mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-12 06:10:44 +01:00
f0ddccff71f3bbda975cac8a86053a67c5d554f3
7726
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Dotta
|
5dee4993f1 |
fix(server): distinguish aborted HTTP responses in access logs (#18781)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Operators use HTTP logs to investigate task updates and database contention. > - Node starts each response with status 200 before it sends any headers. > - The custom HTTP formatter printed that status when the client disconnected before a response. > - This made an aborted request look like a successful task update. > - This pull request labels incomplete responses as aborted and preserves error severity. > - Operators can distinguish HTTP delivery from database commit evidence. ## Linked Issues or Issue Description - Refs: #18753. - Related: #14369 removes writes from legacy comment attribution reads. This change addresses a separate logging defect. The missing-write incident in #18753 remains unconfirmed. We reproduced a misleading `PATCH /issues/:id 200` log when the client received no HTTP response. We did not reproduce an HTTP 200 delivered before the mutation committed. ## What Changed - Label premature connection closure as `request aborted before response` or `request aborted during response` in both success and error log paths. - Log aborts as warnings, including on routes whose successful polling logs are normally silent. Keep explicit errors and 5xx responses at error severity. - Keep private webhook paths and query values out of aborted-request logs. - Add real HTTP disconnect coverage and real PostgreSQL task-update tests with synthetic data. - Document that HTTP completion logs do not establish database commit or rollback. ## Verification - Baseline: the new HTTP disconnect test failed because the log said `PATCH /api/issues/synthetic-task 200` at info level, while the client received no response. - `node node_modules/vitest/vitest.mjs run server/src/__tests__/http-log-aborted-response.test.ts server/src/__tests__/http-log-redaction.test.ts server/src/__tests__/issue-update-lock-contention-postgres.test.ts` passed: 81 tests in 3 files. - `pnpm -r typecheck` and `pnpm build` passed. The build used a second clean clone at the exact PR commit. - GitHub CI completed at `e3e96d836119df5fa1c7f22615484265de5aa700`: 52 checks passed and two Storybook checks were skipped. Greptile returned 5/5 with no findings. - Local `pnpm test:run` passed all six chat shards (1,834 tests). Its general server stage then reported four failures. We stopped the run after preserving these results and comparing the failing cases. The later workspace and serialized stages were not completed locally. The full local suite is not green. - The four observed failures were the two OAuth concurrency cases in `tool-access-service.test.ts`, the bounded workspace conflict case in `workspace-runtime.test.ts`, and the `enableNativeRunner`-disabled stop-control case in `muse-control.test.ts`. - Comparisons used the same sanitized environment on unchanged master `2b14ec03a48f17606db53fd7fc0edc53bc930409` and this PR. Both OAuth cases pass in isolation on both revisions. The full tool-access file passes all 498 tests on the PR; master has a different 15-second timeout, with 497 tests passing. The workspace case passes on the PR but fails port-reservation setup on master. The Muse case passes on both. These results do not explain or clear the original broad-suite OAuth and Muse failures. - Tested on macOS arm64, Node.js 26.4.0, pinned pnpm 9.15.4, and disposable embedded PostgreSQL. The API fixture uses a synthetic process-adapter agent. No model provider runs and no production database or real credentials participate. - The PostgreSQL fixture verifies an actual lock waiter with `pg_blocking_pids`. A blocked issue update does not return 200 before release. After release, status, monitor policy, inline comment, and audit receipt are persisted before the completed response. - Holding the company row with `FOR NO KEY UPDATE` does not block these fixture writes. They return 200 with persisted data while the compatible parent-row lock remains held. - An injected database trigger failure returns 500 and leaves the issue, comment, and mutation audit unchanged. Two influence-attempt receipts remain under the existing documented cap policy. - The combined lock/disconnect/failure test verifies no delivered response, an aborted log, and rollback after an injected database failure. The injected failure is not evidence of the reporter's production root cause. - Source inspection at release `v2026.1023.0` (`1b4c803816e82b0265c8422e1a804327a7ed8c54`) finds the same unconditional success formatter and awaited issue-update transaction. That release was not runtime-tested. ## Risks - Log consumers that parse the literal method/path/status message must handle the aborted-request message. - Client disconnects during streaming or SSE responses also use the aborted warning. Expected disconnects may therefore produce more warnings. - A disconnected request may still commit after the client leaves. This change reports transport completion and does not cancel or retry mutations. - This PR does not close #18753. The original missing-write cause and a received-200-without-commit scenario remain unconfirmed. - Local full-suite failures remain unresolved despite green CI. Keep this PR in draft until that verification gap is accepted or resolved. > This is a targeted bug fix. ROADMAP.md does not list overlapping feature work. ## Model Used - OpenAI GPT-6 through Codex, with reasoning, tool use, code execution, and parallel subagent investigation. - The exact runtime model ID and context window size were not exposed to the agent. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [ ] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge |
||
|
DottaandPaperclip
|
c41869f6df |
test(activity): retain equivalent actor-owned repair descriptions
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
8b8f676e83 |
fix(activity): guard timeline inference with scoped ownership history
Use one ownership projection for window-independent history presence and recorded events. Preserve ordered narrative facts across punctuation. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
27d35314be |
feat: connect experimental Cursor Cloud to native Runner (#18069)
## Thinking Path > - Paperclip helps people manage AI agents at work. > - Native Runner owns execution and task authority. > - Operators need an explicit native Cursor Cloud choice. > - Tools/files need company, task, session and run scope. > - This connects setup and lifecycle to Runner Core. > - Legacy agents retain their selected runtime/history. ## Linked Issues or Issue Description Depends on #18068. Managed-only API placement and API-owned Stop containment are in #18310. Refs #5664, #11803, #17299, and #17471. No duplicate native integration was found. ## What Changed - Add explicit native setup, Test, saved settings and dispatch with no-prompt account/catalog/callback checks. - Scope MCP to company, agent, task, session and run through authenticated HTTPS. Register verified artifacts/work products; preserve unpriced accounting, completion and budgets. - Connect Question/Plan continuation, cold handoff and admitted fresh sessions; recheck waiting admission under the task lock. - Present audited Stop intent while preserving physical containment proof; bound typed Stop metadata. Preserve title edits through preference/navigation readiness. - Select one live/history representation per run to prevent duplicate settled cards. Integrate upstream task/read-state/retry/schema contracts and repair two test fixtures. ## Verification Core #18068 merged normally as `3910352c923a`. Product head `5a60f541c982` has passing repository CI/build/browser checks, including required `ci / verify` and `ci / e2e`, plus a fresh exact-head Greptile 5/5 review with zero unresolved threads. Two non-required external app checks remain pending: Paperclip Review is queued waiting for its assigned agent; Superagent Security Scan is reviewing the diff. These are not claimed completed. The normal merge endpoint enforces the repository rules; no policy or status override is used. The dependent Top PR is addressing its fresh deferred-Stop review finding. Product `5a60f541c982` passed 270 UI and 20 selected Cloud/native/disk-full tests, UI/server typechecks, tokens and maps. Top’s clean normal merge passed tokens/maps/diff checks; its matrix is pending. Earlier receipts retain Core’s 47 Rust, 4 compiled-native retirement and 19 settlement controls, normal Runner build, and runtime 9af build/recursive