preset} />
+          <SummaryLine
+            label={t("wizard.reviewLabelMcp")}
+            value={
+              specs.length === 0
+                ? t("wizard.reviewMcpNone")
+                : t("wizard.reviewMcpServers", { count: specs.length })
+            }
           />
           {specs.map((spec, i) => (
             // biome-ignore lint/suspicious/noArrayIndexKey: review-only render, order fixed
@@ -166,7 +201,7 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
             </Box>
           ))}
           <Box marginTop={1}>
-            <Text>Saves to {defaultConfigPath()}</Text>
+            <Text>{t("wizard.reviewSavesTo", { path: defaultConfigPath() })}</Text>
           </Box>
           {error ? (
             <Box marginTop={1}>
@@ -174,7 +209,7 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
             </Box>
           ) : null}
           <Box marginTop={1}>
-            <Text dimColor>[Enter] save · [Esc] cancel</Text>
+            <Text dimColor>{t("wizard.reviewFooter")}</Text>
           </Box>
         </Box>
         <ReviewConfirm
@@ -195,7 +230,7 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
               setStep("saved");
               onComplete(next);
             } catch (e) {
-              setError(`Could not save config: ${(e as Error).message}`);
+              setError(t("wizard.reviewSaveError", { message: (e as Error).message }));
             }
           }}
         />
@@ -203,17 +238,16 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
     );
   }
 
-  // saved
   return (
     <Box flexDirection="column" borderStyle="round" borderColor="green" paddingX={1}>
       <Text bold color="green">
-        ▸ Saved.
+        {t("wizard.savedTitle")}
       </Text>
       <Box marginTop={1}>
-        <Text>Run `reasonix` any time to start chatting — your settings are remembered.</Text>
+        <Text>{t("ui.welcome")}</Text>
       </Box>
       <Box marginTop={1}>
-        <Text dimColor>[Enter] to exit</Text>
+        <Text dimColor>{t("wizard.savedFooter")}</Text>
       </Box>
       <ExitOnEnter onExit={exit} />
     </Box>
@@ -222,6 +256,38 @@ export function Wizard({ onComplete, onCancel, existingApiKey, initial }: Wizard
 
 // ---------- step components ----------
 
+function LanguageStep({
+  initialValue,
+  onSubmit,
+}: {
+  initialValue: LanguageCode;
+  onSubmit: (lang: LanguageCode) => void;
+}) {
+  const items: SelectItem<LanguageCode>[] = getSupportedLanguages().map((code) => ({
+    value: code,
+    label: LANGUAGE_LABELS[code],
+    hint: code === detectSystemLanguage() ? "(detected)" : undefined,
+  }));
+  return (
+    <Box flexDirection="column" borderStyle="round" borderColor="cyan" paddingX={1}>
+      <Text bold color="cyan">
+        {t("wizard.languageTitle")}
+      </Text>
+      <Box marginTop={1}>
+        <Text dimColor>{t("wizard.languageSubtitle")}</Text>
+      </Box>
+      <Box marginTop={1}>
+        <SingleSelect<LanguageCode>
+          items={items}
+          initialValue={initialValue}
+          onSubmit={onSubmit}
+          footer={t("wizard.selectFooter")}
+        />
+      </Box>
+    </Box>
+  );
+}
+
 function ApiKeyStep({
   onSubmit,
   error,
@@ -235,16 +301,16 @@ function ApiKeyStep({
   return (
     <Box flexDirection="column" borderStyle="round" borderColor="cyan" paddingX={1}>
       <Text bold color="cyan">
-        Welcome to Reasonix.
+        {t("wizard.welcomeTitle")}
       </Text>
       <Box marginTop={1}>
-        <Text>Paste your DeepSeek API key to get started.</Text>
+        <Text>{t("wizard.apiKeyPrompt")}</Text>
       </Box>
-      <Text dimColor>Get one at: https://platform.deepseek.com/api_keys</Text>
-      <Text dimColor>Saved locally to {defaultConfigPath()}</Text>
+      <Text dimColor>{t("wizard.apiKeyGetOne")}</Text>
+      <Text dimColor>{t("wizard.apiKeySavedLocally", { path: defaultConfigPath() })}</Text>
       <Box marginTop={1}>
         <Text bold color="cyan">
-          {"key › "}
+          {t("wizard.apiKeyInputLabel")}
         </Text>
         <TextInput
           value={value}
@@ -252,7 +318,7 @@ function ApiKeyStep({
           onSubmit={(raw) => {
             const trimmed = raw.trim();
             if (!isPlausibleKey(trimmed)) {
-              onError("Doesn't look like a DeepSeek key. They start with 'sk-' and are 30+ chars.");
+              onError(t("wizard.apiKeyInvalid"));
               setValue("");
               return;
             }
@@ -268,7 +334,7 @@ function ApiKeyStep({
         </Box>
       ) : value ? (
         <Box marginTop={1}>
-          <Text dimColor>preview: {redactKey(value)}</Text>
+          <Text dimColor>{t("wizard.apiKeyPreview", { redacted: redactKey(value) })}</Text>
         </Box>
       ) : null}
     </Box>
@@ -287,8 +353,54 @@ function McpArgsStep({
   onError: (e: string | null) => void;
 }) {
   const [value, setValue] = useState("");
+  const [pendingCreate, setPendingCreate] = useState<string | null>(null);
+
+  useInput((input, key) => {
+    if (!pendingCreate) return;
+    const ch = input.toLowerCase();
+    if (ch === "y" || key.return) {
+      try {
+        mkdirSync(pendingCreate, { recursive: true });
+        const created = pendingCreate;
+        setPendingCreate(null);
+        setValue("");
+        onError(null);
+        onSubmit(created);
+      } catch (e) {
+        onError(
+          t("wizard.mcpArgsDirCreateFailed", {
+            path: pendingCreate,
+            message: (e as Error).message,
+          }),
+        );
+        setPendingCreate(null);
+      }
+    } else if (ch === "n" || key.escape) {
+      setPendingCreate(null);
+      onError(null);
+    }
+  });
+
+  if (pendingCreate) {
+    return (
+      <StepFrame title={t("wizard.mcpArgsTitle", { name: entry.name })} step={2} total={3}>
+        <Box flexDirection="column">
+          <Text>{t("wizard.mcpArgsDirMissing", { path: pendingCreate })}</Text>
+          <Box marginTop={1}>
+            <Text dimColor>{t("wizard.mcpArgsDirCreateHint")}</Text>
+          </Box>
+          {error ? (
+            <Box marginTop={1}>
+              <Text color="red">{error}</Text>
+            </Box>
+          ) : null}
+        </Box>
+      </StepFrame>
+    );
+  }
+
   return (
-    <StepFrame title={`Configure ${entry.name}`} step={2} total={3}>
+    <StepFrame title={t("wizard.mcpArgsTitle", { name: entry.name })} step={2} total={3}>
       <Box flexDirection="column">
         <Text>{entry.summary}</Text>
         {entry.note ? (
@@ -297,7 +409,7 @@ function McpArgsStep({
           </Box>
         ) : null}
         <Box marginTop={1}>
-          <Text>Required parameter: </Text>
+          <Text>{t("wizard.mcpArgsRequiredParam")}</Text>
           <Text bold>{entry.userArgs}</Text>
         </Box>
         <Box marginTop={1}>
@@ -311,9 +423,20 @@ function McpArgsStep({
             onSubmit={(raw) => {
               const trimmed = raw.trim();
               if (!trimmed) {
-                onError(`${entry.name} needs a value — got an empty string.`);
+                onError(t("wizard.mcpArgsEmpty", { name: entry.name }));
                 return;
               }
+              if (entry.name === "filesystem") {
+                const check = checkFilesystemPath(trimmed);
+                if (check.kind === "missing") {
+                  setPendingCreate(trimmed);
+                  return;
+                }
+                if (check.kind === "not-a-dir") {
+                  onError(t("wizard.mcpArgsNotADir", { path: trimmed }));
+                  return;
+                }
+              }
               onSubmit(trimmed);
               setValue("");
             }}
@@ -330,6 +453,14 @@ function McpArgsStep({
   );
 }
 
+function checkFilesystemPath(p: string): { kind: "ok" | "missing" | "not-a-dir" } {
+  try {
+    return { kind: statSync(p).isDirectory() ? "ok" : "not-a-dir" };
+  } catch {
+    return { kind: "missing" };
+  }
+}
+
 function ReviewConfirm({ onConfirm }: { onConfirm: () => void }) {
   useInput((_i, key) => {
     if (key.return) onConfirm();
@@ -344,8 +475,6 @@ function ExitOnEnter({ onExit }: { onExit: () => void }) {
   return null;
 }
 
-// ---------- small bits ----------
-
 function StepFrame({
   title,
   step,
@@ -360,9 +489,7 @@ function StepFrame({
   return (
     <Box flexDirection="column" borderStyle="round" borderColor="cyan" paddingX={1}>
       <Box>
-        <Text dimColor>
-          Step {step}/{total} ·{" "}
-        </Text>
+        <Text dimColor>{t("wizard.stepCounter", { step, total })}</Text>
         <Text bold color="cyan">
           {title}
         </Text>
@@ -383,8 +510,6 @@ function SummaryLine({ label, value }: { label: string; value: string }) {
   );
 }
 
-// ---------- data helpers ----------
-
 function presetItems(): SelectItem<PresetName>[] {
   return (["auto", "flash", "pro"] as const).map((name) => ({
     value: name as PresetName,
@@ -396,7 +521,7 @@ function presetItems(): SelectItem<PresetName>[] {
 function mcpItems(): SelectItem<string>[] {
   return MCP_CATALOG.map((entry) => {
     const hintParts: string[] = [entry.summary];
-    if (entry.userArgs) hintParts.push(`(you'll provide ${entry.userArgs})`);
+    if (entry.userArgs) hintParts.push(t("wizard.mcpUserArgsHint", { arg: entry.userArgs }));
     if (entry.note) hintParts.push(entry.note);
     return {
       value: entry.name,
@@ -413,9 +538,6 @@ function placeholderFor(entry: CatalogEntry): string {
 }
 
 function deriveInitialCatalog(existingSpecs: string[]): string[] {
-  // Best-effort recovery: if the user previously picked catalog entries,
-  // the spec strings look like `name=npx -y <pkg> ...`. Match by package
-  // so reconfigure pre-checks the same boxes.
   const packageToName = new Map(MCP_CATALOG.map((e) => [e.package, e.name]));
   const out: string[] = [];
   for (const spec of existingSpecs) {
@@ -436,12 +558,15 @@ function deriveInitialCatalog(existingSpecs: string[]): string[] {
  */
 export function buildSpec(name: string, argsByName: Record<string, string>): string {
   const entry = CATALOG_BY_NAME.get(name);
-  if (!entry) return name; // shouldn't happen; fall back gracefully
+  if (!entry) return name;
   const userArg = entry.userArgs ? argsByName[name] : undefined;
   const tail = userArg ? ` ${quoteIfNeeded(userArg)}` : "";
   return `${entry.name}=npx -y ${entry.package}${tail}`;
 }
 
 function quoteIfNeeded(s: string): string {
-  return /\s|"/.test(s) ? `"${s.replace(/"/g, '\\"')}"` : s;
+  // Escape backslashes BEFORE quotes — otherwise a trailing `\` in the
+  // input would consume the closing quote when a downstream parser
+  // un-escapes the output (CodeQL js/incomplete-sanitization).
+  return /\s|"/.test(s) ? `"${s.replace(/\\/g, "\\\\").replace(/"/g, '\\"')}"` : s;
 }

src/cli/ui/alt-screen.ts

@@ -1,39 +0,0 @@
-/** Alt-screen lifecycle — restore must be idempotent and run on SIGINT/SIGTERM/exit so the user's main buffer always comes back. */
-
-import { useEffect } from "react";
-
-function isInteractiveTty(): boolean {
-  return Boolean(process.stdout?.isTTY);
-}
-
-export function useAltScreen(): void {
-  useEffect(() => {
-    if (!isInteractiveTty()) return;
-    // 1002 = button-event tracking (press + drag + release); 1006 = SGR coords.
-    // Shift+drag bypasses tracking for native terminal selection.
-    process.stdout.write("\x1b[?1049h\x1b[2J\x1b[H\x1b[?1002h\x1b[?1006h");
-
-    let restored = false;
-    const restore = () => {
-      if (restored) return;
-      restored = true;
-      try {
-        process.stdout.write("\x1b[?1006l\x1b[?1002l\x1b[?1000l\x1b[?1049l");
-      } catch {
-        /* terminal closed */
-      }
-    };
-
-    process.once("exit", restore);
-    process.once("SIGINT", () => {
-      restore();
-      process.exit(130);
-    });
-    process.once("SIGTERM", () => {
-      restore();
-      process.exit(143);
-    });
-
-    return restore;
-  }, []);
-}

src/cli/ui/cards/BranchCard.tsx

@@ -1,35 +0,0 @@
-import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React from "react";
-import { Card } from "../primitives/Card.js";
-import { CardHeader } from "../primitives/CardHeader.js";
-import type { BranchCard as BranchCardData } from "../state/cards.js";
-import { CARD, FG, TONE } from "../theme/tokens.js";
-
-const BAR_CELLS = 28;
-
-export function BranchCard({ card }: { card: BranchCardData }): React.ReactElement {
-  const ratio = card.total > 0 ? card.completed / card.total : 0;
-  const filled = Math.max(0, Math.min(BAR_CELLS, Math.round(ratio * BAR_CELLS)));
-  const tone = card.done ? TONE.ok : CARD.branch.color;
-  return (
-    <Card tone={tone}>
-      <CardHeader
-        glyph="⎇"
-        tone={tone}
-        title={card.done ? "branching done" : "branching"}
-        meta={[`${card.completed} of ${card.total} samples`]}
-      />
-      <Box flexDirection="row" gap={1}>
-        <Text color={tone}>{"█".repeat(filled)}</Text>
-        <Text color={FG.faint}>{"░".repeat(BAR_CELLS - filled)}</Text>
-        <Text color={FG.faint}>{`${(ratio * 100).toFixed(0)}%`}</Text>
-      </Box>
-      {!card.done && card.completed > 0 ? (
-        <Text color={FG.faint}>
-          {`latest · #${card.latestIndex} · T=${card.latestTemperature.toFixed(2)} · ${card.latestUncertainties} unc`}
-        </Text>
-      ) : null}
-    </Card>
-  );
-}

src/cli/ui/cards/CardRenderer.tsx

@@ -1,9 +1,7 @@
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
 import type { Card } from "../state/cards.js";
 import { FG } from "../theme/tokens.js";
-import { BranchCard } from "./BranchCard.js";
 import { CtxCard } from "./CtxCard.js";
 import { DiffCard } from "./DiffCard.js";
 import { DoctorCard } from "./DoctorCard.js";
@@ -21,11 +19,14 @@ import { UsageCard } from "./UsageCard.js";
 import { UserCard } from "./UserCard.js";
 import { WarnCard } from "./WarnCard.js";
 
-export function CardRenderer({ card }: { card: Card }): React.ReactElement {
-  // No marginTop here — Card primitive owns inter-card spacing. Doubling it
-  // bakes 2 blank rows into every static-emitted block of bytes.
+// Memoized so the cards array re-rendering (every store update) only
+// reconciles cards whose object identity actually changed — the reducer
+// keeps prior cards reference-stable, so unchanged history skips work.
+export const CardRenderer = React.memo(function CardRenderer({
+  card,
+}: { card: Card }): React.ReactElement {
   return <Box flexDirection="column">{renderCard(card)}</Box>;
-}
+});
 
 function renderCard(card: Card): React.ReactElement {
   switch (card.kind) {
@@ -61,8 +62,6 @@ function renderCard(card: Card): React.ReactElement {
       return <CtxCard card={card} />;
     case "doctor":
       return <DoctorCard card={card} />;
-    case "branch":
-      return <BranchCard card={card} />;
     default:
       return <FallbackCard card={card} />;
   }

src/cli/ui/cards/StreamingCard.tsx

@@ -1,7 +1,9 @@
 import { Box, Text, useStdout } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React from "react";
+import React, { useContext } from "react";
 import { clipToCells, wrapToCells } from "../../../frame/width.js";
+import { countTokens } from "../../../tokenizer.js";
+import { LiveExpandContext } from "../layout/LiveExpandContext.js";
 import { useReserveRows } from "../layout/viewport-budget.js";
 import { Markdown } from "../markdown.js";
 import { Card } from "../primitives/Card.js";
@@ -10,17 +12,49 @@ import { PILL_MODEL, Pill, modelBadgeFor } from "../primitives/Pill.js";
 import { Spinner } from "../primitives/Spinner.js";
 import type { StreamingCard as StreamingCardData } from "../state/cards.js";
 import { FG, TONE, TONE_ACTIVE } from "../theme/tokens.js";
+import { useSlowTick } from "../ticker.js";
 
 /** Streaming preview tail length — bounded live region so chunks don't thrash whole-card layout. */
 const STREAMING_PREVIEW_LINES = 4;
+/** Expanded mode shows up to this many lines so the card can't swallow the whole viewport. */
+const EXPANDED_MAX_LINES = 60;
+
+const MIN_ELAPSED_MS_FOR_RATE = 500;
+const MIN_TOKENS_FOR_RATE = 4;
+
+function formatTokenCount(n: number): string {
+  if (n >= 10000) return `${(n / 1000).toFixed(1)}k`;
+  if (n >= 1000) return `${(n / 1000).toFixed(2)}k`;
+  return String(n);
+}
+
+function tokenRate(
+  text: string,
+  startTs: number,
+  endTs: number,
+): { tokens: number; tps: number | null } {
+  const tokens = countTokens(text);
+  const elapsedMs = endTs - startTs;
+  if (elapsedMs < MIN_ELAPSED_MS_FOR_RATE || tokens < MIN_TOKENS_FOR_RATE) {
+    return { tokens, tps: null };
+  }
+  return { tokens, tps: Math.round((tokens * 1000) / elapsedMs) };
+}
+
+const PILL_RATE = { bg: "#11141a", fg: "#8b949e" } as const;
 
 export function StreamingCard({ card }: { card: StreamingCardData }): React.ReactElement {
   const { stdout } = useStdout();
   const cols = stdout?.columns ?? 80;
+  const expanded = useContext(LiveExpandContext);
+  const reserveCap = expanded ? EXPANDED_MAX_LINES + 2 : STREAMING_PREVIEW_LINES + 2;
   useReserveRows("stream", {
     min: STREAMING_PREVIEW_LINES + 1,
-    max: STREAMING_PREVIEW_LINES + 2,
+    max: reserveCap,
   });
+  // Re-render at 1Hz so the rate keeps updating even when chunks stall.
+  // Frozen once `card.done` is true — settled cards render via Static.
+  useSlowTick();
 
   const modelBadge = card.model ? modelBadgeFor(card.model) : null;
   const modelPill = modelBadge ? (
@@ -28,9 +62,24 @@ export function StreamingCard({ card }: { card: StreamingCardData }): React.Reac
   ) : null;
 
   if (card.done && !card.aborted) {
+    const { tokens, tps } = tokenRate(card.text, card.ts, card.endedAt ?? Date.now());
+    const ratePill =
+      tokens >= MIN_TOKENS_FOR_RATE && tps !== null ? (
+        <Pill label={`${formatTokenCount(tokens)} tok · ${tps} t/s`} {...PILL_RATE} bold={false} />
+      ) : null;
     return (
       <Card tone={TONE.ok}>
-        <CardHeader glyph="‹" tone={TONE.ok} title="reply" right={modelPill} />
+        <CardHeader
+          glyph="‹"
+          tone={TONE.ok}
+          title="reply"
+          right={
+            <>
+              {ratePill}
+              {modelPill}
+            </>
+          }
+        />
         <Markdown text={card.text} />
       </Card>
     );
@@ -39,12 +88,23 @@ export function StreamingCard({ card }: { card: StreamingCardData }): React.Reac
   const lineCells = Math.max(20, cols - 4);
   const allLines = card.text.length > 0 ? card.text.split("\n") : [""];
   const visualLines = allLines.flatMap((l) => wrapToCells(l, lineCells));
-  const visible = visualLines.slice(-STREAMING_PREVIEW_LINES);
+  const cap = expanded ? EXPANDED_MAX_LINES : STREAMING_PREVIEW_LINES;
+  const visible = visualLines.slice(-cap);
+  const droppedAbove = Math.max(0, visualLines.length - visible.length);
   const aborted = !!card.aborted;
   const headColor = aborted ? TONE.err : TONE_ACTIVE.brand;
   const glyph = aborted ? "‹" : "◈";
   const headLabel = aborted ? "aborted" : "writing…";
 
