Commit Graph

316 Commits

Author SHA1 Message Date
Agent Workbench Dev 2fa220735d feat(runtime): default to passthrough and persist execution permissions 2026-11-06 04:58:57 +08:00
Agent Workbench Dev 67c034f4b6 fix(context): pair route capacities and preserve source evidence within bounded requests 2026-11-06 02:37:52 +08:00
Agent Workbench Dev 3e8c2f99fe test(release): preserve full-history diagnostics and recover the user-closed host baseline 2026-11-06 02:06:43 +08:00
Agent Workbench Dev d5de90930e chore(release): freeze 1.25.5 capacity and cache delivery checks 2026-11-06 01:27:39 +08:00
Agent Workbench Dev 72115647d9 fix(runtime): preserve model capacity history and cache evidence with resumable file proposals 2026-11-06 01:18:37 +08:00
Agent Workbench Dev 2b17821138 fix(models): accept streamed thinking signatures and publish valid plan tools 2026-11-06 00:34:04 +08:00
Agent Workbench Dev ada7d36453 docs: seal 1.25.4 output-truncation recovery QA and AC real-model sample 2026-11-05 09:46:58 +08:00
Agent Workbench Dev 2cb6c019fa test(live): add AC driver re-running the sealed AB task on 1.25.4 to observe truncation recovery 2026-11-05 09:41:16 +08:00
Agent Workbench Dev 9b064ce1b0 test(release): extend every 1.25.x verifier branch in verify-release to 1.25.4 2026-11-05 09:22:55 +08:00
Agent Workbench Dev 51402a1a19 test(release): add retained-state 1.25.3 to 1.25.4 formal upgrade driver 2026-11-05 09:12:51 +08:00
Agent Workbench Dev bea76a5f15 test(release): allow 1.25.4 in the isolated release verifier's controlled-Git branch 2026-11-05 09:04:37 +08:00
Agent Workbench Dev 74b0030888 chore(release): bump to 1.25.4 for output-truncation recovery
Pin package.json, package-lock.json and the literal APP_VERSION together,
add 1.25.4 to the controlled-release version allowlist, and record the
output-truncation recovery in the changelog.
2026-11-05 08:55:48 +08:00
Agent Workbench Dev 0fefa65eeb feat(runtime): recover a bounded number of output-ceiling truncations
When a turn stops at the output-token ceiling (output_truncated), the run no
longer dies. The truncated turn is still dropped whole - no tool is executed
and the invalid assistant message is never appended - but the model is given a
standalone instruction to split its proposals and the run continues. Recoveries
are capped per run (2); past that the original diagnostic still fails the run.
The connection is not cooled down for a truncation, since it responded fine and
only the model overran its budget, so the recovery turn can reach it. Both the
failed attempt and the recovering turn are accounted; neither is unknown.
Fixture regression covers recover-then-continue and the exhausted cap.
2026-11-05 05:56:10 +08:00
Agent Workbench Dev dfb6029f33 fix(models): give an actionable diagnostic when output hits the token ceiling
A response stopped at finish_reason=length (or Messages max_tokens /
Responses max_output_tokens) is still rejected whole - a complete leading
tool call is never applied because a later call may be cut mid-argument.
Previously the model only saw a generic incompleteness message and would
repeat the same oversized multi-tool turn. Now these paths raise a shared
output_truncated diagnostic telling the model to split proposals into one
file per response. Regression covers streaming/non-streaming Chat,
Responses and Messages, including the AB sample shape.
2026-11-05 04:47:42 +08:00
Agent Workbench Dev 7ab242dfec test: seal AB delivery failure and preserve validation boundaries 2026-11-05 03:35:06 +08:00
Agent Workbench Dev 658a5c4e37 test: freeze AB delivery with verified DeepSeek low reasoning 2026-11-05 03:23:44 +08:00
Agent Workbench Dev 6726ce1448 test: preserve AA live provider compatibility results 2026-11-05 03:21:49 +08:00
Agent Workbench Dev 4a7af1a6af test: freeze AA live compatibility and delivery validation 2026-11-05 03:15:01 +08:00
Agent Workbench Dev cef636fdbd chore(release): seal 1.25.3 model catalog delivery and QA evidence 2026-11-05 03:02:21 +08:00
Agent Workbench Dev 1bd94856d3 fix(qa): accept 1.25.3 in contained engine release verification 2026-11-05 02:12:05 +08:00
Agent Workbench Dev 6741510b2d fix(release): pair retained upgrades with the current package version 2026-11-05 01:23:34 +08:00
Agent Workbench Dev bb4b497713 chore(release): include model catalog in 1.25.3 acceptance 2026-11-05 00:45:10 +08:00
Agent Workbench Dev d038191832 feat(models): discover upstream catalogs and select shared-source model pools 2026-11-05 00:44:35 +08:00
Agent Workbench Dev 109e51247a chore(release): freeze 1.25.3 reasoning acceptance and retained upgrade 2026-11-05 00:16:46 +08:00
Agent Workbench Dev f1242ee800 feat(models): pair reasoning controls with provider profiles and tool history 2026-11-05 00:13:53 +08:00
Agent Workbench Dev 3ab3a6e46a fix(models): reject empty Messages responses with retained evidence 2026-11-04 23:53:49 +08:00
Agent Workbench Dev b74e5edcd6 docs(qa): seal Claude empty engineering response and retained state 2026-11-04 23:44:03 +08:00
Agent Workbench Dev 23ff3c388b test(live): freeze explicitly requested Claude retry at revision 3 2026-11-04 23:39:12 +08:00
Agent Workbench Dev 31f2d53292 docs(qa): retain corrected Claude request and GPT identity response 2026-11-04 23:37:22 +08:00
Agent Workbench Dev c00c57e58f test(live): freeze user-corrected Claude route as a new sample 2026-11-04 23:34:38 +08:00
Agent Workbench Dev 324ec46623 docs(qa): retain Claude identity typo and DeepSeek reasoning cap failure 2026-11-04 23:34:08 +08:00
Agent Workbench Dev 409d30f6e2 test(live): freeze separate Claude and DeepSeek route samples 2026-11-04 23:24:51 +08:00
Agent Workbench Dev b9f4164909 docs(qa): seal bounded Y outcome and reconcile retained state 2026-11-04 23:22:40 +08:00
Agent Workbench Dev af89f019dd chore(qa): freeze independent apikey sample Y after X connection guard 2026-11-04 23:04:19 +08:00
Agent Workbench Dev 0667458321 docs(release): seal verified 1.25.2 retained upgrade and original failures 2026-11-04 22:56:21 +08:00
Agent Workbench Dev d510cb7049 test(network): synchronize TLS timeout assertions with real handshakes 2026-11-04 21:39:52 +08:00
Agent Workbench Dev 4a49f02daf fix(context): size sealed browser-result reads without mutation reserve 2026-11-04 20:29:02 +08:00
Agent Workbench Dev a3210fa559 fix(qa): freeze live sample against the installed source commit 2026-11-04 20:21:33 +08:00
Agent Workbench Dev 401a39a824 fix(qa): inspect the refreshed HTML preview document 2026-11-04 20:13:54 +08:00
Agent Workbench Dev b4e1569cb8 fix(qa): require 1.25.2 candidate before formal upgrade 2026-11-04 20:03:27 +08:00
Agent Workbench Dev e3f2d8e79a chore(release): prepare 1.25.2 validation and fresh bounded sample X 2026-11-04 19:57:24 +08:00
Agent Workbench Dev 3647296a8d perf(reading): share bounded file and log budgets across independent reads 2026-11-04 19:53:13 +08:00
Agent Workbench Dev d2703ad340 fix(reading): preserve mixed-turn progress without disabling no-op guards 2026-11-04 19:43:48 +08:00
Agent Workbench Dev b4d7d2c965 fix(tasks): honor explicit prohibitions on running engineering checks 2026-11-04 19:43:47 +08:00
Agent Workbench Dev 1037031e82 fix(install): recheck file ownership after the manifest scan 2026-11-04 19:40:33 +08:00
Agent Workbench Dev 13c8fc2d79 fix(upgrade): accept the served current patch in the compatibility gate 2026-11-04 19:40:32 +08:00
Agent Workbench Dev dcea76b732 fix(budget): verify reservation history in ledger order and preserve output bounds 2026-11-04 19:34:36 +08:00
Agent Workbench Dev 78abbc3eb9 docs: plan reliability fixes and bounded delivery verification for 1.25.2 2026-11-04 19:29:37 +08:00
Agent Workbench Dev 532b9d605a docs(agents): record 1.25.1 delivery and sample W as current handoff state 2026-11-04 19:09:59 +08:00
Agent Workbench Dev 62961cf128 docs(qa): record 1.25.1 formal upgrade and bounded real sample W results
Formal 1.25.1 upgrade retained PID 36552/Provider/key and all data exact; measured update 156369ms vs 1.25.0 774098ms (~5x faster). The first upgrade attempt was safely rolled back by the HOST_VERSION_MISMATCH guard before the APP_VERSION constant fix. Sample W network diagnostic and single identity preflight passed; the bounded Node task stopped at no_progress after 6 reading-only requests without proposing the allowed files, so it did not close the loop and did not reach R's reservation wall. v2 benefit remains proven at unit level with R's real receipts; no Purge AB/BA run per plan.

Docs only; evidence stays under gitignored .qa.
2026-11-04 19:06:51 +08:00