dtoro 3c3b12df5e fix(agent): directive assent notes + retry bump to fix empty responses ...

The assent system note said 'Do not re-request or call run again for
these' — the LLM interpreted this as 'don't call run at all' and produced
empty responses (finish_reason=stop, content_len=0) until retries were
exhausted, leaving the session stuck in 'executing'.

Fix: rewrite both assent notes (pending-approval path and pure-plan-approval
path) to be directive about WHAT TO DO NEXT: call update_plan_step(running)
then run for each remaining step. The 'don't re-request' guidance is now
scoped to 'THOSE SPECIFIC' executions, not all run calls.

Also bump maxLLMRetries from 2 to 3 — the empty-response flake on complex
multi-turn flows benefits from one more retry.

2026-07-15 12:50:22 +02:00

..

eval

chore(eval): consolidate evals into single evals/ folder

2026-07-15 09:50:26 +02:00

agent.go

fix(agent): directive assent notes + retry bump to fix empty responses

2026-07-15 12:50:22 +02:00

assent_test.go

fix(agent): word-boundary matching + contracted negatives in chat-assent

2026-07-11 19:51:57 +02:00

assent.go

feat(web): split SPA from oikos binary, require auth on every route

2026-07-12 15:49:42 +02:00

continue_test.go

feat: event-driven auto-continuation — agent runs an approved plan to completion

2026-07-10 14:19:43 +02:00

continue.go

feat(agent): plan-first gate, iterative follow-ups, reasoning persistence

2026-07-15 09:36:27 +02:00

main.go

feat(agent): plan-first gate, iterative follow-ups, reasoning persistence

2026-07-15 09:36:27 +02:00

store_test.go

feat(agent): close knowledge loop — refuse complete_task without writeback (D.1+D.2)

2026-07-14 20:40:49 +02:00

store.go

fix(agent): move plan step replacement from reopenSession to setGoal

2026-07-15 12:32:02 +02:00

tasks.go

fix(agent): read-only plans execute without approval

2026-07-15 10:12:51 +02:00