+  const { tokens: liveTokens, tps: liveTps } = tokenRate(card.text, card.ts, Date.now());
+  const liveRatePill =
+    !aborted && liveTokens >= MIN_TOKENS_FOR_RATE && liveTps !== null ? (
+      <Pill label={`${liveTps} t/s`} {...PILL_RATE} bold={false} />
+    ) : null;
+  const expandPill = !aborted ? (
+    <Pill label={expanded ? "expanded ⌃o" : "preview ⌃o"} {...PILL_RATE} bold={false} />
+  ) : null;
+
   return (
     <Card tone={headColor}>
       <CardHeader
@@ -53,13 +113,20 @@ export function StreamingCard({ card }: { card: StreamingCardData }): React.Reac
         title={headLabel}
         right={
           <>
+            {liveRatePill}
+            {expandPill}
             {aborted ? null : <Spinner kind="braille" color={TONE_ACTIVE.brand} />}
             {modelPill}
           </>
         }
       />
+      {expanded && droppedAbove > 0 ? (
+        <Text
+          color={FG.faint}
+        >{`⋯ ${droppedAbove} earlier line${droppedAbove === 1 ? "" : "s"} above`}</Text>
+      ) : null}
       {visible.map((line, i) => (
-        <Box key={`${card.id}:${allLines.length - visible.length + i}`} flexDirection="row">
+        <Box key={`${card.id}:${visualLines.length - visible.length + i}`} flexDirection="row">
           <Text color={aborted ? FG.meta : FG.body}>{clipToCells(line, lineCells)}</Text>
         </Box>
       ))}

src/cli/ui/cards/UsageCard.tsx

@@ -4,7 +4,7 @@ import React from "react";
 import { Card } from "../primitives/Card.js";
 import { CardHeader } from "../primitives/CardHeader.js";
 import type { UsageCard as UsageCardData } from "../state/cards.js";
-import { FG, TONE, formatCNY } from "../theme/tokens.js";
+import { FG, TONE, formatBalance, formatCost } from "../theme/tokens.js";
 
 const BAR_CELLS = 30;
 
@@ -32,7 +32,7 @@ export function UsageCard({ card }: { card: UsageCardData }): React.ReactElement
   const reasonRatio = card.tokens.reason / cap;
   const outputRatio = card.tokens.output / cap;
 
-  const headerMeta: string[] = [`turn ${card.turn}`, formatCNY(card.cost)];
+  const headerMeta: string[] = [`turn ${card.turn}`, formatCost(card.cost, card.balanceCurrency)];
   if (card.elapsedMs !== undefined) headerMeta.push(`${(card.elapsedMs / 1000).toFixed(1)}s`);
   return (
     <Card tone={FG.meta}>
@@ -66,11 +66,15 @@ export function UsageCard({ card }: { card: UsageCardData }): React.ReactElement
       </Box>
       <Box flexDirection="row" gap={1}>
         <Text color={FG.faint}>session</Text>
-        <Text bold color={FG.body}>{`⛁ ${formatCNY(card.sessionCost, 3)}`}</Text>
+        <Text bold color={FG.body}>
+          {`⛁ ${formatCost(card.sessionCost, card.balanceCurrency, 3)}`}
+        </Text>
         {card.balance !== undefined ? (
           <>
             <Text color={FG.faint}>· balance</Text>
-            <Text bold color={TONE.brand}>{`¥${card.balance.toFixed(2)}`}</Text>
+            <Text bold color={TONE.brand}>
+              {formatBalance(card.balance, card.balanceCurrency)}
+            </Text>
           </>
         ) : null}
       </Box>
@@ -89,9 +93,9 @@ function CompactUsageRow({ card }: { card: UsageCardData }): React.ReactElement
       </Text>
       <Text color={FG.faint}>· cache</Text>
       <Text color={TONE.ok}>{`${(card.cacheHit * 100).toFixed(0)}%`}</Text>
-      <Text color={FG.faint}>{`· ${formatCNY(card.cost)}${elapsed}`}</Text>
+      <Text color={FG.faint}>{`· ${formatCost(card.cost, card.balanceCurrency)}${elapsed}`}</Text>
       {card.balance !== undefined ? (
-        <Text color={TONE.brand}>{`· ¥${card.balance.toFixed(2)}`}</Text>
+        <Text color={TONE.brand}>{`· ${formatBalance(card.balance, card.balanceCurrency)}`}</Text>
       ) : null}
     </Box>
   );

src/cli/ui/clipboard.ts

@@ -1,6 +1,6 @@
 /** OSC 52 clipboard write + temp-file fallback. */
 
-import { writeFileSync } from "node:fs";
+import { mkdtempSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
 
@@ -13,7 +13,11 @@ export interface ClipboardWrite {
 }
 
 export function writeClipboard(text: string): ClipboardWrite {
-  const filePath = join(tmpdir(), `reasonix-clip-${Date.now()}.txt`);
+  // mkdtemp creates a private 0700 directory atomically — keeps the
+  // file out of the shared tmp namespace where a sibling process can
+  // race or read it (CodeQL js/insecure-temporary-file).
+  const dir = mkdtempSync(join(tmpdir(), "reasonix-clip-"));
+  const filePath = join(dir, "clip.txt");
   let osc52 = false;
   if (text.length <= OSC_52_LIMIT) {
     const b64 = Buffer.from(text, "utf8").toString("base64");

src/cli/ui/dashboard/use-picker-broadcast.ts

@@ -0,0 +1,90 @@
+import type { MutableRefObject } from "react";
+import { useEffect } from "react";
+import type {
+  DashboardEvent,
+  PickerAction,
+  PickerItem,
+  PickerResolution,
+} from "../../../server/context.js";
+
+export interface PickerSnapshot {
+  pickerKind: string;
+  title: string;
+  query?: string;
+  items: PickerItem[];
+  actions: PickerAction[];
+  hasMore?: boolean;
+  hint?: string;
+}
+
+export interface ViewerSnapshot {
+  viewerKind: string;
+  title: string;
+  body?: string;
+  steps?: Array<{ id: string; title: string; status: "done" | "queued" }>;
+  meta?: string;
+}
+
+export interface ViewerBroadcastPorts {
+  broadcast: (ev: DashboardEvent) => void;
+  resolverRef: MutableRefObject<(() => void) | null>;
+  snapshotRef: MutableRefObject<ViewerSnapshot | null>;
+}
+
+/** Read-only sibling of `usePickerBroadcast` — viewer modals carry no selection so only `close` flows back. */
+export function useViewerBroadcast(
+  active: boolean,
+  snapshot: ViewerSnapshot,
+  onClose: () => void,
+  ports: ViewerBroadcastPorts,
+): void {
+  const { broadcast, resolverRef, snapshotRef } = ports;
+
+  useEffect(() => {
+    if (!active) return;
+    return () => {
+      broadcast({ kind: "modal-down", modalKind: "viewer" });
+      if (resolverRef.current) resolverRef.current = null;
+      if (snapshotRef.current) snapshotRef.current = null;
+    };
+  }, [active, broadcast, resolverRef, snapshotRef]);
+
+  useEffect(() => {
+    if (!active) return;
+    snapshotRef.current = snapshot;
+    resolverRef.current = onClose;
+    broadcast({ kind: "modal-up", modal: { kind: "viewer", ...snapshot } });
+  }, [active, snapshot, onClose, broadcast, resolverRef, snapshotRef]);
+}
+
+export interface PickerBroadcastPorts {
+  broadcast: (ev: DashboardEvent) => void;
+  resolverRef: MutableRefObject<((res: PickerResolution) => void) | null>;
+  snapshotRef: MutableRefObject<PickerSnapshot | null>;
+}
+
+/** Mirrors a TUI picker into the dashboard via modal-up/down events. Caller passes stable refs from App.tsx so identity does not churn the effect. */
+export function usePickerBroadcast(
+  active: boolean,
+  snapshot: PickerSnapshot,
+  onResolve: (res: PickerResolution) => void,
+  ports: PickerBroadcastPorts,
+): void {
+  const { broadcast, resolverRef, snapshotRef } = ports;
+
+  useEffect(() => {
+    if (!active) return;
+    return () => {
+      broadcast({ kind: "modal-down", modalKind: "picker" });
+      if (resolverRef.current) resolverRef.current = null;
+      if (snapshotRef.current) snapshotRef.current = null;
+    };
+  }, [active, broadcast, resolverRef, snapshotRef]);
+
+  useEffect(() => {
+    if (!active) return;
+    snapshotRef.current = snapshot;
+    resolverRef.current = onResolve;
+    broadcast({ kind: "modal-up", modal: { kind: "picker", ...snapshot } });
+  }, [active, snapshot, onResolve, broadcast, resolverRef, snapshotRef]);
+}

src/cli/ui/drain-tty.ts

@@ -0,0 +1,39 @@
+/** stdin-queue drain on exit — eats stuck terminal-feature-detection responses (#365). */
+
+import process from "node:process";
+
+/** Eats stuck OSC/CPR/DA replies on exit so fish/bash don't print them as input (#365). */
+export async function drainTtyResponses(timeoutMs = 50): Promise<void> {
+  const stdin = process.stdin;
+  if (!stdin.isTTY && typeof (stdin as { setRawMode?: unknown }).setRawMode !== "function") {
+    return;
+  }
+  let raised = false;
+  try {
+    stdin.setRawMode(true);
+    raised = true;
+  } catch {
+    return;
+  }
+  stdin.resume();
+
+  await new Promise<void>((resolve) => {
+    const onData = (_chunk: Buffer | string): void => {
+      // Discard — anything pending here is a terminal-feature reply.
+    };
+    stdin.on("data", onData);
+    const timer = setTimeout(() => {
+      stdin.off("data", onData);
+      stdin.pause();
+      if (raised) {
+        try {
+          stdin.setRawMode(false);
+        } catch {
+          /* stdin may already be closed; ignore */
+        }
+      }
+      resolve();
+    }, timeoutMs);
+    timer.unref();
+  });
+}

src/cli/ui/hash-memory.ts

@@ -1,6 +1,6 @@
 /** `#` writes project memory, `#g` global; `##+` stays a markdown heading; `\#` escapes and submits the literal `#`. */
 
-import { appendFileSync, existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
+import { closeSync, fstatSync, mkdirSync, openSync, readSync, writeSync } from "node:fs";
 import { homedir } from "node:os";
 import { dirname, join } from "node:path";
 import { PROJECT_MEMORY_FILE } from "../../memory/project.js";
@@ -79,18 +79,28 @@ function appendBulletToFile(path: string, note: string, newFileHeader: string):
   const trimmed = note.trim();
   if (!trimmed) throw new Error("note body cannot be empty");
   const bullet = `- ${trimmed}\n`;
-  if (!existsSync(path)) {
-    mkdirSync(dirname(path), { recursive: true });
-    writeFileSync(path, `${newFileHeader}${bullet}`, "utf8");
-    return { path, created: true };
-  }
-  let prefix = "";
+  mkdirSync(dirname(path), { recursive: true });
+  // One `a+` open covers both branches: O_APPEND lands every write
+  // atomically at end-of-file (concurrent appenders interleave whole
+  // bullets), O_CREAT creates the file when it's missing, and we use
+  // `fstat().size === 0` as the "we just created it" signal to decide
+  // whether to emit the file header. Single fd from open through
+  // write — no path-based check between (CodeQL js/file-system-race).
+  const fd = openSync(path, "a+");
   try {
-    const existing = readFileSync(path, "utf8");
-    if (existing.length > 0 && !existing.endsWith("\n")) prefix = "\n";
-  } catch {
-    // Unreadable but exists — let appendFileSync surface the real error.
+    const stat = fstatSync(fd);
+    if (stat.size === 0) {
+      writeSync(fd, `${newFileHeader}${bullet}`);
+      return { path, created: true };
+    }
+    // Existing file — peek the trailing byte to decide whether to
+    // insert a leading newline. Same fd → no separate stat→read race.
+    const tail = Buffer.alloc(1);
+    readSync(fd, tail, 0, 1, stat.size - 1);
+    const prefix = tail[0] !== 0x0a ? "\n" : "";
+    writeSync(fd, `${prefix}${bullet}`);
+    return { path, created: false };
+  } finally {
+    closeSync(fd);
   }
-  appendFileSync(path, `${prefix}${bullet}`, "utf8");
-  return { path, created: false };
 }

src/cli/ui/hooks/apply-slash-result.ts

@@ -6,13 +6,6 @@ import type { Scrollback } from "./useScrollback.js";
 
 export type SlashOutcome = { kind: "consumed" } | { kind: "resubmit"; text: string };
 
