fix(confidence-engine): preserve successful chat detection

This commit is contained in:
2026-09-05 19:41:01 +01:00
parent cd1c6f4fc5
commit 7070342fb1
3 changed files with 14 additions and 8 deletions
+6
View File
@@ -92,6 +92,12 @@
- Provider execution diagnostics now distinguish chat skipped due capability state, chat attempted and failed before generate fallback, and successful chat without fallback. They are deterministically verified through `/api/cases/start`; endpoint selection and fallback behaviour remain unchanged.
- Next restart point: one observation-only fixed-scenario production call to identify why `/api/generate` is reached.
## Chat capability body disposal
- A fresh-process debugger observation proved capability `/api/chat` returned HTTP 200, then invalid `res.body?.consume()` threw and the catch incorrectly cached `_chatSupported = false`.
- This skipped schema-constrained reconstruction `/api/chat` and used unconstrained `/api/generate`. Response-body disposal now uses Fetch-compatible consumption without changing capability or fallback semantics.
- Next boundary: one fresh-process fixed-scenario production observation.
## Current product architecture
Three distinct routes, not a single page: