Files
zero/internal/cli/plugin_activate.go
Anandan f401c66e81 polish: cut per-turn overhead, stabilize cache, add /new (#478)
* polish: cut per-turn overhead, stabilize cache, add /new

Context economy
- Compaction: preserveLast 8->6, triggerRatio 0.8->0.7 (both copies in sync).
- System prompt: kill the "ELABORATE summary" directive; scale summary to
  the work. Comment discipline moved into the core prompt for all model
  families (was Anthropic-only); Minimax/DeepSeek/Qwen/etc. now get it.
- Token-budget regression test ratchets the fixed per-turn overhead (base
  prompt + eager tool schemas) using the runtime's own estimator, so silent
  prompt/schema growth fails CI.

Response quality
- defaultResponseStyle balanced->concise. Reversible via /style.

Tool discovery
- <available_skills> block in the system prompt lists installed skills
  (name + one-line description) so the model loads the right one on the
  first call instead of guessing and reading the failure. Budget-capped at
  640 bytes with an "...and N more" fallback; renders nothing when no
  skills are installed (byte-identical to prior prompt).

Session lifecycle
- /new starts a fresh session in place; previous session stays on disk and
  is named in the confirmation for /resume. Guards against mid-run reset.
- /clear now notes that context persists and points to /new.

Performance & caching
- partitionTools no longer alpha-sorts the whole tools array every turn,
  which inserted a loaded deferred tool mid-array and shifted the cached
  prefix on every load. It now builds append-only regions
  [eager sorted] + [tool_search] + [loaded deferred sorted] so a
  mid-session load grows the tail instead of busting the eager prefix.
  Regression test asserts the eager block precedes any loaded tool.
- /context now shows cache hit rate (flags low rate over >5 turns as a
  possible unstable prefix) and rolling average turn latency with
  time-to-first-token. Both reset on /new.

Measured baselines (auto-mode, no workspace): base prompt 3162 tokens,
eager tool schemas 2433 across 12 tools. Budget ceilings set ~10% above.

Deliberately deferred (need a decision): inverted safety-policy rewrite
(sandbox doesn't catch ~half the B-tier rules), core-tool/web_fetch
deferral (activation-semantics blocker), aggressive prompt slim.

* fix(tui): clear staged attachments + queued message on /new

CodeRabbit: startNewSession reset conversation/usage/latency state but not
the staged input (pendingImages, pendingImageLabels, pendingDocuments,
queuedMessage), which is only consumed at prompt-submit. Without clearing it,
the fresh session's first prompt silently inherited the previous session's
images/documents/queued text. Test now stages all four and asserts they clear.
2026-07-04 20:56:52 +08:00

108 lines
4.2 KiB
Go

package cli
import (
"fmt"
"io"
"strings"
"github.com/Gitlawb/zero/internal/agent"
"github.com/Gitlawb/zero/internal/hooks"
"github.com/Gitlawb/zero/internal/plugins"
"github.com/Gitlawb/zero/internal/tools"
)
// pluginActivation holds what plugin activation contributed to a bootstrap so the
// later dispatcher + skill wiring can consume it: the plugin hook definitions and
// the plugin skill search roots. The zero value (no plugins) is inert — the
// dispatcher gets no extra hooks and the skill tool keeps only the default dir.
type pluginActivation struct {
hooks []hooks.Definition
skillRoots []string
}
// activatePlugins loads the workspace's plugins and makes their declared
// extensions live: plugin tools are registered into registry here, and the
// returned activation carries the plugin hooks + skill roots for the caller to
// fold into the hook dispatcher and the skill tool.
//
// It fails OPEN: any load error (or a malformed plugin) is surfaced as a warning
// on stderr and otherwise skipped, so a broken plugin can never wedge startup —
// mirroring how newHookDispatcher and skills.Load tolerate bad input.
func activatePlugins(workspaceRoot string, registry *tools.Registry, deps appDeps, stderr io.Writer) pluginActivation {
loaded, err := deps.loadPlugins(plugins.LoadOptions{Cwd: workspaceRoot})
if err != nil {
writePluginActivationWarning(stderr, "failed to load plugins: "+err.Error())
return pluginActivation{}
}
// Load fails OPEN per plugin: a malformed manifest is recorded as a diagnostic
// and the plugin is dropped rather than aborting the whole load. Surface those
// diagnostics so a skipped plugin is never silent.
for _, diagnostic := range loaded.Diagnostics {
writePluginActivationWarning(stderr, formatLoadDiagnostic(diagnostic))
}
result := plugins.Activate(registry, loaded.Plugins, plugins.ActivateOptions{Cwd: workspaceRoot})
for _, warning := range result.Warnings {
writePluginActivationWarning(stderr, warning)
}
// Re-register the skill tool to also resolve plugin-declared skills. The core
// skill tool only reads the default skills dir; the plugin-aware replacement
// merges the default dir with the plugin skill roots (default dir wins a name
// clash), so plugin skills appear in the agent's skill list. With no plugin
// skill roots this is byte-equivalent to the default skills surface.
if len(result.SkillRoots) > 0 {
registry.Register(plugins.NewSkillTool(deps.skillsDir(), result.SkillRoots))
}
return pluginActivation{hooks: result.Hooks, skillRoots: result.SkillRoots}
}
// skillInfos resolves the reusable skills the model can load via the skill tool —
// the default skills dir merged with any plugin skill roots, the same set the
// plugin-aware skill tool resolves against — as plain data for the agent's system
// prompt. It returns nil when no skills are installed, so a skill-less run leaves
// the prompt byte-identical.
func (a pluginActivation) skillInfos(defaultDir string) []agent.SkillInfo {
merged, _ := plugins.MergedSkills(defaultDir, a.skillRoots)
if len(merged) == 0 {
return nil
}
infos := make([]agent.SkillInfo, 0, len(merged))
for _, skill := range merged {
name := strings.TrimSpace(skill.Name)
if name == "" {
continue
}
infos = append(infos, agent.SkillInfo{Name: name, Description: strings.TrimSpace(skill.Description)})
}
return infos
}
// formatLoadDiagnostic renders a plugin load diagnostic into a single warning
// line, prefixing the diagnostic kind and appending the most specific locator
// available (manifest path, then plugin path, then plugin ID) so an operator can
// locate the offending plugin even when the manifest path is unknown.
func formatLoadDiagnostic(diagnostic plugins.Diagnostic) string {
message := fmt.Sprintf("[%s] %s", diagnostic.Kind, diagnostic.Message)
locator := diagnostic.ManifestPath
if locator == "" {
locator = diagnostic.PluginPath
}
if locator == "" {
locator = diagnostic.PluginID
}
if locator != "" {
message += " (" + locator + ")"
}
return message
}
func writePluginActivationWarning(stderr io.Writer, message string) {
if stderr == nil {
return
}
_, _ = fmt.Fprintf(stderr, "[zero] WARNING: plugin activation: %s\n", message)
}