-interface ActiveLoopShape {
-  prompt: string;
-  intervalMs: number;
-  nextFireAt: number;
-  iter: number;
-}
-
 export interface ApplySlashResultContext {
   log: Scrollback;
   stdoutWrite: (chunk: string) => void;
@@ -20,10 +13,10 @@ export interface ApplySlashResultContext {
   syncPendingCount: () => void;
   session: string | null;
   codeModeOn: boolean;
-  activeLoopRef: MutableRefObject<ActiveLoopShape | null>;
+  isLoopActive: () => boolean;
   stopLoop: () => void;
   quitProcess: () => void;
-  promptHistory: MutableRefObject<string[]>;
+  pushHistory: (text: string) => void;
   /** The verbatim text the user typed; used for promptHistory bookkeeping. */
   text: string;
 }
@@ -34,14 +27,12 @@ export function applySlashResult(result: SlashResult, ctx: ApplySlashResultConte
     // the process is exiting. Use quitProcess (process.exit) rather than
     // Ink's exit(): the singleton stdin reader keeps a `data` listener
     // attached, so exit() unmounts React but leaves the event loop alive.
-    if (ctx.activeLoopRef.current) ctx.stopLoop();
+    if (ctx.isLoopActive()) ctx.stopLoop();
     ctx.quitProcess();
     return { kind: "consumed" };
   }
   if (result.clear) {
-    // 2J + 3J + H: nuke viewport AND scrollback (visible buffer + the
-    // off-screen lines). Then reset state.cards so React doesn't
-    // immediately repaint everything that was just wiped.
+    // 2J + 3J + H: visible buffer + scrollback + cursor home.
     ctx.stdoutWrite("\x1b[2J\x1b[3J\x1b[H");
     ctx.log.reset();
     if (result.info) ctx.log.pushInfo(result.info);
@@ -50,7 +41,7 @@ export function applySlashResult(result: SlashResult, ctx: ApplySlashResultConte
       clearPendingEdits(ctx.session);
       ctx.syncPendingCount();
     }
-    if (ctx.activeLoopRef.current) ctx.stopLoop();
+    if (ctx.isLoopActive()) ctx.stopLoop();
     return { kind: "consumed" };
   }
   if (result.info) {
@@ -77,6 +68,6 @@ export function applySlashResult(result: SlashResult, ctx: ApplySlashResultConte
   if (result.resubmit) {
     return { kind: "resubmit", text: result.resubmit };
   }
-  ctx.promptHistory.current.push(ctx.text);
+  ctx.pushHistory(ctx.text);
   return { kind: "consumed" };
 }

src/cli/ui/hooks/useChatScroll.ts

@@ -0,0 +1,51 @@
+import { useCallback, useEffect, useState } from "react";
+
+/** Rows advanced per PgUp / PgDn / arrow / wheel — small for smooth feel. */
+export const SCROLL_PAGE_ROWS = 3;
+
+export interface ChatScrollState {
+  /** How many rows of content are above the visible viewport. */
+  scrollRows: number;
+  /** True when the user is following the bottom (auto-advances on new content). */
+  pinned: boolean;
+  scrollUp: () => void;
+  scrollDown: () => void;
+  /** Jump straight to the latest content and resume auto-follow (End key). */
+  jumpToBottom: () => void;
+  /** CardStream calls this once it has measured inner/outer Box heights. */
+  setMaxScroll: (rows: number) => void;
+}
+
+/** Row-precision scroll state. CardStream reports maxScroll back via setMaxScroll once Ink has laid out the inner column. */
+export function useChatScroll(): ChatScrollState {
+  const [scrollRows, setScrollRows] = useState(0);
+  const [pinned, setPinned] = useState(true);
+  const [maxScroll, setMaxScrollState] = useState(0);
+
+  useEffect(() => {
+    if (pinned) setScrollRows(maxScroll);
+  }, [pinned, maxScroll]);
+
+  useEffect(() => {
+    if (scrollRows > maxScroll) setScrollRows(maxScroll);
+  }, [scrollRows, maxScroll]);
+
+  const scrollUp = useCallback(() => {
+    setPinned(false);
+    setScrollRows((o) => Math.max(0, o - SCROLL_PAGE_ROWS));
+  }, []);
+
+  const scrollDown = useCallback(() => {
+    setScrollRows((o) => {
+      const next = Math.min(maxScroll, o + SCROLL_PAGE_ROWS);
+      if (next >= maxScroll) setPinned(true);
+      return next;
+    });
+  }, [maxScroll]);
+
+  const jumpToBottom = useCallback(() => setPinned(true), []);
+
+  const setMaxScroll = useCallback((rows: number) => setMaxScrollState(rows), []);
+
+  return { scrollRows, pinned, scrollUp, scrollDown, jumpToBottom, setMaxScroll };
+}

src/cli/ui/hooks/useCodeMode.ts

@@ -0,0 +1,93 @@
+import { type MutableRefObject, useCallback } from "react";
+import {
+  type ApplyResult,
+  type EditBlock,
+  type EditSnapshot,
+  applyEditBlocks,
+  snapshotBeforeEdits,
+} from "../../../code/edit-blocks.js";
+import { clearPendingEdits, savePendingEdits } from "../../../code/pending-edits.js";
+import { formatEditResults, partitionEdits } from "../edit-history.js";
+
+export interface UseCodeModeResult {
+  /** /apply callback. Empty `indices` means "all"; specific 1-based indices apply only those. */
+  codeApply: (indices?: readonly number[]) => string;
+  /** /discard callback. Same indices semantics as codeApply. */
+  codeDiscard: (indices?: readonly number[]) => string;
+}
+
+export interface UseCodeModeOptions {
+  codeMode: boolean;
+  pendingEdits: MutableRefObject<EditBlock[]>;
+  currentRootDir: string;
+  session: string | null;
+  syncPendingCount: () => void;
+  recordEdit: (
+    source: string,
+    blocks: readonly EditBlock[],
+    results: readonly ApplyResult[],
+    snaps: readonly EditSnapshot[],
+  ) => void;
+}
+
+/** Slash-command callbacks for `/apply` and `/discard` over the pending-edits queue. Owns the partition / snapshot / save / sync sequence; AppInner just forwards the strings to its log. */
+export function useCodeMode(opts: UseCodeModeOptions): UseCodeModeResult {
+  const { codeMode, pendingEdits, currentRootDir, session, syncPendingCount, recordEdit } = opts;
+
+  const codeApply = useCallback(
+    (indices?: readonly number[]): string => {
+      if (!codeMode) return "not in code mode";
+      const blocks = pendingEdits.current;
+      if (blocks.length === 0) {
+        return "nothing pending — the model hasn't proposed edits since the last /apply or /discard.";
+      }
+      const useSubset = indices !== undefined && indices.length > 0;
+      const { selected, remaining } = useSubset
+        ? partitionEdits(blocks, indices)
+        : { selected: blocks, remaining: [] as EditBlock[] };
+      if (selected.length === 0) {
+        return "▸ no edits matched those indices — nothing applied. Use /apply with no args to commit them all.";
+      }
+      const snaps = snapshotBeforeEdits(selected, currentRootDir);
+      const results = applyEditBlocks(selected, currentRootDir);
+      const anyApplied = results.some((r) => r.status === "applied" || r.status === "created");
+      if (anyApplied) recordEdit("review-apply", selected, results, snaps);
+      pendingEdits.current = remaining;
+      if (remaining.length === 0) clearPendingEdits(session ?? null);
+      else savePendingEdits(session ?? null, remaining);
+      syncPendingCount();
+      const tail =
+        remaining.length > 0
+          ? `\n▸ ${remaining.length} edit block(s) still pending — /apply or /discard to clear them.`
+          : "";
+      return formatEditResults(results) + tail;
+    },
+    [codeMode, currentRootDir, session, syncPendingCount, recordEdit, pendingEdits],
+  );
+
+  const codeDiscard = useCallback(
+    (indices?: readonly number[]): string => {
+      const blocks = pendingEdits.current;
+      if (blocks.length === 0) return "nothing pending to discard.";
+      const useSubset = indices !== undefined && indices.length > 0;
+      const { selected, remaining } = useSubset
+        ? partitionEdits(blocks, indices)
+        : { selected: blocks, remaining: [] as EditBlock[] };
+      if (selected.length === 0) {
+        return "▸ no edits matched those indices — nothing discarded.";
+      }
+      pendingEdits.current = remaining;
+      if (remaining.length === 0) clearPendingEdits(session ?? null);
+      else savePendingEdits(session ?? null, remaining);
+      syncPendingCount();
+      const tail =
+        remaining.length > 0
+          ? `  (${remaining.length} block(s) still pending)`
+          : ". Nothing was written to disk.";
+      return `▸ discarded ${selected.length} pending edit block(s)${tail}`;
+    },
+    [session, syncPendingCount, pendingEdits],
+  );
+
+  return { codeApply, codeDiscard };
+}

src/cli/ui/hooks/useInputRecall.ts

@@ -0,0 +1,41 @@
+import { useCallback, useRef } from "react";
+
+export interface UseInputRecallResult {
+  recallPrev: () => void;
+  recallNext: () => void;
+  pushHistory: (text: string) => void;
+  /** Reset cursor to the "fresh input" position — call after a successful submit. */
+  resetCursor: () => void;
+}
+
+/** Bash-style ↑/↓ recall over a turn-local prompt history. Cursor is `useRef` so toggles don't re-render. */
+export function useInputRecall(setInput: (s: string) => void): UseInputRecallResult {
+  const promptHistory = useRef<string[]>([]);
+  const historyCursor = useRef<number>(-1);
+
+  const recallPrev = useCallback(() => {
+    const hist = promptHistory.current;
+    if (hist.length === 0) return;
+    const nextCursor = Math.min(historyCursor.current + 1, hist.length - 1);
+    historyCursor.current = nextCursor;
+    setInput(hist[hist.length - 1 - nextCursor] ?? "");
+  }, [setInput]);
+
+  const recallNext = useCallback(() => {
+    if (historyCursor.current < 0) return;
+    const hist = promptHistory.current;
+    const nextCursor = historyCursor.current - 1;
+    historyCursor.current = nextCursor;
+    setInput(nextCursor < 0 ? "" : (hist[hist.length - 1 - nextCursor] ?? ""));
+  }, [setInput]);
+
+  const pushHistory = useCallback((text: string) => {
+    promptHistory.current.push(text);
+  }, []);
+
+  const resetCursor = useCallback(() => {
+    historyCursor.current = -1;
+  }, []);
+
+  return { recallPrev, recallNext, pushHistory, resetCursor };
+}