typecheck. Reproduced loading/shutdown CI fixture failures passed 7/7 browser and 121/121 unit tests, including a broken-unsubscribe negative control. No timeout/oracle was relaxed. The actual-parent restack adopts 75 mainline paths, including schema/permission and generic runtime changes, beyond the prior two fixture repairs and client projection fix. The full difference from hosted `9af1a5de42a2` is therefore larger than four files; backend source also differs. The merged config-read checks match Core391, and the Cursor placement/runtime/relay/selection modules match frozen9af; selected affected regressions pass. Paid proof keeps its actual 9af source/image identity. The separately tested live/history card fix is not claimed deployed on that host. On 9af, original saved settings survived reopening and fresh no-prompt Test authenticated catalog/model/binding; no new Save is claimed. Fresh Linux admission passed. With `claude-sonnet-4-5`, Artifact downloaded exact 27 bytes and matched 121 canonical identities in complete S3-served history. First Stop’s strict quiet FAILED on a 10.373-second gap despite independent cancellation/no late artifact. One targeted unchanged-sampler retry retained all 24 samples: 35.076 seconds quiet / maximum gap 4.584 seconds, 67 ordered canonical identities, and separate CANCELLED/artifact-absent status after the original 180-second deadline. The pre-reload duplicate settled card is disclosed. Exact-ID cleanup for the three new provider agents is awaiting authorization, so full grades and the fresh Artifact’s post-cleanup download remain blocked. A historical deleted-provider artifact was newly downloaded on 9af; original F226 Question/Plan/reconnect and cleanup proof keep their identities. Retained failures: stale fixture 132/133; E8 48/135 with 87 reset-hook failures; Mac 17 SQL failures/skips; earlier title/startup/custody/Stop/deletion failures with unknown causes/bounded recoveries; `dea023` Stop’s warning/50 duplicate identities; pre-repair ACK and original Core CI failures. WSP’s 251-assertion, 7/8-suite attempt failed physical PostgreSQL teardown; the unchanged wrapper rerun passed 128. Retry alone establishes no cause or full local-suite pass. All paid attempts remain recorded; vendor dollar cost is UNKNOWN. ## Risks Experimental: Cursor independently acknowledges neither the effective model nor instruction digest before the first prompt. Explicit values are submitted and checked through behavior; unpriced provider dollar cost stays UNKNOWN. Callbacks need authenticated reachable HTTPS. Scope, revocation, budgets, completion and cleanup remain authoritative. Operator Stop intent is separate from physical containment. Saved legacy agents retain their selected runtime. Prior shared title-save/deletion failures remain disclosed. ## Model Used OpenAI Codex, GPT-6 family, assisted with reasoning, repository tools, code execution, browser verification, and delegated code review. The serving model’s exact API ID and context window are not exposed. No value is inferred from the model catalog. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used and the limits of available model metadata - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have linked public issues or described the issue in this PR - [x] I have not referenced internal or instance-local Paperclip issues or links - [x] My branch describes the change and contains no internal ticket ID - [x] I have run the scoped local tests reported above; pending gates are stated separately - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation - [x] I have considered and documented the risks - [x] All required CI gates pass on the current head - [x] Fresh Greptile is 5/5 with no unresolved findings on the current head - [x] I will address all reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
3d5878980b |
fix(ui): show agent avatars in the chat picker (#18709)
## Thinking Path > - Paperclip helps people manage AI agents and their work. > - Chat lets a person choose an agent for a conversation. > - The shared agent picker shows a role icon by default. > - An icon does not show the agent's saved visual identity. > - This pull request uses the shared avatar component in that picker. > - The picker now matches the avatar in the agent profile and chat sidebar. ## Linked Issues or Issue Description **What happened?** The chat header switcher and main sidebar picker show agent role icons. The conversation sidebar already supplies avatars through its existing render option. **Expected behavior** All chat pickers show the agent's saved avatar. Uploaded images and default appearances use the same shared component. **Steps to reproduce** 1. Open an agent conversation. 2. Use the agent switcher in the chat header. 3. Compare each choice with the agent's profile avatar. Related work: [chat header profile navigation](https://github.com/paperclipai/paperclip/pull/17169) changes profile links. [Task list avatars](https://github.com/paperclipai/paperclip/pull/18607) changes a separate surface. ## What Changed - Use `AgentAvatar` at size 32 as the default in `AgentChatPicker`. - Keep the existing custom render option and selection behavior. - Test custom and default avatars inside the picker. Test the custom render option. - Keep the feature map unchanged for this small UI fix. ## Verification - Passed: focused picker, avatar, sidebar, and chat page tests (36 tests across 6 files). - Regression proof: the new picker avatar test fails with the original icon rendering and passes with this fix. - Passed: `pnpm --filter @paperclipai/ui typecheck`. - Passed: `pnpm --filter @paperclipai/ui build`. - Passed: `pnpm check:token-gates`, `node scripts/check-maps.mjs`, and `git diff --check`. - Full local typecheck and build could not finish because this environment has no Rust `cargo` toolchain. - The full test run was stopped after concurrent validation exceeded the workspace memory limit. The first UI build and local server were killed. The UI build passed when run on its own. Focused tests passed before the full run. - Browser verification passed against a fresh local source-checkout instance with real API responses and embedded PostgreSQL. No agent runs were started. Chromium checked header and sidebar pickers, an uploaded custom image, role search, Enter selection, reload, and a 390-pixel mobile viewport. No page errors occurred. - Desktop, mobile, and sidebar screenshots were captured. They are not committed to the repository. - Previous head `9bf268c308e51113b0a10fb1314444cc99882b04`: all CI checks passed; Greptile was 5/5 with no outstanding findings. - Head `4e85eedd155ccf1624233009905d5ebfe8c3849b`: all 36 focused tests, UI typecheck/build, token gates, map checks and whitespace checks passed again after merging master. Fresh review gave 4/5 because this small fix added feature-map detail. - Current head `e13af2d5650e95ff552b1bfe392820c490eccf23`: removed all four added feature-map lines. The map matches the base version. Map checks and whitespace checks pass. The code and tests did not change. All 30 check runs are complete: 25 passed, 4 were skipped, and 1 was neutral. Build, typecheck, tests, all eight E2E shards, and both CI summary gates passed. The legacy commit status passed. Greptile gave 5/5 with no open findings. All review threads are resolved. The separate Paperclip review ended early with a neutral check; its prior map finding is fixed. ## Risks - Low risk. The change uses the existing avatar component and its image-error fallback. Picker rows now use the same avatar size as the conversation sidebar. ## Model Used - OpenAI Codex, GPT-6, with reasoning, repository editing, shell tools, and browser verification. The runtime does not expose a more specific model ID or context window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #123` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
80318e4567 |
fix(evals): share bounded local review object grammar
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
e16de641ba |
