Heretek-AI/GDevelop-Heretek
1
0
Fork 0
mirror of https://github.com/Heretek-AI/GDevelop-Heretek.git synced 2026-10-12 02:02:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore(autonomous): refresh the run log through cycle 192

Browse Source
This commit is contained in:
John Smith
2026-09-24 01:52:26 -04:00
parent 05c108e671
commit caff144931
1 changed files with 15 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
runs/RUNS.md
+15 -12
View File
@@ -34,27 +34,30 @@ work; `endpoint` is the AI endpoint under test for that cycle.
| 181 | Cover the security-relevant role resolution and role predicates (Tier 1) | `test(ai): cover role resolution and role predicates` | green (gates) |
| 182 | Cover spawnSubAgent, the untested delegation glue (Tier 1) | `test(ai): cover spawnSubAgent delegation glue` | green (gates) |
| 183 | Execute the Phase 5 WebUI audit (dev server + Chrome DevTools MCP) | `chore(autonomous): execute the Phase 5 WebUI audit` | green (WebUI-audited) |
| 184 | WebUI-verify the studio transcript rendering (reasoning + tool flow) | `chore(autonomous): verify the studio transcript rendering` | green (gates) |
| 184 | WebUI-verify the studio transcript rendering (reasoning + tool flow) | `chore(autonomous): verify the studio transcript rendering` | green (WebUI-audited) |
| 185 | Harden the function-call-output builder (null entry + non-object spread) | `fix(ai): harden the function-call-output builder` | green (gates) |
| 186 | Cover the prompt builder and per-chat config resolver (Tier 1) | `test(ai): cover the prompt builder and config resolver` | green (gates) |
| 187 | Unit-test the preferences backfill by extracting it to the tested leaf | `refactor(preferences): extract the stored-preferences merge to the tested leaf` | green (gates) |
| 188 | Sanitize plan tasks at the getLatestActivePlan chokepoint | `fix(ai): sanitize plan tasks at the getLatestActivePlan chokepoint` | green (gates) |
| 189 | Live Phase 5 harness run vs local Ollama; fix the CORS preflight header | `fix(ai): stop sending attribution headers that break local CORS` | green (live local Ollama: preflight 204 + POST 200, real reply) |
| 190 | Bounded live studio-run attempt (plan + sub-agent) via the WebUI | `chore(autonomous): record the bounded live studio-run attempt` | blocked (UI drawer flaky for new chat) |
| 191 | Stream reasoning as progress so the cold-start hint clears (Tier 2) | `fix(ai): stream reasoning as progress before the first answer token` | green (gates; live endpoint verified) |
| 192 | Make provider telemetry inspectable per request (Tier 2 audit data layer) | `feat(ai): expose per-request provider telemetry` | green (gates; live endpoint verified) |
## Session context (cycles 155-188)
## Session context (cycles 155-192)
- **model**: `deepseek-v4.1-flash:cloud` (Ollama provider).
- **endpoint**: `llm.heretek.one` reachable but answers 401 without a key; no
`config.local.json`, so a live model turn is blocked on credentials.
- **endpoint**: local Ollama `:11434` is live and works (cycle 189 fixed a CORS
preflight bug that blocked every local request); hosted `llm.heretek.one` answers
401 without `config.local.json`.
- **turns per phase**: Phase 1 ingest, Phase 2 scoring, Phase 3 implementation,
Phase 4 gates, Phase 6 consolidation; Phase 5 executed cycle 183.
- **token efficiency**: fork AI suite 1489 -> 1627 passing tests across 81 suites.
- **evidence**: cycle 183/184 loaded the editor via the dev server + Chrome DevTools MCP,
opened the Ask AI drawer, and read the persisted HarborTown run (114 messages, 35
assistant turns, 146 tool calls, 12 sub-agent spawns, status ready, no error); console
had only upstream MUI deprecation warnings; no 4xx/5xx requests.
Phase 4 gates, Phase 6 consolidation; Phase 5 executed cycles 183/184/189.
- **token efficiency**: fork AI suite 1489 -> 1630 passing tests across 81 suites.
- **evidence**: cycle 189 live BYOK round-trip (real model reply) after the CORS fix;
cycle 183/184 WebUI audit of the persisted HarborTown run (12 sub-agent spawns).
- **gates every cycle**: `npm test` (react-app-rewired), `eslint --max-warnings=0`,
`prettier --list-different`, `check-fork-divergence.js`; `npm run build` twice,
`npm run flow` on Flow-relevant changes, dev-server compile live.
`prettier --list-different`, `check-fork-divergence.js`; `npm run build` x2;
`npm run flow` on Flow-relevant changes; dev-server compile live.
## Gate definitions