src/cli/ui/hooks/useLoopMode.ts

@@ -0,0 +1,143 @@
+import { type MutableRefObject, useCallback, useEffect, useRef, useState } from "react";
+import type { Scrollback } from "./useScrollback.js";
+
+interface ActiveLoop {
+  prompt: string;
+  intervalMs: number;
+  nextFireAt: number;
+  iter: number;
+}
+
+export interface ActiveLoopSnapshot {
+  prompt: string;
+  intervalMs: number;
+  iter: number;
+  nextFireMs: number;
+}
+
+export interface UseLoopModeResult {
+  startLoop: (intervalMs: number, prompt: string) => void;
+  stopLoop: () => void;
+  /** Snapshot for the `/loop` (no-arg) status branch. Returns null when no loop is active. */
+  getLoopStatus: () => ActiveLoopSnapshot | null;
+  /** Cheap predicate — used by handleSubmit's cancel-on-user-input guard and by apply-slash-result. */
+  isLoopActive: () => boolean;
+  /** True only during the timer's `handleSubmit` invocation — tells handleSubmit's cancel guard to skip itself so the loop's own re-submit doesn't kill the loop. */
+  isLoopFiring: () => boolean;
+  /** Reset by handleSubmit at the top of every call so the firing flag is one-shot. */
+  clearFiringFlag: () => void;
+  /** Reactive state for the LoopStatusRow render — null when no loop is active. */
+  activeLoop: ActiveLoop | null;
+}
+
+export interface UseLoopModeOptions {
+  log: Scrollback;
+  busyRef: MutableRefObject<boolean>;
+  /** Forward-ref to the latest `handleSubmit` — the closure shifts as state changes, so the timer dereferences fresh on each fire. */
+  handleSubmitRef: MutableRefObject<((raw: string) => Promise<void>) | null>;
+}
+
+/** Owns the active /loop config + its setTimeout-based scheduler. Re-issuing /loop replaces the slot; cancellation is centralized in stopLoop. */
+export function useLoopMode(opts: UseLoopModeOptions): UseLoopModeResult {
+  const { log, busyRef, handleSubmitRef } = opts;
+  const [activeLoop, setActiveLoop] = useState<ActiveLoop | null>(null);
+  const activeLoopRef = useRef<ActiveLoop | null>(null);
+  const loopTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
+  const loopFiringRef = useRef<boolean>(false);
+
+  useEffect(() => {
+    activeLoopRef.current = activeLoop;
+  }, [activeLoop]);
+
+  const stopLoop = useCallback(() => {
+    if (loopTimerRef.current) {
+      clearTimeout(loopTimerRef.current);
+      loopTimerRef.current = null;
+    }
+    const cur = activeLoopRef.current;
+    if (!cur) return;
+    setActiveLoop(null);
+    log.pushInfo(`▸ loop stopped (after ${cur.iter} iter${cur.iter === 1 ? "" : "s"}).`);
+  }, [log]);
+
+  const startLoop = useCallback((intervalMs: number, prompt: string) => {
+    if (loopTimerRef.current) {
+      clearTimeout(loopTimerRef.current);
+      loopTimerRef.current = null;
+    }
+    setActiveLoop({
+      prompt,
+      intervalMs,
+      nextFireAt: Date.now() + intervalMs,
+      iter: 0,
+    });
+  }, []);
+
+  const getLoopStatus = useCallback((): ActiveLoopSnapshot | null => {
+    const cur = activeLoopRef.current;
+    if (!cur) return null;
+    return {
+      prompt: cur.prompt,
+      intervalMs: cur.intervalMs,
+      iter: cur.iter,
+      nextFireMs: Math.max(0, cur.nextFireAt - Date.now()),
+    };
+  }, []);
+
+  const isLoopActive = useCallback(() => activeLoopRef.current !== null, []);
+  const isLoopFiring = useCallback(() => loopFiringRef.current, []);
+  const clearFiringFlag = useCallback(() => {
+    loopFiringRef.current = false;
+  }, []);
+
+  // /loop scheduler. Re-runs whenever activeLoop's `nextFireAt` shifts —
+  // either because startLoop set a fresh schedule or because a previous
+  // firing bumped the next-fire time. Cleanup clears the in-flight
+  // timer so a stopLoop / replacement doesn't leak a fire after cancel.
+  useEffect(() => {
+    if (!activeLoop) return;
+    const delay = Math.max(0, activeLoop.nextFireAt - Date.now());
+    const timer = setTimeout(async () => {
+      loopTimerRef.current = null;
+      // Skip the firing entirely when a prior turn is still running.
+      // Re-arm in 1s so the loop catches up the moment busy clears,
+      // rather than waiting a full interval after a slow turn.
+      if (busyRef.current) {
+        setActiveLoop((cur) => (cur ? { ...cur, nextFireAt: Date.now() + 1000 } : cur));
+        return;
+      }
+      const cur = activeLoopRef.current;
+      if (!cur) return;
+      const nextIter = cur.iter + 1;
+      // Schedule the NEXT firing now (independent of how long this turn
+      // takes). Keeps the cadence honest even when individual turns run
+      // long.
+      setActiveLoop((c) =>
+        c ? { ...c, iter: nextIter, nextFireAt: Date.now() + cur.intervalMs } : c,
+      );
+      log.pushInfo(`▸ /loop iter ${nextIter} → ${cur.prompt}`);
+      loopFiringRef.current = true;
+      try {
+        await handleSubmitRef.current?.(cur.prompt);
+      } catch {
+        // Persistent submission errors → kill the loop rather than spam
+        // the screen. User can re-issue /loop once they fix the cause.
+        stopLoop();
+      } finally {
+        loopFiringRef.current = false;
+      }
+    }, delay);
+    loopTimerRef.current = timer;
+    return () => clearTimeout(timer);
+  }, [activeLoop, stopLoop, log, busyRef, handleSubmitRef]);
+
+  return {
+    startLoop,
+    stopLoop,
+    getLoopStatus,
+    isLoopActive,
+    isLoopFiring,
+    clearFiringFlag,
+    activeLoop,
+  };
+}

src/cli/ui/hooks/useQuit.ts

@@ -0,0 +1,19 @@
+import type { WriteStream } from "node:fs";
+import { type MutableRefObject, useCallback, useEffect } from "react";
+
+/** Ctrl+C / SIGINT → flush transcript + `process.exit(0)`. We call `process.exit` directly rather than Ink's `exit()` because the singleton stdin reader keeps a `data` listener attached — `exit()` would unmount the React tree but leave the event loop alive and the terminal would hang. */
+export function useQuit(transcriptRef: MutableRefObject<WriteStream | null>): () => void {
+  const quitProcess = useCallback(() => {
+    transcriptRef.current?.end();
+    process.exit(0);
+  }, [transcriptRef]);
+
+  useEffect(() => {
+    process.on("SIGINT", quitProcess);
+    return () => {
+      process.off("SIGINT", quitProcess);
+    };
+  }, [quitProcess]);
+
+  return quitProcess;
+}

src/cli/ui/hooks/useScrollback.ts

@@ -38,6 +38,7 @@ export interface Scrollback {
     cost: number;
     sessionCost: number;
     balance?: number;
+    balanceCurrency?: string;
     elapsedMs?: number;
   }): string;
   showPlan(args: {
@@ -58,24 +59,11 @@ export interface Scrollback {
     topTools: ReadonlyArray<{ name: string; tokens: number; turn: number }>;
   }): string;
 
-  startBranch(total: number): string;
-  updateBranch(
-    id: string,
-    p: {
-      completed: number;
-      total: number;
-      latestIndex: number;
-      latestTemperature: number;
-      latestUncertainties: number;
-    },
-  ): void;
-  endBranch(id: string, aborted?: boolean): void;
-
-  startReasoning(): string;
+  startReasoning(model?: string): string;
   appendReasoning(id: string, chunk: string): void;
   endReasoning(id: string, paragraphs: number, tokens: number, aborted?: boolean): void;
 
-  startStreaming(): string;
+  startStreaming(model?: string): string;
   appendStreaming(id: string, chunk: string): void;
   endStreaming(id: string, aborted?: boolean): void;
 
@@ -219,6 +207,7 @@ export function useScrollback(): Scrollback {
           cost: args.cost,
           sessionCost: args.sessionCost,
           balance: args.balance,
+          balanceCurrency: args.balanceCurrency,
           elapsedMs: args.elapsedMs,
         });
         return id;
@@ -236,20 +225,9 @@ export function useScrollback(): Scrollback {
         dispatch({ type: "ctx.show", id, ...args, topTools: [...args.topTools] });
         return id;
       },
-      startBranch(total) {
-        const id = nextId("branch");
-        dispatch({ type: "branch.start", id, total });
-        return id;
-      },
-      updateBranch(id, p) {
-        dispatch({ type: "branch.progress", id, ...p });
-      },
-      endBranch(id, aborted) {
-        dispatch({ type: "branch.end", id, aborted });
-      },
-      startReasoning() {
+      startReasoning(model) {
         const id = nextId("r");
-        dispatch({ type: "reasoning.start", id });
+        dispatch({ type: "reasoning.start", id, ...(model ? { model } : {}) });
         return id;
       },
       appendReasoning(id, chunk) {
@@ -258,9 +236,9 @@ export function useScrollback(): Scrollback {
       endReasoning(id, paragraphs, tokens, aborted) {
         dispatch({ type: "reasoning.end", id, paragraphs, tokens, aborted });
       },
-      startStreaming() {
+      startStreaming(model) {
         const id = nextId("s");
-        dispatch({ type: "streaming.start", id });
+        dispatch({ type: "streaming.start", id, ...(model ? { model } : {}) });
         return id;
       },
       appendStreaming(id, chunk) {

src/cli/ui/hooks/useTranscriptWriter.ts

@@ -0,0 +1,20 @@
+import type { WriteStream } from "node:fs";
+import { type MutableRefObject, useCallback } from "react";
+import type { LoopEvent } from "../../../loop.js";
+import { recordFromLoopEvent, writeRecord } from "../../../transcript/log.js";
+
+/** Returns a `LoopEvent` writer that no-ops when no transcript was opened. Wraps `recordFromLoopEvent` + `writeRecord` so callers don't carry the model/prefix metadata. */
+export function useTranscriptWriter(
+  transcriptRef: MutableRefObject<WriteStream | null>,
+  model: string,
+  prefixHash: string,
+): (ev: LoopEvent) => void {
+  return useCallback(
+    (ev: LoopEvent) => {
+      const stream = transcriptRef.current;
+      if (!stream) return;
+      writeRecord(stream, recordFromLoopEvent(ev, { model, prefixHash }));
+    },
+    [transcriptRef, model, prefixHash],
+  );
+}

src/cli/ui/key-normalize.ts

@@ -27,7 +27,13 @@ const CSI_TAIL_TO_FLAGS: ReadonlyArray<{ tail: string; flags: CsiKeyFlags }> = [
   // Forward-delete (the key labelled Delete on most keyboards).
   { tail: "[3~", flags: { delete: true } },
   // Shift+Tab — terminal sends `\x1b[Z` rather than tab-with-shift.
+  // `[1;2Z` is the modifier-encoded variant some Windows PowerShell
+  // hosts emit; `[27;2;9~` and `[9;2u` cover modifyOtherKeys / Kitty
+  // forms. Issue #373.
   { tail: "[Z", flags: { shift: true, tab: true } },
+  { tail: "[1;2Z", flags: { shift: true, tab: true } },
+  { tail: "[27;2;9~", flags: { shift: true, tab: true } },
+  { tail: "[9;2u", flags: { shift: true, tab: true } },
 ];
 
 function alreadyStructured(flags: CsiKeyFlags): boolean {

src/cli/ui/layout/CardStream.tsx

@@ -1,48 +1,74 @@
-import { Static } from "ink";
+import { Box, type DOMElement, Text, useBoxMetrics } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React from "react";
+import React, { useEffect, useRef } from "react";
 import { CardRenderer } from "../cards/CardRenderer.js";
-import { ActiveCardContext } from "../primitives/Card.js";
 import type { Card } from "../state/cards.js";
 import { useAgentState } from "../state/provider.js";
+import { FG } from "../theme/tokens.js";
 