fix(heartbeat): retain Stop and receipt debt after setup failure (#18776)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The server owns execution, Stop delivery, and usage settlement. > - A restored execution can fail during preparation. > - A preparation failure does not prove that earlier provider work never started. > - Replacing its result metadata can remove an existing Stop and falsely release its accounting hold. > - This pull request preserves those obligations at the terminal write. > - The benefit is honest cleanup and receipt debt without new execution authority. ## Linked Issues or Issue Description **What happened?** A recovered native run can fail a configuration check before dispatch in the new controller. The setup failure writes bootstrap metadata with `providerWorkStarted: false`. It also replaces the previous result metadata. This can remove a pending native Stop. Accounting then accepts the new bootstrap flag and releases the original hold despite an incomplete receipt. **Expected behavior** Keep the current run-Stop and receipt debt when setup fails after earlier native execution. Preserve an acknowledgement that commits during the failure write. Fresh setup that never dispatched must remain eligible for pre-provider settlement. Keep account-identity checks effective. **Steps to reproduce** 1. Save admitted native execution, an active provider checkpoint, an original held reservation, and an audited pending run-only Stop. 2. Recover the run and fail preparation with an account configuration error before the new controller dispatches. 3. Run ordinary cost reconciliation with an incomplete receipt. 4. Observe that the base removes the Stop and releases the hold as if the run never started. Related: #18682 repairs exact-session Stop cleanup after restart. This change protects setup failure before that path. #18702 handles AI connection-request waits. #18035 proposes broader accounting policy. #18778 separately handles retained cleanup across account changes. Account, budget, and receipt authority are unchanged here. ## What Changed - Mark setup-failure status writes through an internal lifecycle option. At the atomic write, a prior coordinator attempt bound to the same run, company and task prevents false pre-provider evidence, even with a stale loaded profile. Lock the coordinator before the run so an in-flight claim cannot be missed. Selected execution input alone does not imply provider work. - Keep genuine pre-dispatch setup eligible for settlement after fresh native input is persisted, whether the loaded profile is current or stale. - Preserve only exactly bound native run-Stop fields from the current database row during failure writes, including the intentional handoff marker. Keep the latest acknowledgement when the update waits behind another writer. Normal Stop delivery still advances its own metadata. Replace unrelated stale result fields. - Add real PostgreSQL regressions for receipt debt, original holds, stale snapshots, late acknowledgement, fresh setup, dispatched work, and foreign or task-scoped metadata. - Update the execution contract. ## Verification - On untouched master, the recovered-setup regression fails six assertions: Stop removal, false bootstrap evidence, false accounting settlement, lost debt, and released hold. The fresh-setup control passes. - A second regression fails four assertions with the first repair present when the loaded profile lacks the current saved execution. The atomic current-row check fixes that case. - The focused safety selection passes. It includes real row-lock contention. A controlled concurrent writer seeds audited acknowledgement state while the terminal update waits. The product code retains that state and creates no acknowledgement or provider turn. This race test does not prove physical provider termination. - Fresh native and legacy setup can settle without a receipt. Dispatched native and legacy work keeps its debt. Foreign run, company, issue and task-scope metadata stays excluded. Unrelated old result fields are replaced. - After the final review fixes, all three completion, Stop and budget-cancellation suites pass: 121 tests. The exact `pnpm run typecheck:build-gaps` entry point passes after the final source and service-test edits. Module-boundary checks pass. An ordinary adapter pre-provider receipt control also passes. - Native review also found that fresh selected input could be mistaken for prior execution. Real-admission loaded/stale cases fail against the earlier head; a prior-attempt-only control passes. The correction uses the bound coordinator attempt and includes a foreign-task negative. Reservation `providerStartedAt` is stamped at reservation creation and is not used as provider-work evidence. - A row-lock regression also exposes a first execution claim committing during the terminal write: the prior statement snapshot misses it and falsely releases the hold. Locking the exact coordinator before the run fixes that race and preserves the existing claim lock order. Concurrent Stop acknowledgement tests remain green. - Review found two preservation gaps. New real-database reds show handoff classification loss and overwritten budget Stop delivery progress. Failure-only preservation and the bounded handoff marker address them. - Initial helper-import and reservation-order fixture errors are retained separately from causal failures. No assertion or deadline was relaxed to pass the regression. - Current-head Greptile is 5/5. The optional native review stopped before completing and returned NEUTRAL; this is not substantive approval. Its prior finding was reproduced and fixed. Coordinator and independent source/evidence reviews found no additional issue; neither is a hosted recovery test. - These are local database tests with saved provider state. No hosted recovery, physical provider termination, historical-state repair, or deployment is claimed. Full repository tests and build passed in current-head CI: 55 successful checks, three expected skips, and the explicitly incomplete optional native review returned NEUTRAL. Required verify and e2e aggregates passed. ## Risks - The current database row must retain canonical Stop state while replacing other result metadata. The preservation is limited to the same native run, company, issue and run-only Stop scope. - Setup-failure history checks keep uncertain work pending. They do not create a receipt, acknowledgement, reservation or charge. - The setup-only write option leaves ordinary adapter pre-provider receipts unchanged. No account-identity, budget, dispatch, or provider authorization gate is bypassed. - Already overwritten historical state needs separate evidence and repair. This change does not reconstruct it. - No schema, public API, provider profile, or deployment setting changes are required. ## Model Used OpenAI Codex, `gpt-6.1-sol`, Extra High reasoning, with source inspection, code execution, and regression testing. Context-window capacity is not exposed in this session. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing>canary/v2026.1103.0-canary.10 |
||
|
DottaandPaperclip
|
3910352c92 |
feat(runner): add experimental Cursor Cloud provider (#18068)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Runner gives each provider durable execution, tool, and completion contracts. > - Cursor Cloud currently has a direct adapter outside that native contract. > - Cloud work needs exact ownership across HTTP, SSE, Stop, and recovery. > - This PR adds a qualified Cloud API driver and verified proxy to Runner. > - Product setup and hosted placement remain separate dependent changes. ## Linked Issues or Issue Description Refs #5664, #11803, #17299, and #17471. Product integration is #18069. Managed-only API placement is #18310. The related direct-adapter model/transcript work is not replaced. No duplicate native Cursor Cloud integration was found. ## What Changed - Add native `cursor_cloud` / `cursor_cloud_api` with explicit model/repository validation, bounded HTTP/SSE, durable per-turn checkpoints and a qualified Node proxy launched through the Rust facade. - Commit terminal custody before publication/release; coalesce consumers, preserve ambiguous create/Stop/recovery fences, reject late tools, and distinguish definitely-unsent oversized requests from ambiguous responses. - Support scoped cold attachment, current authority, reconnect and verified artifact staging. Keep unsupported warm/external and Windows paths closed. - Preserve early source cursors and verify immutable legacy prefixes before suffix replay, including supported compaction/masking. Optional proof publication failure falls back to complete cold verification. - Preserve canonical records at the durable native producer with JSON-safe privacy and the existing 8MiB bound; drain admitted successor flushes. Reconcile upstream semantic/pack/schema/lifecycle changes and retain S3 and legacy CLI checks. ## Verification Core head `37c442ec8dd4`: all 66 check contexts are terminal successful/neutral/skipped, including required `ci / verify` and `ci / e2e`; fresh exact-head Greptile is 5/5 with zero unresolved threads. Dependent Product `81e828d20a77` and Top `dd2959689627` include the later client handoff correction and are running their own fresh gates. Core: 47/47 Rust, 4/4 compiled-native retirement, 19/19 settlement and normal