Commit Graph

14 Commits

Author SHA1 Message Date
matevip
1affbd7b82 feat(agent): loop-engineering robustness — goal continuation, plan re-plan, stall detection
- goal: continue (not skip) on max-iterations and evidence-insufficient turns.
  A max-iterations turn grants a fresh iteration budget ("hard continuation"),
  bounded per run and sized into the graph recursion ceiling, so a task too big
  for one budget keeps going instead of stalling until the next user message.
- plan-execute: re-plan the remaining work on a step exception, and on a
  signature-based stall (repeated failures / identical results / no usable
  result) instead of advancing dependent steps with junk; bounded by a per-run
  re-plan cap, with a graduated change-strategy nudge before the hard stop.
- plan-execute: auto-derive a goal from a genuine multi-step plan, seeding the
  acceptance criteria from the plan steps, so the goal subsystem engages without
  the model calling setGoal; broadcast goal_created so the UI hydrates.
- react: refund the iteration for setup-only rounds (load_skill / enable_tool)
  so a tight budget is not eaten by the load-then-use two-step.
- ui: re-fetch the active goal when a turn finishes so a goal created or mutated
  mid-conversation surfaces without depending on an SSE event.
- streaming: make retry backoff / total-time budget instance fields with a
  test-only seam; clarify that the wall-clock budget (not max-retries) bounds a
  sustained SERVER_ERROR loop to ~8 attempts, fixing the slow/flaky retry test.
2026-06-17 06:37:08 +08:00
matevip
3883b68cca fix(goal): drop score gate, bill failed evaluator calls, fix Evaluator SPI + tool prompt + bootstrap cap 2026-06-03 21:19:27 +08:00
matevip
77196acbcd test(goal): checklist codec + service coverage; final internal-ref sweep 2026-06-03 21:19:27 +08:00
matevip
c5ccce8ebc feat(goal): deterministic completion + remaining-criteria followup + auto-followup gates 2026-06-03 21:19:26 +08:00
matevip
0fc8579a3e feat(goal): persist checklist end-to-end + GoalResponse wire DTO 2026-06-03 21:19:26 +08:00
matevip
b65887f93d feat(goal): dual-mode checklist evaluator with structured output + Evaluator SPI 2026-06-03 21:19:26 +08:00
matevip
cd1c66fc0e feat(goal): structured checklist data model — criteria column, criterion records, dual-carrier evaluation result 2026-06-03 21:19:26 +08:00
matevip
a9c2d45790 Harden goal approval and workspace flows 2026-05-23 22:55:16 +08:00
matevip
66af70388b fix(goal): give reasoning models enough budget for the evaluator JSON 2026-05-21 22:27:43 +08:00
matevip
9e93c52d9a fix(goal): real evaluator, retry refactor, hardened node + extra edges 2026-05-21 22:27:00 +08:00
matevip
efce3bc209 fix(goal): pin JSON wire form for GoalStatus to lowercase 2026-05-21 14:43:40 +08:00
matevip
746ade0410 feat(goal): flip enabled flag on, forward completions to long-term memory 2026-05-21 14:43:34 +08:00
matevip
ce74a0ae48 feat(goal): graph topology + evaluation node wired into ReAct + Plan-Execute 2026-05-21 14:43:13 +08:00
matevip
ac6c8a18e3 feat(goal): persistent cross-turn goal with self-evaluation scaffolding 2026-05-21 14:43:01 +08:00