-/** Settled = no future event can mutate it; safe to commit via Ink's Static. */
-function isSettled(card: Card): boolean {
+/**
+ * Row-precision virtual scroll: outer Box clips with overflow="hidden",
+ * inner Box holds all cards and slides up via negative marginTop. Yoga
+ * computes inner height from card layout; useBoxMetrics reports it back
+ * so App can clamp scroll bounds. New cards arriving while the user is
+ * scrolled up sit below the visible window — bytes outside outer bounds
+ * aren't written, so terminal text selection survives streaming activity.
+ */
+export function CardStream({
+  scrollRows,
+  onMaxScrollChange,
+  suppressLive = false,
+}: {
+  scrollRows: number;
+  onMaxScrollChange: (rows: number) => void;
+  suppressLive?: boolean;
+}): React.ReactElement {
+  const cards = useAgentState((s) => s.cards);
+  // useBoxMetrics types its ref as RefObject<DOMElement> (non-null); the
+  // null! assertion satisfies the signature — Ink populates the ref on
+  // mount before the metrics hook reads it.
+  const outerRef = useRef<DOMElement>(null!);
+  const innerRef = useRef<DOMElement>(null!);
+  const outer = useBoxMetrics(outerRef);
+  const inner = useBoxMetrics(innerRef);
+  const maxScroll = Math.max(0, inner.height - outer.height);
+
+  useEffect(() => {
+    onMaxScrollChange(maxScroll);
+  }, [maxScroll, onMaxScrollChange]);
+
+  let visible = cards;
+  if (suppressLive && cards.length > 0 && !isFullySettled(cards[cards.length - 1]!)) {
+    visible = cards.slice(0, -1);
+  }
+
+  return (
+    <>
+      {scrollRows > 0 ? <Text color={FG.faint}>{" ↑ earlier — PgUp / wheel / ↑"}</Text> : null}
+      <Box ref={outerRef} flexDirection="column" flexGrow={1} overflow="hidden">
+        <Box ref={innerRef} flexDirection="column" marginTop={-scrollRows} flexShrink={0}>
+          {visible.map((card) => (
+            <CardRenderer key={card.id} card={card} />
+          ))}
+        </Box>
+      </Box>
+    </>
+  );
+}
+
+function isFullySettled(card: Card): boolean {
   switch (card.kind) {
     case "streaming":
     case "tool":
-    case "branch":
       return card.done || !!card.aborted;
     case "reasoning":
       return !card.streaming || !!card.aborted;
+    case "task":
+    case "subagent":
+      return card.status !== "running";
+    case "plan":
+      return card.steps.every((s) => s.status === "done" || s.status === "skipped");
     default:
       return true;
   }
 }
-
-export function CardStream(): React.ReactElement {
-  const cards = useAgentState((s) => s.cards);
-  const lastIdx = cards.length - 1;
-  const lastCard = lastIdx >= 0 ? (cards[lastIdx] as Card) : null;
-  const lastIsLive = !!lastCard && !isSettled(lastCard);
-  const committed: Card[] = lastIsLive ? cards.slice(0, lastIdx) : cards.slice();
-  const live: Card[] = lastIsLive && lastCard ? [lastCard] : [];
-  // Static items are emitted via bridge.emitStatic, which renders them in an
-  // off-tree React reconciler — context from the live tree does NOT propagate.
-  // The ActiveCardContext.Provider must therefore live inside the children
-  // function so it travels with the rendered subtree.
-  return (
-    <>
-      <Static items={committed}>
-        {(card) => (
-          <ActiveCardContext.Provider value={false} key={card.id}>
-            <CardRenderer card={card} />
-          </ActiveCardContext.Provider>
-        )}
-      </Static>
-      {live.map((card) => (
-        <CardRenderer key={card.id} card={card} />
-      ))}
-    </>
-  );
-}

src/cli/ui/layout/LiveExpandContext.ts

@@ -0,0 +1,4 @@
+import { createContext } from "react";
+
+/** ctrl-o toggles this; live streaming card swaps 4-line tail for full-tail view. */
+export const LiveExpandContext = createContext<boolean>(false);

src/cli/ui/layout/LiveRows.tsx

@@ -1,4 +1,4 @@
-import { Box, Text, useStdout } from "ink";
+import { Box, Text } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React as a runtime value (classic transform)
 import React from "react";
 import type { ApplyResult } from "../../../code/edit-blocks.js";
@@ -11,6 +11,7 @@ import { PILL_MODEL, PILL_SECTION, Pill, modelBadgeFor } from "../primitives/Pil
 import { Spinner } from "../primitives/Spinner.js";
 import { CARD, FG, TONE } from "../theme/tokens.js";
 import { useElapsedSeconds, useSlowTick, useTick } from "../ticker.js";
+import type { SubagentActivity } from "../useSubagent.js";
 
 export const SPINNER_FRAMES = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
 
@@ -79,17 +80,7 @@ export function ModeStatusBar({
 }
 
 function ModeBarFrame({ children }: { children: React.ReactNode }) {
-  const { stdout } = useStdout();
-  const cols = stdout?.columns ?? 80;
-  const ruleWidth = Math.max(20, cols - 2);
-  return (
-    <Box flexDirection="column">
-      <Box paddingX={1}>
-        <Text color={FG.faint}>{"╌".repeat(ruleWidth)}</Text>
-      </Box>
-      <Box paddingX={1}>{children}</Box>
-    </Box>
-  );
+  return <Box paddingX={1}>{children}</Box>;
 }
 
 function ModePill({
@@ -112,17 +103,20 @@ function ModePill({
 export function UndoBanner({
   banner,
 }: {
-  banner: { results: ApplyResult[]; expiresAt: number };
+  banner: { results: ApplyResult[]; expiresAt: number; pausedRemainingMs: number | null };
 }) {
   useTick();
   const totalMs = 5000;
-  const remainingMs = Math.max(0, banner.expiresAt - Date.now());
+  const paused = banner.pausedRemainingMs !== null;
+  const remainingMs = paused
+    ? (banner.pausedRemainingMs ?? 0)
+    : Math.max(0, banner.expiresAt - Date.now());
   const remainingSec = Math.ceil(remainingMs / 1000);
   const ok = banner.results.filter((r) => r.status === "applied" || r.status === "created").length;
   const total = banner.results.length;
-  const urgent = remainingSec <= 1;
+  const urgent = !paused && remainingSec <= 1;
   const pct = (remainingMs / totalMs) * 100;
-  const tone = urgent ? TONE.err : TONE.accent;
+  const tone = paused ? TONE.warn : urgent ? TONE.err : TONE.accent;
   return (
     <Box marginY={1} paddingX={1}>
       <Text backgroundColor={TONE.accent} color="black" bold>
@@ -132,11 +126,15 @@ export function UndoBanner({
       <Text backgroundColor={TONE.brand} color="black" bold>
         {" u "}
       </Text>
-      <Text color={FG.faint}>{" to undo  "}</Text>
+      <Text color={FG.faint}>{paused ? " to undo · " : " to undo · "}</Text>
+      <Text backgroundColor={paused ? TONE.warn : FG.faint} color="black" bold>
+        {" space "}
+      </Text>
+      <Text color={FG.faint}>{paused ? " to resume  " : " to pause  "}</Text>
       <CharBar pct={pct} width={20} color={tone} showLabel={false} />
       <Text color={FG.faint}>{"  "}</Text>
-      <Text color={tone} bold={urgent}>
-        {`${remainingSec}s`}
+      <Text color={tone} bold={urgent || paused}>
+        {paused ? `${remainingSec}s · paused` : `${remainingSec}s`}
       </Text>
     </Box>
   );
@@ -159,20 +157,8 @@ function subagentTitle(skillName: string | undefined, task: string): string {
   return `Sub-agent · ${short || "anonymous"}`;
 }
 
-/** Live block for a running subagent — Card-styled, fixed row count for flicker-safety. */
-export function SubagentRow({
-  activity,
-}: {
-  activity: {
-    task: string;
-    iter: number;
-    elapsedMs: number;
-    skillName?: string;
-    model?: string;
-    phase?: "exploring" | "summarising";
-    lastInner: { glyph: string; color: string; label: string; meta?: string } | null;
-  };
-}) {
+/** Live block for a single in-flight subagent — rich layout, used when only one is running. */
+export function SubagentRow({ activity }: { activity: SubagentActivity }) {
   useTick();
   const seconds = (activity.elapsedMs / 1000).toFixed(1);
   const phase = subagentPhaseLabel(activity.phase, activity.iter, activity.elapsedMs);
@@ -222,6 +208,78 @@ export function SubagentRow({
   );
 }
 
+/** 1 → rich; 2-max → compact rows; >max → compact + "+N more" fold. */
+export function SubagentLiveStack({
+  activities,
+  max = 3,
+}: {
+  activities: ReadonlyArray<SubagentActivity>;
+  max?: number;
+}) {
+  const tick = useTick();
+  if (activities.length === 0) return null;
+  if (activities.length === 1) return <SubagentRow activity={activities[0]!} />;
+  const visible = activities.slice(0, max);
+  const overflow = activities.length - visible.length;
+  const summarising = activities.filter((a) => a.phase === "summarising").length;
+  const metaParts = [`${activities.length} running`];
+  if (summarising > 0) metaParts.push(`${summarising} summarising`);
+  return (
+    <Card tone={CARD.subagent.color}>
+      <CardHeader
+        glyph="⌬"
+        tone={CARD.subagent.color}
+        title="subagents"
+        titleColor={PILL_SECTION.plan.fg}
+        titleBg={PILL_SECTION.plan.bg}
+        subtitle={metaParts.join(" · ")}
+        right={<Spinner kind="braille" color={CARD.subagent.color} />}
+      />
+      {visible.map((a, i) => (
+        <CompactSubagentLine key={a.runId} activity={a} tick={tick} index={i} />
+      ))}
+      {overflow > 0 ? <Text color={FG.faint}>{`  +${overflow} more running…`}</Text> : null}
+    </Card>
+  );
+}
+
+function CompactSubagentLine({
+  activity,
+  tick,
+  index,
+}: {
+  activity: SubagentActivity;
+  tick: number;
+  index: number;
+}) {
+  const summarising = activity.phase === "summarising";
+  const spinnerFrame = SPINNER_FRAMES[(tick + index) % SPINNER_FRAMES.length] ?? "·";
+  const glyph = summarising ? "▶" : spinnerFrame;
+  const glyphColor = summarising ? TONE.brand : CARD.subagent.color;
+  const seconds = (activity.elapsedMs / 1000).toFixed(1).padStart(5);
+  const title = activity.skillName ?? truncate(activity.task, 28);
+  const titlePadded = title.padEnd(28);
+  const last = activity.lastInner;
+  return (
+    <Box flexDirection="row">
+      <Text color={glyphColor} bold>
+        {`  ${glyph} `}
+      </Text>
+      <Text color={FG.body}>{titlePadded}</Text>
+      <Text color={FG.faint}>{`  iter ${String(activity.iter).padStart(2)} · ${seconds}s · `}</Text>
+      {last ? (
+        <>
+          <Text color={last.color}>{`${last.glyph} `}</Text>
+          <Text color={FG.body}>{truncate(last.label, 18)}</Text>
+          {last.meta ? <Text color={FG.faint}>{`  ${last.meta}`}</Text> : null}
+        </>
+      ) : (
+        <Text color={FG.faint}>queued…</Text>
+      )}
+    </Box>
+  );
+}
+
 function truncate(text: string, max: number): string {
   return text.length > max ? `${text.slice(0, max)}…` : text;
 }

src/cli/ui/layout/StatusRow.tsx

@@ -4,25 +4,17 @@ import React from "react";
 import { Countdown } from "../primitives/Countdown.js";
 import { useAgentState } from "../state/provider.js";
 import type { Mode, NetworkState, StatusBar } from "../state/state.js";
-import { FG, TONE, USD_TO_CNY } from "../theme/tokens.js";
+import { FG, TONE, formatCost } from "../theme/tokens.js";
 
 const RULE_PAD = 4;
 const RULE_MIN = 20;
 
-function balanceColor(cny: number): string {
-  if (cny < 5) return TONE.err;
-  if (cny < 20) return TONE.warn;
-  return TONE.brand;
-}
-
 export function StatusRow(): React.ReactElement {
   const status = useAgentState((s) => s.status);
   const session = useAgentState((s) => s.session);
   const { stdout } = useStdout();
   const cols = stdout?.columns ?? 80;
   const ruleWidth = Math.max(RULE_MIN, cols - RULE_PAD);
-  const turnCny = status.cost * USD_TO_CNY;
-  const sessionCny = status.sessionCost * USD_TO_CNY;
   const hasTurn = status.cost > 0;
 
   return (
@@ -48,16 +40,9 @@ export function StatusRow(): React.ReactElement {
             <Text bold color={TONE.brand}>
               {"▸ "}
             </Text>
-            <Text bold color={FG.body}>{`¥${turnCny.toFixed(4)} turn`}</Text>
-          </>
-        )}
-        <Sep />
-        <Text color={FG.sub}>{`¥${sessionCny.toFixed(3)} session`}</Text>
-        {status.balance !== undefined && (
-          <>
-            <Sep />
-            <Text color={FG.faint}>{"wallet "}</Text>
-            <Text bold color={balanceColor(status.balance)}>{`¥${status.balance.toFixed(2)}`}</Text>
+            <Text bold color={FG.body}>
+              {`${formatCost(status.cost, status.balanceCurrency)} turn`}
+            </Text>
           </>
         )}
         <Sep />

src/cli/ui/markdown-view.tsx

@@ -1,6 +1,6 @@
+import { Box, Text } from "ink";
 // biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
-import { inkCompat } from "../../renderer/index.js";
 import { type InlineSpan, type MdLine, markdownToLines } from "./markdown-lines.js";
 
 const FG_BODY = "#c9d1d9";
@@ -13,9 +13,6 @@ const TONE_WARN = "#f0b07d";
 const SURFACE_ELEV = "#161b22";
 
 export function MarkdownView({ text }: { text: string }): React.ReactElement {
-  // No useMemo: the renderer's Static-path uses a one-shot tree walker that
-  // doesn't set up a React dispatcher, so hook calls would throw. Parsing on
-  // every render is cheap (marked.lexer is allocation-light).
   return <MarkdownLines lines={markdownToLines(text)} />;
 }
 
@@ -25,35 +22,34 @@ export function MarkdownLines({
   lines: ReadonlyArray<MdLine>;
 }): React.ReactElement {
   return (
-    <inkCompat.Box flexDirection="column">
+    <Box flexDirection="column">
       {lines.map((line, i) => (
-        // biome-ignore lint/suspicious/noArrayIndexKey: lines are positional + stable per render
         <LineRow key={`md-${i}-${line.kind}`} line={line} />
       ))}
-    </inkCompat.Box>
+    </Box>
   );
 }
 
 function LineRow({ line }: { line: MdLine }): React.ReactElement | null {
   switch (line.kind) {
     case "blank":
-      return <inkCompat.Text> </inkCompat.Text>;
+      return <Text> </Text>;
     case "hr":
-      return <inkCompat.Text color={FG_FAINT}>──────</inkCompat.Text>;
+      return <Text color={FG_FAINT}>──────</Text>;
     case "heading":
       return (
-        <inkCompat.Box>
-          <inkCompat.Text bold color={FG_STRONG}>
+        <Box>
+          <Text bold color={FG_STRONG}>
             {`${"#".repeat(line.level)} `}
-          </inkCompat.Text>
+          </Text>
           <Spans spans={line.spans} bold strongColor />
-        </inkCompat.Box>
+        </Box>
       );
     case "paragraph":
       return (
-        <inkCompat.Box>
+        <Box>
           <Spans spans={line.spans} />
-        </inkCompat.Box>
+        </Box>
       );
     case "list": {
       const indent = " ".repeat(line.depth * 2);
@@ -68,20 +64,20 @@ function LineRow({ line }: { line: MdLine }): React.ReactElement | null {
       const markerColor =
         line.task === "done" ? TONE_OK : line.task === "todo" ? FG_FAINT : FG_META;
       return (
-        <inkCompat.Box>
-          <inkCompat.Text color={markerColor}>{`${indent}${marker} `}</inkCompat.Text>
+        <Box>
+          <Text color={markerColor}>{`${indent}${marker} `}</Text>
           <Spans spans={line.spans} dim={line.task === "done"} strike={line.task === "done"} />
-        </inkCompat.Box>
+        </Box>
       );
     }
     case "code":
       return <CodeBlock lang={line.lang} text={line.text} />;
     case "blockquote":
       return (
-        <inkCompat.Box>
-          <inkCompat.Text color={TONE_BRAND}>{"▎ "}</inkCompat.Text>
+        <Box>
+          <Text color={TONE_BRAND}>{"▎ "}</Text>
           <Spans spans={line.spans} italic />
-        </inkCompat.Box>
+        </Box>
       );
   }
 }
@@ -93,15 +89,15 @@ function spanKey(span: InlineSpan, i: number): string {
 function CodeBlock({ lang, text }: { lang: string; text: string }): React.ReactElement {
   const lines = text.split("\n");
   return (
-    <inkCompat.Box flexDirection="column">
-      {lang.length > 0 ? <inkCompat.Text color={FG_META}>{` ${lang}`}</inkCompat.Text> : null}
+    <Box flexDirection="column">
+      {lang.length > 0 ? <Text color={FG_META}>{` ${lang}`}</Text> : null}
       {lines.map((ln, i) => (
         // biome-ignore lint/suspicious/noArrayIndexKey: code lines are positional + stable per render
-        <inkCompat.Text key={`code-${i}`} backgroundColor={SURFACE_ELEV}>
+        <Text key={`code-${i}`} backgroundColor={SURFACE_ELEV}>
           {` ${ln} `}
-        </inkCompat.Text>
+        </Text>
       ))}
-    </inkCompat.Box>
+    </Box>
   );
 }
 
@@ -115,7 +111,7 @@ interface SpansProps {
 }
 
 function Spans({ spans, bold, italic, dim, strike, strongColor }: SpansProps): React.ReactElement {
-  if (spans.length === 0) return <inkCompat.Text> </inkCompat.Text>;
+  if (spans.length === 0) return <Text> </Text>;
   return (
     <>
       {spans.map((span, i) => (
@@ -150,9 +146,9 @@ function SpanText({
 }): React.ReactElement {
   if (span.code) {
     return (
-      <inkCompat.Text color={FG_STRONG} backgroundColor={SURFACE_ELEV}>
+      <Text color={FG_STRONG} backgroundColor={SURFACE_ELEV}>
         {` ${span.text} `}
-      </inkCompat.Text>
+      </Text>
     );
   }
   const color = span.fileRef
@@ -163,7 +159,7 @@ function SpanText({
         ? FG_STRONG
         : FG_BODY;
   return (
-    <inkCompat.Text
+    <Text
       color={color}
       bold={!!(span.bold || ambientBold)}
       italic={!!(span.italic || ambientItalic)}
@@ -172,6 +168,6 @@ function SpanText({
       underline={!!(span.link || span.fileRef)}
     >
       {span.text}
-    </inkCompat.Text>
+    </Text>
   );
 }