Runner build passed. Runtime Top `9af1a5de42a2`: build and recursive typecheck passed. Earlier source-scoped evidence includes 398 focused tests, five relay tests and token/type gates. The 251-assertion, 7/8-suite run failed physical PostgreSQL teardown; the unchanged stable WSP suite passed 128 through the documented wrapper. The downstream fixture corrections passed 7/7 browser and 121/121 startup tests without timeout/oracle relaxation; a broken-unsubscribe control fails. Hosted runtime `9af1a5de42a2` has a passing normal build and recursive typecheck. Latest dependent heads additionally contain a pure client fix for stale live/history overlap, with 48 focused tests, UI typecheck and token gates passing. Its new client is not deployed in the frozen hosted image; server/native/API/provider source is unchanged by these four downstream files. Fresh hosted Linux admission on `9af1a5de42a2` retired its owned runner/proxy/socket with zero prompts. The saved configuration survived reopening, and Test authenticated the bound key/catalog/selected model without a prompt. Artifact with `claude-sonnet-4-5` completed, downloaded the exact 27 bytes, matched 121 canonical identities against complete logs and independently reached FINISHED. The active Stop retry was clicked within its immutable boundary, produced one canonical interrupted turn and acknowledged run-only cancellation, retained a valid 35.076s quiet window (maximum gap 4.584s), and independently remained CANCELLED/artifact-absent beyond the original 180-second command deadline. All provider-observer credential callbacks, DB clients and loaders closed; accounting is settled and explicitly unpriced. The first fresh Stop retains its failed 10.373s sampling gap and every sample. On the upgraded9af server, the historical F226 artifact was reopened and downloaded with its exact original hash after its provider agent had been deleted/GET404 earlier today. Historical F226 full Artifact/Stop cleanup, Question, corrected Plan and reconnect retain their original identities. Deleting the three newly created disposable provider agents and the new-case post-delete grade remain blocked on explicit user approval after automatic review rejected expanding the two original approved IDs. No new-case cleanup or complete fresh matrix pass is claimed; this is separate from the directly observed task/Stop behavior and retained exact cleanup proof. Retained failures: stale fixture 132/133; E8 48/135 with 87 reset-hook failures; Mac 17 SQL failures/skips; earlier startup-title, title-save, custody/Stop and deletion failures with unknown causes and bounded recoveries. `dea023` Artifact passed 17 checks but Stop failed its false warning and 50 duplicate log identities. Pre-repair ACK failure and original Core CI failure remain recorded; retry alone establishes no cause or full local-suite pass. Provider dollar cost stays UNKNOWN. ## Risks Experimental: Cursor independently acknowledges neither the effective model nor instruction digest before the first prompt. The caller’s explicit values are submitted and behavior checked live. Unpriced provider dollar cost stays UNKNOWN. Only qualified per-turn controllers are admitted. Uncertain work and incomplete cleanup stay fenced. Existing damaged/duplicate archives remain failed evidence. Prefix proof adds bounded object-store I/O; the 8MiB producer bound does not guarantee future providers’ output fits. ## Model Used OpenAI Codex, GPT-6 family, assisted with reasoning, repository tools, code execution, browser verification, and delegated code review. The serving model’s exact API ID and context window are not exposed. No value is inferred from the model catalog. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used and the limits of available model metadata - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have linked public issues or described the issue in this PR - [x] I have not referenced internal or instance-local Paperclip issues or links - [x] My branch describes the change and contains no internal ticket ID - [x] I have run the scoped local tests reported above; pending gates are stated separately - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation - [x] I have considered and documented the risks - [x] All required CI gates pass on the current head - [x] Fresh Greptile is 5/5 with no unresolved findings on the current head - [x] I will address all reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
18c139f182 |
fix(evals): classify explicitly local review exceptions
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
8c00414cf6 |
fix(evals): scope local drafts and negated exceptions
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
c89461e34e |
test(activity): accept present-perfect human reassignment
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
2b14ec03a4 |
fix(ai-connections): remove account execution locks (#18778)
## Thinking Path > - Paperclip manages AI agents and the accounts that supply their model credentials. > - Several runs can use the same AI connection at the same time. > - Runtime preparation treated unfinished credential cleanup as exclusive account ownership. > - Manual account tests also held a company credential mutex for the entire provider test. > - A failed cleanup could therefore cancel fresh runs before the provider started and keep them in automatic retries. > - This pull request removes that execution gate and those read and probe locks. > - Each invocation keeps its own credential home. Credential updates still use the existing freshness and authorization checks. ## Linked Issues or Issue Description **What happened?** Fresh runs stopped with `ai_connection_busy` and the message “The selected AI account is still completing credential cleanup. Retry shortly.” Pending cleanup from an earlier invocation blocked every new invocation of that account. Concurrent manual account tests also waited on a company mutex. **Expected behavior** AI connections support concurrent runs and account tests. Cleanup for one private runtime must not prevent another private runtime from using the same account. **Steps to reproduce** 1. Prepare a managed OpenAI account and retain an earlier runtime whose cleanup has not finished. 2. Prepare another run with that account. The old implementation rejects it before provider work starts. 3. Alternatively, keep one manual account test open and start a second test with the same account. The old implementation prevents the second provider callback from starting. **Paperclip version or commit** Reproduced with two PostgreSQL regressions on `17e8bd773543d62ab3f444c55255acfd82575efd`. Both fail before the fix and pass after it. **Deployment mode** The reported deployment used authenticated Cloud. Local reproduction uses disposable PostgreSQL and fixture credentials. No live provider credentials are required for these regressions. Related search found #15952, which concerns account removal and provider application archival. It does not remove this runtime cleanup gate. This fix does not introduce the account concurrency cap requested in #14564. ROADMAP.md was checked; this change restores concurrent account use. ## What Changed - Remove the account-wide cleanup assertion and cleanup mutex from run admission. - Read ordinary subscription credentials without acquiring the company file mutex or grant and secret write locks. Only an actual durable replacement recovery enters the existing write transaction. - Remove the mutex held across manual account tests. - Retry retained cleanup without making new runtime preparation wait for it. A home-specific cleanup check considers only that exact private home. - Keep cleanup attempts for one runtime idempotent. Keep physical retirement, credential freshness, reconnect, revocation and company checks. - Update concurrency, refresh and cleanup regressions, and the implementation contract. ## Verification - Before/after: the pending-cleanup and simultaneous-probe regressions both fail on the original code and pass with the fix. - Affected suites: 186 tests passed across `ai-connections.test.ts`, `antigravity-ai-connections.integration.test.ts`, and `retained-google-credential-cleanup.test.ts`. They use real PostgreSQL, encrypted fixture credentials and controlled provider callbacks. - `pnpm -r typecheck` passed. - `node scripts/check-maps.mjs` and `git diff --check` passed. - `pnpm build` passed on the rebased branch. The server typecheck and production server build both passed again on final commit `03bd3973bdf85108bbd82355abdb217f0e1c83bd` after the runtime-home ownership changes. - Caller coverage: all 166 cases pass across native credential handoff, managed Google subscription writeback, model discovery, heartbeat configuration and hired-agent execution. The full hiring file passed all 79 cases after its admission assertion was corrected. Together with the affected suites, 352 targeted cases pass. - The final home-binding check passed all 188 cases across the account, managed Google writeback, model discovery and retained-cleanup suites. - Review follow-up: two real-PostgreSQL cases pass for quota refresh and concurrent Google account tests. The quota case observes a credential row update in progress, verifies that the runtime takes no account file mutex, and confirms that the private home receives the committed replacement token. The Google case completes a second provider callback while the first callback remains open in a separate private home. - Greptile rated the runtime change 4/5 and found only misplaced map patch notes. Those additions were removed. The implementation documentation remains in `doc/connections/AI-CONNECTIONS.md` and the implementation contract. - `pnpm test:run` was attempted. Chat partition 1 passed 351 expanded cases. Partition 2 passed 324 cases and failed one Teams file-receipt API case at its 15-second deadline: `refuses generic resolution with stale version transfer preconditions without changing receipts`. That case exercises publication resolution, not AI account execution. Its isolated rerun passed. The remaining local partitions did not run after that failure. This is not a claim that the full local suite passed. - Wider caller checks initially identified old serialization expectations and two cleanup retry cases. Updated assertions now require concurrent account use. The partial-removal correction lets a retained home finish removal after a durable credential handoff. All affected caller cases pass, including the full hiring file. - Live provider concurrency and deployment recovery have not been qualified. This PR does not claim that the reported Cloud deployment contains the fix. ## Risks - A new invocation can use the currently saved credential while another runtime still owns a refreshed copy. The existing writeback checks decide which credential can replace the saved value. - Actual credential updates and durable replacement recovery retain atomic storage transactions. This change removes execution serialization, not the integrity checks on stored credentials. - Unfinished cleanup retains its private home and its existing idle hold. This change does not certify a previous provider process as stopped. - There is no database migration, credential format change, permission expansion or ambient-account fallback. ## Model Used - OpenAI GPT-6 through Codex desktop, with source editing, shell execution, PostgreSQL fixture tests and read-only browser diagnosis. The session does not expose the exact deployment ID or context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
1eee137d16 |
fix(activity): use authoritative changes for timeline assignments
Exercise the real task PATCH producer before timeline reads, and bind narrative attribution directly to the reassignment action. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
7b2a720986 |
fix(evals): keep shared negation and future approval conditions
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
d8f5e2bcdc |
test(activity): synchronize selected timeline metadata and calibration
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
adde1ae0e9 |
test(activity): distinguish recorded person from other human actors
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
29487a08b0 |
fix(activity): preserve recorded handoffs in work timelines
Calibrate complete narrative attribution without first-match wording, and retain independent command, audit, privacy and link checks. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
2d2c916633 |
fix(evals): retain external locations in proposal claims
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
234632836b |
fix(evals): verify declined action guidance and attributed denials
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
a45d8268b2 |
test(evals): use production previous-question navigation
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
7bbb1939c8 |
fix(opencode): correlate closed finish validation stages
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
17c89f48f6 |
fix(opencode): retain closed finish input diagnostics
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
86b365f992 |
feat(auth): add independent agent configuration read permission (#18766)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Agents need to inspect configuration and instructions when they set up teammates. > - Configuration reads currently depend on configure or suggestion grants. > - Legacy standard agents did not receive those newer grants. Some self-read routes also lose the target agent identity. > - This pull request adds an independent read grant and applies only that grant to eligible existing agents. > - The benefit is that agents can inspect and verify configuration without new write authority. ## Linked Issues or Issue Description Refs #15283. Related earlier proposals: #8064 and #12050. Those proposals target older authorization code; this change uses the current configure/suggest model and adds a read-only rollout for existing agents. Related bounded peer metadata proposal: #14160. **What existing behavior does this improve?** Agent configuration, configured skill, instruction, and revision reads through the existing APIs and native instruction tools. **Current behavior** An agent needs `agents:configure` or `agents:suggest-changes` to inspect a peer. Older standard agents can hire a teammate but cannot verify the saved instructions. Configuration and revision routes also reject self-reads because they authorize a company resource instead of the target agent. **Proposed behavior** Add `agents:read` for redacted configuration and instruction reads. Keep `agents:configure` as read/write and preserve suggestion-grant compatibility. Standard new agents receive the read grant. A one-time migration grants only reads to eligible existing standard agents. Individual routes authorize the target agent and grant scope. **Reason and benefit** Read access can be granted without configuration changes, rollback, or environment probes. Legacy agents can verify teammates without receiving new write permissions. ## What Changed - Add `agents:read` to the shared permission contract and standard-agent defaults. - Accept read, configure, or suggestion grants for configuration reads, with target scope and the responsible-user ceiling. - Fix target-aware self and scoped configuration, skill, and revision reads. - Add an audited data migration. Skip low-trust, built-in, inactive, pending, terminated, setup, and scoped configuration/suggestion cases. Preserve existing read scopes and later revocation. - Update the specifications, default-permission documentation, and agent access map. ## Verification - Focused acceptance: 276 tests passed across seven suites on commit `a37314df88`: authorization, real-key HTTP access, route authorization, canonical instructions, default grants, migration replay, and lifecycle configuration. - The HTTP journey uses a real hashed agent API key and a disposable PostgreSQL database. It verifies redaction, scoped reads, denied edits and rollback, cross-company isolation, and both agent and responsible-user revocation. - New read-grant regressions failed before the authorization change. A realistic paused/pausing/resuming migration fixture also failed before the lifecycle filter correction and passed afterward. Setup-held paused agents stay excluded. - `pnpm -r typecheck` and `pnpm build` passed on the rebased branch; the final lifecycle migration correction also passed the database package typecheck and build. The latest commit changes only a test assertion. - `pnpm --filter @paperclipai/db check:migrations`, `node scripts/check-maps.mjs`, and `git diff --check` passed. - Full current-head [CI run](https://github.com/paperclipai/paperclip/actions/runs/41641531426) passed: typecheck, build, all general and serialized test lanes, Runner checks, all eight browser shards, canary dry run, policy, and aggregate gates. The local full-suite attempt was stopped after exposing the outdated cross-company 403 assertion; interrupted local full-suite attempts are not counted as passes. ## Risks The migration deliberately expands redacted configuration and instruction visibility for eligible existing standard agents in the same company. An absent older write grant does not deny the new read capability. It adds no write grant. Scoped configure/suggestion grants and all existing read-grant scopes stay intact. Administrators can remove the independent read grant after the one-time rollout; ordinary updates and startup do not restore it. Configure and suggestion grants still imply reads within their own scopes. Cross-company configuration/history reads now return 404 through the common target-aware helper. No schema change or credential-value disclosure is added. ## Model Used OpenAI GPT-6 through Codex, with reasoning, repository tools, and code execution. The host does not expose a more specific runtime model ID or context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
22ce1a8c32 |
fix(native): refresh prior warm controller projections (#18718)
> - Paperclip is the open source app people use to manage AI agents for work. > - Native remote sessions keep their controller state in a database checkpoint. > - An admission can also leave a full local copy of a prior run's state. > - The prior owner can commit more state without changing its run ID. > - Warm admission skipped a refresh when the local and saved run IDs matched. > - This pull request refreshes that prior-run copy when no retained warm owner exists. > - The next turn can use the committed state while existing ownership checks remain effective. **What happened?** A prior warm run can complete after an earlier admission reads its controller checkpoint. The local copy can still contain a pending command or lack the final terminal event. A later admission sees the same run ID and skips restoration. It can then hold work against that old copy even though the database checkpoint has settled. **Expected behavior** Read the canonical prior-run checkpoint before reuse when no retained in-memory owner exists. Keep current-run state with its controller. Keep unresolved canonical work and invalid checkpoint bindings blocked. **Steps to reproduce** 1. Create a full prior-run local projection while a command is pending. 2. Complete that command and record the terminal in the canonical checkpoint under the same run ID. 3. Start an ordinary remote warm continuation without a retained in-memory owner. 4. Before this change, the local command remains pending. With this change, the completed command and terminal replace the stale copy. **Paperclip version or commit** Reproduced on master `36df8933f0` before changing production code. Related work: #18018 restored per-turn checkpoints; #18343 added controller replacement recovery; #18554 restored missing remote suspension receipts. This change handles a full prior-run projection whose run ID already matches the saved supervisor. It does not change the settlement guard or MCP compatibility policy. - Refresh a prior-run warm controller projection even when its run ID has not changed. - Keep current-run and retained warm owner authority on the existing path. - Add five executor cases for a settled source, a pending canonical command, a current run, legacy file-backed state, and an invalid checkpoint binding. - Add an opt-in Linux continuity case with a stale full local projection and two follow-up turns. - Update execution documentation and the paired session verification recipes. - RED before the production edit: the executor fixture retained `pending` in its local command record when the canonical record was `completed`. - GREEN on the final commit: 48 focused tests passed across executor restoration, retained warm owner rotation, per-turn restoration, and remote settlement guards. - `pnpm --filter @paperclipai/server typecheck` passed. The server TypeScript check also passed after the integration fixture was added. - `node scripts/check-maps.mjs` and `git diff --check` passed. - The unit fixture uses the real durable controller serializer and the real executor entry. It stops at current tool authority before provider/backend creation. It proves restoration and guard behavior, not full task acceptance. - Actual protocol RED/GREEN: the same opt-in continuity fixture uses real PostgreSQL, TLS, the current Linux Rust Runner, and a deterministic fake Codex backend. With only the executor restored to its parent revision, continuation fails with `prior_owner_terminal_state_indeterminate` in the settlement guard after one provider turn (20.16 seconds). The PR source completes all three turns with one provider thread and final suspension (34.77 seconds). Hard recovery permits one provider-process replacement; the healthy final follow-up starts no additional process. Both runs use the same Runner binary and owned container. - The exact PR executor source was restored after the baseline comparison, its Git blob was verified, and the worktree is clean. The disposable local container was removed. This fixture proves the controlled stale-projection recovery path; it does not establish a live model or Daytona API result. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on the current commit, with no unresolved review threads. No original task was retried and no staging sandbox lifecycle was changed. - Admission without a retained owner performs one additional scoped checkpoint read when a full prior-run projection already exists. - The canonical checkpoint must still pass its company, task, session and record validation. A pending canonical command stays held. Current-run and retained owner state are not replaced. - The regression proves a stale local projection defect. It does not establish that a historical incident had those same local bytes. OpenAI Codex, GPT-6 family. The harness does not expose the exact serving model ID or context-window size. Capabilities: repository inspection, code generation, tools, and test execution. - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
ae36d63757 |
fix(native): publish fresh retry sessions for later turns (#18713)
> - Paperclip is the open source app people use to manage AI agents for work. > - Native tasks save a provider session for later comments and monitor checks. > - The goal projection also records the run that owns the session event sequence. > - A fresh retry uses a new native session, but the projection accepted only the old session. > - The accepted checkpoint then rejected session publication because the goal source still named the old run. > - This pull request lets the current task executor advance that source after a terminal predecessor with no unfinished goal. > - Later turns can then use the fresh retry's saved session. Related: #12669, #18343. I searched open and closed PRs for source epochs and native task-session publication. I found no duplicate repair. **What happened?** A successful fresh native retry could retain its predecessor's task-session pointer. A later comment or monitor selected the old provider session. The goal source transition required both runs to have the same native session. The terminal checkpoint then saved a `suppressed_newer` publication result. **Expected behavior** The current task executor can publish its accepted fresh session. A later turn resumes that session. Delayed events cannot restore the old source or revive a cleared goal. **Steps to reproduce** 1. Complete a native turn and retain its task-session row. 2. Admit a fresh session on the same task without deleting the row. 3. Complete that turn, then send an ordinary follow-up. 4. Before this change, the saved pointer can still name the first session. **Paperclip version or commit** Reproduced on source commit `36df8933f0` with a PostgreSQL regression before the production change. **Deployment mode** Source checkout. The affected producer runs for local and remote native sessions. - Permit a different native session to advance the goal source only when its run holds the task execution lock and assignee, and the task is in progress. - Keep the terminal predecessor, native runtime, runner identity, absent or completed goal, source sequence, and conversation reset checks. - Test accepted session publication, a fresh heartbeat followed by an ordinary resumed turn, stale events, current-owner loss, and unfinished or unreadable goals. - Add an opt-in live Codex file-persistence check. Update execution documentation and both recovery maps. - Red: the new PostgreSQL publication regression failed before the producer change. The saved session, display ID and goal source still named the predecessor. - Green: 72 distinct tests passed across the final focused runs of `runner-goals.test.ts`, `native-task-session-publication.test.ts`, `native-session-resumption.test.ts`, and `remote-warm-session-recovery.test.ts`. - The assembled heartbeat test admits three turns, opens two provider sessions, resumes the second session once, and stores three results. It uses a deterministic provider backend. - `node scripts/check-maps.mjs` and `git diff --check` passed. - Real local Codex acceptance passed on macOS arm64 with `gpt-6.1-sol` and explicit warm mode. Three heartbeat turns saved three exact file lines. The fresh second turn published its new normalized and provider session IDs. The ordinary third turn retained both IDs. The fixture closed its warm owners and removed its private credential copy and workspace. The production code was at `450e15e`; the final test-only commit selects explicit warm mode and uses the existing owned-home cleanup helper. The production change is identical in the final head. - The earlier per-turn fixture preserved all three file steps and selected the correct normalized session, but provider-thread reuse failed. `run.attach` reported that the Codex provider had pending events; the safe fallback used another provider thread. This remains a separate failure. The first fixture cleanup also failed on immutable instruction assets. The owned-home helper fixed cleanup, and the second per-turn attempt exposed the provider identity assertion directly. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on the current commit, with no unresolved review threads. Broad local workspace checks were not duplicated; the focused regressions and actual provider acceptance above were run locally. No merge, deployment, or recovery of an original task is included. - Source epoch changes affect saved goal state. The fresh-session path requires current task authority and refuses unfinished or unreadable saved goals. - The change starts no provider work and changes no prior-owner recovery guard. Unknown external outcomes retain their reconciliation hold. - A prior session remains held when it lacks the required authority or state proof. - Per-turn provider-thread reuse remains an open coverage gap described above. This change does not alter that attachment guard. OpenAI Codex, GPT-6 family. The exact model ID and context window are unavailable in this session. The agent used code inspection, tools, and executable tests. - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
ebd6181e53 |
fix(opencode): keep native tool calls pending until Runner settlement (#18711)
> - Paperclip is the open source app people use to manage AI agents for work. > - Native OpenCode sends semantic tool calls through the durable Runner protocol. > - The Runner retains a call until its result is known. > - OpenCode and its HTTP bridge each stopped waiting after 30 seconds. > - A slow response could leave a call pending after the provider had moved on. > - A later run then failed the pending-call guard during attachment. > - This pull request keeps native calls under their Runner owner until settlement or cancellation. > - The benefit is one delivered result and a safe next run. Related: #18343 applies the durable wait policy to native ACPX calls. This change applies that policy to native OpenCode. **What happened?** An assigned tool call took more than 30 seconds. OpenCode reported an MCP timeout while the Runner still retained the request. The source turn could finish, but suspension failed. A later `run.attach` failed because provider tool calls were still pending. **Expected behavior** The native proxy must wait for the authoritative result under the Runner owner. Stop, bridge close, and lease expiry must still end the wait. Unknown effects must still block attachment. **Steps to reproduce** 1. Start a native OpenCode run with an assigned semantic tool. 2. Hold the controller's result for 31 seconds. 3. Observe the provider's MCP result and attach a later run to the session. **Paperclip version or commit** Reproduced against `6d72243d8af0362d73fbcb5331fd42a384800d4e`. The branch is based on `36df8933f0`. - The native OpenCode proxy enables a driver capability for durable tool requests. - That capability removes the bridge's separate request timer and sets OpenCode's fixed MCP client ceiling to one hour, consistent with the native ACPX sidecar. - Ordinary driver calls keep their 30-second deadline. Assigned external MCP connections keep their existing budget. - Cancellation, result replay, tool authorization, and the pending-call attachment guard keep their existing contracts. - The native regression holds a result for 31 seconds, then completes a second run with the same Runner process and no pending provider calls. - The deterministic provider fixture honors the emitted MCP budget and retains the initial completion contract across ordinary follow-ups. - Execution semantics and the session verification maps describe the wait and its safety checks. - Red before production edits: the bridge aborted a durable request at the ordinary 30-second boundary. The driver emitted `30000` instead of the expected `3600000` client budget. - The final combined bridge, driver, proxy, and native protocol run passed: 95 tests in five files (71.17 seconds). - The native delayed-result probe passed with the current Rust Runner and current bundled proxy. It completed the later run with the same Runner PID and an empty pending-call ledger. - Actual provider check passed on macOS arm64 with `opencode-ai@1.18.34`, the changed source, and a local deterministic OpenAI-compatible model. One assigned write waited 31 seconds and saved one file. Both turns completed. The second turn retained the saved tool receipt. The write ran once across five model requests. - Bridge safety cases cover result replay without another action, ordinary timeout behavior, MCP cancellation, and teardown of an unanswered durable request. - Both runner TypeScript projects passed typecheck. The first attempt lacked the eval-kernel build; it passed after building that declared dependency. - Map alignment, module boundaries, and `git diff --check` passed. - The first native follow-up probe failed because its synthetic provider forgot the initial completion-contract revision. This caused a repair turn and another discovery call. The fixture now retains the contract, and the same probe passed. The failure is retained here. - Runtime acceptance used the implementation at `583d316ad4`. The later clarification commit changes comments, documentation, and a test name only. It does not change runtime behavior. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on this commit, with no unresolved review threads. No deployment or retry of the affected historical task was performed. The actual provider check used a local model; it does not qualify a deployed Linux sandbox or a live model service. - A native call can wait longer while its Runner owner remains valid. Stop, teardown, and lease expiry still end the wait. The MCP client has a separate fixed one-hour ceiling. Runner ownership can renew beyond that ceiling. If the client deadline expires first, the operation outcome remains unknown and the existing hold must still block replay and run attachment. - This prevents new premature timeouts. It does not classify or replay an old request whose effect remains unknown. - No schema, migration, credential, or permission change is required. - OpenAI Codex, GPT-6 family. The harness does not expose the exact model ID or context window. Capabilities used: code generation, tool-assisted diagnosis, local execution, and test analysis. - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
44a9a043b8 |
fix(ui): make the full chat agent selector clickable (#18708)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Chat gives each person a standing conversation with an agent. > - The Coordinator chat header shows an agent avatar, name, and caret. > - Only the caret opened the agent picker. The avatar and name were separate text. > - This pull request puts all three parts in one button that opens the existing picker. > - The benefit is a larger click target with one keyboard focus stop. ## Linked Issues or Issue Description **What happened?** In Coordinator Chat, the caret opened the agent picker. The agent avatar and name did not open it. **Expected behavior** The avatar, name, and caret must form one control. Each part must open the same picker. **Steps to reproduce** 1. Enable the personal Coordinator workspace and open an agent chat. 2. Click the agent avatar or name in the header. 