kami 07819ec82d fix(inference): fold non-L0 SYSTEM entries into the leading system message ...

Strict chat templates (Qwen3.5) reject any system message after index 0.
Recalled L3 memory, L2 stage summaries, and retrieval entries carry
EntryRole.SYSTEM and rendered as standalone mid-conversation system turns,
making llama-server 400 on every router turn once L3 recall returned hits.

2026-06-11 21:55:28 +04:00

..

approvals

fix: harden event store, serialization, and grant engine; wire router chat

2026-05-28 22:39:15 +04:00

contracts

feat(kernel): retried stages receive explicit failure feedback; default retry backoff 1s

2026-06-10 21:53:45 +04:00

deterministic

fix(inference): fold non-L0 SYSTEM entries into the leading system message

2026-06-11 21:55:28 +04:00

fixtures

chore(clean up): remove unused imports, unnecessary object impls, concurrency/coroutine issues, trailing commas, etc.

2026-05-24 22:56:19 +04:00

integration

test(replay,integration): repo-knowledge replay derivation + state-keyed scan reuse

2026-06-11 13:52:06 +04:00

kernel

feat(kernel): RepoKnowledgeRetriever port + relevance-retrieved repo context with recency fallback

2026-06-11 13:39:57 +04:00

projections

feat(events,sessions): ProjectProfileBoundEvent reduced into session state

2026-06-10 21:47:30 +04:00

replay

test(replay,integration): repo-knowledge replay derivation + state-keyed scan reuse

2026-06-11 13:52:06 +04:00

transitions

fix: complete all P2 audit findings — dead code, NPE guard, hygiene

2026-05-29 01:01:45 +04:00

build.gradle

epic-12: after epic audit and init commit

2026-05-17 03:19:39 +04:00