src/cli/ui/markdown.tsx

@@ -127,7 +127,6 @@ function ListItem({
             // biome-ignore lint/suspicious/noArrayIndexKey: list-item children are positional and stable per render
             return <List key={`l-${i}`} token={tok as Tokens.List} depth={depth + 1} />;
           }
-          // biome-ignore lint/suspicious/noArrayIndexKey: list-item children are positional and stable per render
           return <BlockToken key={`b-${i}-${tok.type}`} token={tok} />;
         })}
       </Box>
@@ -328,31 +327,29 @@ function FallbackTable({
 }): React.ReactElement {
   return (
     <Box flexDirection="column">
-      {headerCells.map((h, ci) => {
-        const label = `${padToCells(h, labelPad - 2)}: `;
-        const values = bodyCells.map((r) => r[ci] ?? "");
-        return (
-          // biome-ignore lint/suspicious/noArrayIndexKey: header cells positional
-          <Box key={`fk-${ci}`} flexDirection="column">
-            {values.map((v, ri) => {
-              const lines = wrapToCells(v, valueCells);
-              return lines.map((line, li) => (
-                // biome-ignore lint/suspicious/noArrayIndexKey: fallback table lines are positional
-                <Box key={`fv-${ri}-${li}`}>
-                  {li === 0 ? (
-                    <Text bold color={FG.sub}>
-                      {label}
-                    </Text>
-                  ) : (
-                    <Text>{padToCells("", labelPad)}</Text>
-                  )}
-                  <Text color={FG.body}>{line}</Text>
-                </Box>
-              ));
-            })}
-          </Box>
-        );
-      })}
+      {bodyCells.map((row, ri) => (
+        // biome-ignore lint/suspicious/noArrayIndexKey: body rows positional
+        <Box key={`fr-${ri}`} flexDirection="column">
+          {ri > 0 ? <Text> </Text> : null}
+          {headerCells.map((h, ci) => {
+            const label = `${padToCells(h, labelPad - 2)}: `;
+            const lines = wrapToCells(row[ci] ?? "", valueCells);
+            return lines.map((line, li) => (
+              // biome-ignore lint/suspicious/noArrayIndexKey: fallback table lines are positional
+              <Box key={`fc-${ri}-${ci}-${li}`}>
+                {li === 0 ? (
+                  <Text bold color={FG.sub}>
+                    {label}
+                  </Text>
+                ) : (
+                  <Text>{padToCells("", labelPad)}</Text>
+                )}
+                <Text color={FG.body}>{line}</Text>
+              </Box>
+            ));
+          })}
+        </Box>
+      ))}
     </Box>
   );
 }
@@ -370,12 +367,17 @@ function Inline({ tokens }: { tokens: Token[] }): React.ReactElement {
 const FILE_REF_RE = /\b([A-Za-z0-9_./@\-]+\.[A-Za-z0-9]{1,6})(?::(\d+)(?:-(\d+))?)?\b/g;
 const MENTION_RE = /(?<![A-Za-z0-9_])@([A-Za-z0-9_./\-]+\.[A-Za-z0-9]{1,6})/g;
 
-function osc8(label: string, target: string, color: string): React.ReactElement {
-  const open = `\x1b]8;;${target}\x07`;
-  const close = "\x1b]8;;\x07";
+function looksLikeFileRef(path: string, hasLine: boolean): boolean {
+  if (hasLine) return true;
+  if (path.includes("/")) return true;
+  const ext = path.split(".").pop() ?? "";
+  return ext.length >= 2;
+}
+
+function osc8(label: string, _target: string, color: string): React.ReactElement {
   return (
     <Text color={color} underline>
-      {`${open}${label}${close}`}
+      {label}
     </Text>
   );
 }
@@ -402,6 +404,7 @@ function renderInlineText(raw: string): React.ReactElement {
     if (hits.some((h) => start < h.end && end > h.start)) continue;
     const path = m[1]!;
     const line = m[2];
+    if (!looksLikeFileRef(path, line !== undefined)) continue;
     const target = line ? `file://${path}:${line}` : `file://${path}`;
     hits.push({ start, end, node: osc8(m[0], target, TONE.brand) });
   }

src/cli/ui/multiline-keys.ts

@@ -28,8 +28,10 @@ export interface MultilineAction {
   /** When `true`, fire `onSubmit(submitValue ?? value)`. */
   submit: boolean;
   submitValue?: string;
-  /** Set when ↑/↓ hits a buffer boundary — parent handles history recall. */
+  /** Set on Ctrl+P / Ctrl+N — parent handles prompt-history recall. */
   historyHandoff?: "prev" | "next";
+  /** Set when ↑/↓ fires on an empty buffer — parent scrolls the chat history. */
+  chatScrollHandoff?: "up" | "down";
   /** Reducer is pure — hands raw paste to PromptInput which allocates a sentinel and inserts that. */
   pasteRequest?: { content: string };
 }
@@ -68,12 +70,11 @@ export function processMultilineKey(
     return cursor === value.length ? NOOP : { next: null, cursor: value.length, submit: false };
   }
 
-  // History recall is bound to Ctrl+P / Ctrl+N (bash readline
-  // convention) rather than ↑/↓. Reason: Windows Terminal + ConPTY
-  // translates mouse wheel events to ↑/↓ keystrokes for raw-mode
-  // child processes — indistinguishable at the byte level. If ↑/↓
-  // recalled history, every wheel-scroll would corrupt the prompt
-  // with a prior message. Ctrl+P/N is wheel-immune.
+  // Prompt-history recall is on Ctrl+P / Ctrl+N (bash readline
+  // convention). ↑/↓ on an empty buffer scrolls chat history because
+  // Windows Terminal + ConPTY translates mouse wheel events to ↑/↓ in
+  // raw mode — the only way to make wheel-up scroll chat is to map ↑
+  // to chat-scroll. Ctrl+P/N is wheel-immune.
   if (key.ctrl && key.input === "p") {
     return { ...NOOP, historyHandoff: "prev" };
   }
@@ -90,12 +91,12 @@ export function processMultilineKey(
     return { next: null, cursor: Math.min(value.length, cursor + 1), submit: false };
   }
   if (key.upArrow) {
-    if (value.length === 0) return { ...NOOP, historyHandoff: "prev" };
+    if (value.length === 0) return { ...NOOP, chatScrollHandoff: "up" };
     const moved = moveCursorUp(value, cursor);
     return moved === cursor ? NOOP : { next: null, cursor: moved, submit: false };
   }
   if (key.downArrow) {
-    if (value.length === 0) return { ...NOOP, historyHandoff: "next" };
+    if (value.length === 0) return { ...NOOP, chatScrollHandoff: "down" };
     const moved = moveCursorDown(value, cursor);
     return moved === cursor ? NOOP : { next: null, cursor: moved, submit: false };
   }

src/cli/ui/presets.ts

@@ -1,48 +1,27 @@
-/** Preset table — branch and harvest stay off here so users never get ambushed into N× cost without opting in. */
-
 import type { PresetName } from "../../config.js";
 
 export interface PresetSettings {
   model: string;
   reasoningEffort: "high" | "max";
   autoEscalate: boolean;
-  /** Pillar-2 harvest. Always false in presets — opt-in via /harvest. */
-  harvest: boolean;
-  /** Branch budget. Always 1 in presets — opt-in via /branch. */
-  branch: number;
 }
 
 /** Old names `fast`/`smart`/`max` aliased via `resolvePreset` so legacy configs still load. */
 export const PRESETS: Record<"auto" | "flash" | "pro", PresetSettings> = {
-  // auto — flash baseline + auto-escalate to pro when the model emits
-  // <<<NEEDS_PRO>>> OR after 3+ tool failure signals in one turn.
-  // The default: cheap when easy, smart when hard.
   auto: {
     model: "deepseek-v4-flash",
     reasoningEffort: "max",
     autoEscalate: true,
-    harvest: false,
-    branch: 1,
   },
-  // flash — always flash, never escalate. `/pro` still arms a single
-  // manual turn; auto-promotion is the thing this disables. Use when
-  // you want predictable cost per turn.
   flash: {
     model: "deepseek-v4-flash",
     reasoningEffort: "max",
     autoEscalate: false,
-    harvest: false,
-    branch: 1,
   },
-  // pro — always pro. Hard pin; the model never downgrades. Use for
-  // multi-turn architecture work where flash is just going to keep
-  // escalating anyway and the back-and-forth wastes turns.
   pro: {
     model: "deepseek-v4-pro",
     reasoningEffort: "max",
     autoEscalate: false,
-    harvest: false,
-    branch: 1,
   },
 };
 

src/cli/ui/primitives/Card.tsx

@@ -1,8 +1,7 @@
 import { Box } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React, { useContext } from "react";
 
-/** Settled cards (in scrollback) drop the border + paddingX so history reads as flat lines. */
+/** Settled cards (in scrollback) drop border + padding + margin so history collapses to flat lines. */
 export const ActiveCardContext = React.createContext(true);
 
 export interface CardProps {
@@ -24,11 +23,7 @@ const STRIPE_BORDER = {
 export function Card({ tone, children }: CardProps): React.ReactElement {
   const active = useContext(ActiveCardContext);
   if (!active) {
-    return (
-      <Box flexDirection="column" marginTop={1}>
-        {children}
-      </Box>
-    );
+    return <Box flexDirection="column">{children}</Box>;
   }
   return (
     <Box

src/cli/ui/primitives/CardHeader.tsx

@@ -1,7 +1,7 @@
 import { Box, Text } from "ink";
-// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
-import React from "react";
+import React, { useContext } from "react";
 import { FG } from "../theme/tokens.js";
+import { ActiveCardContext } from "./Card.js";
 
 export type MetaItem = string | { text: string; color: string };
 
@@ -31,6 +31,9 @@ export function CardHeader({
   meta,
   right,
 }: CardHeaderProps): React.ReactElement {
+  const active = useContext(ActiveCardContext);
+  // Settled scrollback drops faint string meta + spinners; colored badges (rejected, retry) stay.
+  const visibleMeta = active ? meta : meta?.filter((item) => typeof item !== "string");
   return (
     <Box flexDirection="row" gap={1}>
       <Text color={tone}>{glyph}</Text>
@@ -44,7 +47,7 @@ export function CardHeader({
         </Text>
       )}
       {subtitle ? <Text color={FG.body}>{subtitle}</Text> : null}
-      {meta?.map((item, i) => {
+      {visibleMeta?.map((item, i) => {
         const isStr = typeof item === "string";
         const text = isStr ? item : item.text;
         const color = isStr ? FG.faint : item.color;
@@ -56,7 +59,7 @@ export function CardHeader({
           </React.Fragment>
         );
       })}
-      {right}
+      {active ? right : null}
     </Box>
   );
 }

src/cli/ui/primitives/Spinner.tsx

@@ -1,16 +1,13 @@
 import { Text } from "ink";
+// biome-ignore lint/style/useImportType: tsconfig jsx=react needs React in value scope for JSX compilation
 import React from "react";
+import { useTick } from "../ticker.js";
 
 const FRAMES = {
   circle: ["◐", "◓", "◑", "◒"] as const,
   braille: ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧"] as const,
 };
 
-const CADENCE_MS = {
-  circle: 200,
-  braille: 80,
-} as const;
-
 export interface SpinnerProps {
   kind?: keyof typeof FRAMES;
   color?: string;
@@ -19,12 +16,8 @@ export interface SpinnerProps {
 
 export function Spinner({ kind = "circle", color, bold }: SpinnerProps): React.ReactElement {
   const frames = FRAMES[kind];
-  const [frame, setFrame] = React.useState(0);
-
-  React.useEffect(() => {
-    const id = setInterval(() => setFrame((f) => (f + 1) % frames.length), CADENCE_MS[kind]);
-    return () => clearInterval(id);
-  }, [kind, frames.length]);
+  const tick = useTick();
+  const frame = tick % frames.length;
 
   return (
     <Text bold={bold} color={color}>