Skip to content

v0.80.0

Choose a tag to compare

@github-actions github-actions released this 16 Oct 19:49
· 92 commits to main since this release
caeae1b

Changelog

Features

  • 464b645 feat(chatgptauth): device-code sign-in
  • de4686a feat(chatgptauth): token file, claims and status
  • 5a6e012 feat(chatgptauth): token refresh with in-process and file locks
  • bf461d0 feat(cli): --autopilot-model picks the model autopilot uses at gates
  • 5d2250c feat(doctor): report the ChatGPT subscription sign-in
  • a1feca3 feat(llm): usage-limit error, subscription usage label, keyless provider registration
  • 1d1533b feat(openai): ChatGPT-subscription mode for the Responses adapter
  • b214fbc feat(setup): sign in to a ChatGPT subscription for openai
  • 32fbd6f feat(trackerbot): refuse to start with TRACKER_OPENAI_AUTH=chatgpt
  • f0fe123 feat: add the TRACKER_OPENAI_AUTH switch (config tier)
  • 0e83b7f feat: route openai through the ChatGPT login when TRACKER_OPENAI_AUTH=chatgpt
  • 2a3977b feat: warn that dollar ceilings skip ChatGPT-subscription calls; estimate prices them at $0

Fixes

  • 6c38e39 fix(chatgptauth): finish a refresh the caller cancels mid-flight
  • ebd724d fix(chatgptauth): never follow redirects from the auth server
  • b8d2b06 fix(chatgptauth): refuse a sign-in whose id_token names no account
  • ae4a158 fix(chatgptauth): remove fixed sleeps from the cancel-flake test
  • 7d93599 fix(chatgptauth): stop Save from following a planted temp-file symlink
  • 163cf13 fix(doctor): keep openai out of the gateway notes in ChatGPT mode
  • b4433f6 fix(doctor): make ChatGPT-mode notes hints, not warnings
  • e0dc268 fix(llm): classify billing_error, merge native error shapes, surface unread compat objects (#685 review)
  • d1d9a84 fix(llm): classify stream errors by HTTP-status code and type (#685 review)
  • 0afa6ae fix(llm): decode numeric stream error codes (#684)
  • a05e286 fix(llm): exhaustion text pauses a retryable stream error; classify non-object compat chunks (#685 review)
  • db481a7 fix(llm): pause on exhaustion under any retryable stream code; ignore placeholder errors (#685 review)
  • 135ff4e fix(llm): retry stream errors whose code is unrecognized (#684)
  • 8995a7f fix(openai): harden ChatGPT-mode Complete, credential errors and redirects
  • 9ebc871 fix(openai): surface the ChatGPT backend's detail message on status errors
  • 6516d1c fix(resume): keep only live graph attributes through resume compaction (#703 review)
  • 43bd2d5 fix(resume): read the artifact-dir snapshot bounded and non-blocking (#706 review)
  • 2f84b66 fix(setup): open the OpenAI page on the API key field
  • cafeb25 fix(setup): show the sign-in age as "refreshed 2h ago"
  • 7427b6c fix(version): don't list openai in ChatGPT mode without a sign-in
  • a0c0522 fix: -r fails when the run is not in the artifact dir; hint repeats -w (#696 review)
  • 82ff350 fix: create TRACKER_RUN_DIR before the first node runs (#686)
  • 23d855f fix: fan-in recovery hint names the run ID after -r (#697 review)
  • aae12bc fix: resume compaction keeps parallel.results, inputs and graph attrs (#697)
  • 818fd18 fix: resume honors --artifact-dir and reads the secure checkpoint (#696)

Other

  • 3891da0 Merge branch 'feat/chatgpt-subscription': ChatGPT-subscription auth for the openai provider
  • 2c87531 Merge origin/main into fix/696-resume-artifact-dir
  • 4e0e69e Merge origin/main into fix/697-fanin-results-on-halt
  • c2d9e9d Merge origin/main into release/v0.80.0: fold in #697, #696, #686, #684
  • bbbf41d Merge pull request #685 from 2389-research/fix/684-stream-error-unknown-code
  • 1ff956d Merge pull request #694 from 2389-research/docs/post-v0.79.0-audit
  • d81b230 Merge pull request #695 from 2389-research/fix/build-product-dippin-tests
  • 840e235 Merge pull request #703 from 2389-research/fix/697-fanin-results-on-halt
  • 7723a46 Merge pull request #704 from 2389-research/fix/686-run-dir-before-first-node
  • caeae1b Merge pull request #705 from 2389-research/release/v0.80.0
  • 7313bfc Merge pull request #706 from 2389-research/fix/696-resume-artifact-dir
  • 9cac586 Merge remote-tracking branch 'origin/main' into fix/684-stream-error-unknown-code
  • 88de6a9 Merge remote-tracking branch 'origin/main' into fix/684-stream-error-unknown-code
  • 417bf07 Merge remote-tracking branch 'origin/main' into fix/684-stream-error-unknown-code
  • b90b0b0 Merge remote-tracking branch 'origin/main' into fix/684-stream-error-unknown-code
  • 6f329c1 Merge remote-tracking branch 'origin/main' into merge-686
  • efb3ef7 Merge remote-tracking branch 'origin/main' into merge-686
  • 58fb5a7 Merge remote-tracking branch 'origin/main' into merge-696
  • 8e048bf chore: apply the pre-commit hooks to every file
  • e82a3c3 chore: run goimports through go run at a pinned version
  • 0997b5a docs(gotchas): tests must start with TRACKER_OPENAI_AUTH unset
  • a78e426 docs(trackerbot): list TRACKER_OPENAI_AUTH on the website page
  • f7cc1f6 docs: CHANGELOG states the #686 run-dir create failure as a behavior change (#686 review)
  • 3b42d6c docs: audit README, website and architecture docs against v0.77.0–v0.79.0
  • 823c76a docs: changelog and gotchas for ChatGPT-subscription auth
  • 88573a4 docs: correct #686 CHANGELOG blast radius and finalizeCapture comment (#686 review)
  • 281eb00 docs: design ChatGPT-subscription auth for the openai provider
  • 1d045dc docs: explicit CheckpointDir resumes with the same path; legacy fallback (#696 review)
  • 477e8fe docs: fix CHANGELOG's autopilot-caveat sub-bullet to name the real levers
  • 007f62e docs: implementation plan for ChatGPT-subscription auth
  • 4ef3cdf docs: mark the ChatGPT spec implemented and fix stale comments
  • 89a7b04 docs: narrow simulate's validation claim; ACP request env is appended unfiltered (#694 review)
  • d8a8674 docs: plan Now — Task 1 complete
  • 209a4df docs: plan Now — Task 10 complete
  • 73d0023 docs: plan Now — Task 11 complete
  • 0545511 docs: plan Now — Task 12 complete
  • 4c9f71b docs: plan Now — Task 2 complete
  • c280325 docs: plan Now — Task 3 complete
  • 523b96c docs: plan Now — Task 4 complete
  • 47414e7 docs: plan Now — Task 5 complete
  • d880712 docs: plan Now — Task 6 complete
  • 820bad0 docs: plan Now — Task 7 complete
  • 39095fd docs: plan Now — Task 8 complete
  • e6b72d0 docs: plan Now — Task 9 complete
  • cfb3b83 docs: plan Now — Task 9 waiting on sign-in
  • 14ed27f docs: plan Now — all tasks complete; final review
  • 4599660 docs: plan Now — approved local merge into main
  • 34d3b1d docs: plan Now — done; record the rulings made during execution
  • 6689b73 docs: plan Now — releasing v0.80.0; record "tag it."
  • 2314062 docs: plan Now — shipped; follow-ups #699-#702; gotcha on device-code timing
  • b6bb2d2 docs: plan Now — v0.80.0 release prepared; record the release approvals
  • d785c75 docs: point compactResumeContext refs at engine_resume_compact.go (#697 review)
  • b07920f docs: reads: keys survive resume compaction only at summary:high/medium (#697 review)
  • 854aee0 docs: record plan approval; start Task 1
  • 0947c60 docs: record the unserved-model status (HTTP 400) from the live run
  • c6f364e docs: say ChatGPT mode drops four request settings
  • 4c61bbc refactor: move the XDG config-home resolver into internal/xdgconfig
  • 6d9b42c release: v0.80.0 — ChatGPT-subscription auth, --autopilot-model (#699)
  • 5a54b22 test(estimate): isolate TestExecuteEstimate_BuiltinWorkflow from the real config .env
  • 3cf71d8 test(llm): drop a stale comment contradicting the bare-quota rows (#685 review)
  • 567ec52 test(openai): live ChatGPT-subscription smoke test (opt-in)
  • 3f44f08 test(openai): opt-in unserved-model subtest; resolve ChatGPT spec items
  • 144be6e test: cover the resume hint's -w wiring from flags; scope the snapshot CHANGELOG entry (#696 review)
  • e1eb933 test: dippin scenario gate fails on an orphaned .test.json (#695 review)
  • 402f622 test: rewrite build_product's dippin scenarios; gate every .test.json
  • 3f21d5e test: start every test with TRACKER_OPENAI_AUTH unset
  • cd96799 test: subgraph #686 test exercises the child engine's run-dir creation (#686 review)