Commit Graph

1252 Commits

Author SHA1 Message Date
mateaix
5e1248432d fix(goal): reject trailing model evaluation documents 2026-09-14 01:12:26 +08:00
mateaix
b58fd93430 fix(goal): reject ambiguous model checklist results 2026-09-14 00:52:24 +08:00
mateaix
9f89db8262 fix(evidence): seal observation state and rows atomically 2026-09-14 00:46:10 +08:00
mateaix
6efae9e71c test(evaluation): reject duplicate keys in offline suites 2026-09-14 00:33:41 +08:00
mateaix
ebe4981476 test(evaluation): record compiled provenance in goal replay reports 2026-09-14 00:30:49 +08:00
mateaix
02f31fdaf6 fix(evidence): preserve failures across command observations 2026-09-14 00:24:52 +08:00
mateaix
ea1a28399c fix(files): authorize ownership before loading generated content 2026-09-14 00:14:34 +08:00
mateaix
7e45a26755 fix(files): isolate generated preview content with sandbox policy 2026-09-14 00:05:01 +08:00
mateaix
0498530f3a fix(goal): refresh progress when requirements are appended 2026-09-13 23:46:31 +08:00
mateaix
47066c6091 fix(files): reject symlinks when reloading generated downloads 2026-09-13 23:43:00 +08:00
mateaix
52a2843bdf fix(goal): sync completion memory after transaction commit 2026-09-13 23:39:45 +08:00
mateaix
542f075759 fix: avoid completion side effects on goal no-op transitions 2026-09-13 23:10:45 +08:00
mateaix
8cc7eaaae1 test: replay goal service boundaries against real H2 transactions 2026-09-13 23:04:31 +08:00
mateaix
96ed9cda27 fix: include configured goal success-check guidance 2026-09-13 22:47:29 +08:00
mateaix
0c045eaff8 fix: fence goal evaluations against definition revisions 2026-09-13 22:40:59 +08:00
mateaix
5c91dcc234 test: add replayable JSON artifact check scenarios 2026-09-13 22:20:49 +08:00
mateaix
fb7bb04f7a feat: add user-invoked bounded JSON artifact checks 2026-09-13 22:15:13 +08:00
mateaix
6a00c3869d fix: derive goal progress from the current merged checklist 2026-09-13 22:04:57 +08:00
mateaix
c024e2d9f0 fix: reject symlink artifacts and bound collection reads 2026-09-13 21:57:34 +08:00
mateaix
511b689387 fix(scheduling): cancel unstarted delayed jobs during shutdown 2026-09-13 21:39:40 +08:00
mateaix
3843dc04b7 feat(evidence): inspect artifact version changes on demand 2026-09-13 21:30:39 +08:00
mateaix
b04b0f673b fix(goal): preserve criteria initialized while bootstrap is in flight 2026-09-13 21:21:45 +08:00
mateaix
bb3d477aa2 test(evaluation): replay artifact version boundaries with real file IO 2026-09-13 21:16:18 +08:00
mateaix
501a1f8bdb fix(goal): recheck current criteria before automatic completion 2026-09-13 21:00:40 +08:00
mateaix
7d26eaf528 fix(artifacts): isolate registered file bytes from caller mutations 2026-09-13 20:53:41 +08:00
mateaix
35a241a62d test(evaluation): add offline goal task replay and baseline 2026-09-13 20:48:02 +08:00
mateaix
88cd78af5e fix(goal): reject checklist passes without evidence 2026-09-13 20:38:19 +08:00
mateaix
00d774fb85 test(webchat): align stream assertions with delta batching 2026-09-12 18:15:52 +08:00
mateaix
bfc5417508 feat(memory): harden recall and provider resilience 2026-09-12 13:03:39 +08:00
matevip
7e36ac7749 perf(execution): batch evidence attempt queries 2026-09-11 04:39:40 -04:00
mateaix
dfae7edbe0 fix(memory): isolate recall ledger by owner 2026-09-08 22:23:45 +08:00
matevip
7406fc99bf feat(execution): add persistent execution evidence in observe mode 2026-09-07 06:38:58 -04:00
matevip
26150b68e8 feat(team): add controlled worker intervention 2026-09-04 04:18:41 -04:00
matevip
bd41a71826 fix(cron): harden long-task reliability 2026-09-02 05:56:06 -04:00
matevip
b706f9a610 fix(memory): prevent transient constraints from becoming durable (#625) 2026-09-01 05:52:41 -04:00
matevip
d8976ddf1e docs(runtime): document restart-safe long tasks 2026-08-31 03:03:51 -04:00
matevip
dd4db18ed8 fix(runtime): harden long-task recovery and checkpoint writes 2026-08-30 23:13:42 -04:00
mateaix
97125b991e docs(delegation): clarify long-task acceptance 2026-08-30 14:47:40 +08:00
mateaix
701b99b3e9 fix(delegation): bound and cancel async agent tasks 2026-08-30 13:32:22 +08:00
mateaix
08a5bf6996 release: v2.2.0 2026-08-29 13:53:11 +08:00
matevip
7eb44731e7 fix(chat): stabilize long-task input recovery 2026-08-27 05:21:59 -04:00
matevip
01ed4a4fcd feat(goal): make long tasks recoverable and pace continuation dispatch 2026-08-27 03:49:15 -04:00
taobig
2090d09704 fix(goal): preserve continuation instructions and verified progress
Keep autonomous prompts separate from persisted user messages while retaining
conversation history and queued user attachment routing. Carry verified
checklist evidence across segments and evaluate only changed criteria.

Cancel model streams without interrupting checkpoint database writes. Fence
late worker admission during shutdown, persist accepted queued messages and
attachments, and leave interrupted execution leases recoverable.

Add regressions for prompt selection, cumulative evidence, cancellation I/O,
queued input durability and shutdown recovery. Verify 250 focused tests,
200 real General Assistant conversation rounds, 12 checkpoints across 13
autonomous segments, and pause/resume/disconnect/restart boundaries.
2026-08-26 06:17:05 -04:00
taobig
2fa2e60170 feat(goal): persist continuous execution across bounded turns
- Add a database-backed supervisor with durable scheduling, fenced leases,
  cooldowns, bounded worker concurrency and expired-lease restart recovery.
- Default new goals to persistent execution with zero meaning unlimited
  cumulative budget; preserve legacy goals and explicit positive limits.
- Yield bounded graph segments to the supervisor instead of ending unfinished
  goals at graph-local continuation limits. Require persisted checklist
  evidence before accepting completion, including concurrent criterion edits.
- Share conversation admission across interactive and background execution;
  preserve partial replies, usage and queued user input during interruption.
- Persist Stop and missing-input pauses, respect approval boundaries, and
  commit resume and approval-denial transitions with correct transactions.
- Retry identifiable transient failures with backoff; retain visible pauses
  for budget limits and errors that require review instead of replaying tools.
- Expose owner-authorized execution status and reconnectable scheduling events;
  add H2, MySQL and Kingbase migrations, API types and bilingual documentation.

Validation: 298 focused backend tests passed, including persistence, restart
scheduling, approval races, cancellation, admission and existing runtime tests.
Frontend type checking, bundled-doc parity and ID precision checks passed.
V188 is registered and all three dialects have unique migration versions;
the migration-map audit still reports 95 pre-existing missing registrations.

Scope: single-backend native runtime. Recovery checks existing state before
repeating effects; this does not promise exactly-once external tool execution.
2026-08-26 05:46:15 -04:00
matevip
fb94027038 fix(dsh): isolate runtime child environment 2026-08-26 05:13:55 -04:00
matevip
987bc2001a fix(agent): complete long-form responses reliably 2026-08-24 04:16:57 -04:00
matevip
e88be95cd2 fix: pass workspace context to tool guard (#617) 2026-08-24 02:43:57 -04:00
mateaix
dd7e561e48 fix(agent): harden tool completion and team retries (#606) 2026-08-23 21:43:35 +08:00
matevip
14df331ce6 fix: resolve workspace context for file mutations (#617) 2026-08-21 22:05:26 -04:00
matevip
8d32a60fdb fix(files): restrict generated downloads by workspace (#611) 2026-08-21 06:06:59 -04:00