3. Compare the result with a click on the caret. ## What Changed - Put the avatar, name, and caret inside one button in `CoordinatorAgentSwitcher`. Keep its current-page marker for screen readers. - Add an optional trigger to `AgentChatPicker`. The shared dialog handles keyboard access and focus return. - Add an optional breadcrumb control slot. The chat owns its button, while the text label still sets the document title. - Add regression cases for both header layouts on desktop and mobile. Check all three click areas, Escape, focus return, and selection of another agent. - Add a Coordinator chat Storybook case with production components and fixture API data. Update the feature recipe. ## Verification - Passed 235 focused tests for `BreadcrumbBar`, `AgentChatPicker`, `BreadcrumbContext`, and `IssueDetail`. The header cases also check the current-page marker. - Passed UI typecheck and UI production build. - Passed token gates, map checks, and Storybook catalog checks. All 30 map and catalog regression tests passed. - The new header cases failed before the change and passed after it. - Full local typecheck could not finish because this workspace has no `cargo` command. - Full local build and the first UI typecheck attempt reached the 8 GB workspace memory limit. The separate UI build and typecheck passed. The full local test run was stopped to free memory for browser verification; it has no complete verdict. - Passed 18 browser checks in Chromium across desktop light, mobile light, and desktop dark. Each case checks avatar/name/caret clicks, Enter, Space, Escape/focus return, and selection of QAChecker. No page render errors occurred. - Captured six screenshots of the header and picker across these viewports and themes. Review images remain outside the repository. - Browser review first found stale fixture responses for recent comments, decision receipts, chat handoffs, and ordinary-task email state. Correct API response shapes removed those fixture render failures before the successful review. ## Risks - Low risk. The shared breadcrumb renderer has a new optional control slot. Callers that omit it retain their existing rendering. - Chat selection uses the existing picker and route. No database or API contract changes are required. - The Storybook API is simulated. It does not prove provider execution or server persistence. ## Model Used OpenAI Codex, GPT-6. The runtime did not expose the exact deployed model ID or context window. The agent used code editing, terminal checks, and browser automation. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
a61cf38106 |
docs: keep the feature map focused on substantial features (#18767)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The feature map describes product capabilities and user journeys. > - Small fixes and test changes have added long notes to that map. > - These notes make substantial features harder to find. > - This pull request removes those additions from the last week's changes and sets a clear rule for future additions. > - The benefit is a smaller map with stable capability IDs and accurate entry points. ## Linked Issues or Issue Description **Issue type** Unclear or confusing documentation. **Where is the issue?** `maps/feature-map/`, the map maintenance instructions, and repository rules. **What's wrong?** Between October 27 and November 3, 2026, 115 commits changed the feature map. Many additions describe bug fixes, performance details, small UI changes, or new tests. They do not introduce substantial product features. The old synchronization instructions also encouraged a map edit for every behavior change. **Suggested fix** Add a rule that requires a substantial new feature before expanding the map. Remove minor-update notes. Keep genuine feature additions and minimal corrections to existing instructions. The review covers master through `b22ef22d56`; it does not revert application code. Conflict resolution preserves later map changes from master outside that audit snapshot. Searched open GitHub PRs and issues for related feature-map work. No duplicate cleanup was found. Substantial feature additions retained here include Coordinator (#17983), the agent YAML editor (#18036), document image galleries (#18157), remote workspace storage (#18403), native Kimi (#18391), proactive chat messages (#18374), and the Sentry receiver added in #18714. ## What Changed - Add `rules/feature-map-scope.mdc` and register its `maps/**` scope with Greptile. - Align `AGENTS.md` and the shared map contract with that rule. Routine fixes and test changes no longer require incidental map edits. - Remove 755 net lines from 23 feature recipes. Preserve all 35 areas and 192 capability IDs. - Keep substantial features and the corrections needed for current navigation, provider availability, and access boundaries. Examples from the reviewed changes: | Removed or shortened addition | Why it does not need a feature-map expansion | | --- | --- | | Rounded composer image thumbnails (#18689) | Small UI change within existing task creation. | | Connection profile reads and review-count queries (#18687, #18688) | Performance details within existing app permissions. | | Warm-session, Retry, export-capacity, and tool-wait cases | Regression details for existing execution and recovery. | | New task-search, project-context, document-persistence, and routine test journeys | Test additions for existing capabilities. The tests remain in place. | | New task button placement (#18703) | Another entry to an existing task form. | | Local browser verification notes | Run evidence belongs in the PR or test report. | ## Verification - `node scripts/check-maps.mjs` passes: 35 areas, 192 capability IDs, valid references and contract fields. - `node scripts/storybook-catalog.mjs` passes: 328 files and 3,068 stories. Generated indexes remain ignored. - Compared capability keys and navigation metadata in every edited recipe with the original source. All are unchanged. - Checked rule frontmatter, Greptile registration, representative scope matches, and `git diff --check`. - `pnpm -r typecheck` and `pnpm build` pass on `9a60eef8eb8da9bc7c3a02203c79bec12e8c8b6b`. The final `a0801e8a939494ecde3739097a80258917428e94` commit only corrects two lines in the plugin-panel recipe; map and catalog checks pass again. - Greptile reviewed the final `a0801e8a939494ecde3739097a80258917428e94` head at 5/5 with no findings. The earlier review found one inaccurate phrase about unavailable plugin panels. The hook confirms that readiness gates automatic opening, while declared panels remain selectable. The final documentation commit corrects that phrase and the thread is resolved. Paperclip's refreshed review is advisory and remains queued. - The earlier CI run failed two assertions: [server 4/12](https://github.com/paperclipai/paperclip/actions/runs/41634761842/job/125914173960) counted three Anthropic identity calls instead of two in `connection-management.integration.test.ts`; [chat 1/3](https://github.com/paperclipai/paperclip/actions/runs/41634761842/job/125914174257) observed zero publications instead of one in the same-principal endpoint-settlement case. Both exact tests pass locally on the rebased head. This PR changes neither their source nor their expectations. These failed results remain historical evidence. - Final-head CI passed server 4/12 and all other suites except the first [chat 1/3 attempt](https://github.com/paperclipai/paperclip/actions/runs/41637834618/job/125925926131), which repeated the same zero-publication assertion. The browser gate passed. The single failed-job retry was accepted for this unchanged head in [workflow attempt 2](https://github.com/paperclipai/paperclip/actions/runs/41637834618/attempts/2). [Chat 1/3 passed on that retry](https://github.com/paperclipai/paperclip/actions/runs/41637834618/job/125933199635), and the aggregate verification gate passed. Both required gates, `ci / verify` and `ci / e2e`, are green on `a0801e8a939494ecde3739097a80258917428e94`. The one retry allowance is consumed. - The first full local `pnpm test:run` attempt was interrupted before completion so master could be rebased without changing source under running tests. It is not a passing test result. Full current-head CI remains required. - No application code, tests, schemas, or lockfiles changed. ## Risks - Low runtime risk. This change only edits documentation and review guidance. - Detailed regression instructions become less visible in the feature map. Their tests and technical documentation remain available. - Some additions mix a new feature with a small fix. Those blocks retain the feature and shorten the incidental detail. ## Model Used - OpenAI Codex, GPT-6 family, with reasoning, Git history inspection, repository tools, and code execution. The session does not expose a more specific model deployment ID or context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass — map, catalog, capability-ID, and rule-scope checks; full Vitest limits are stated above - [x] I have added or updated tests where applicable — no new runtime tests apply to this documentation change - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
|
17e8bd7735 |
fix(runner): save native checkpoints when the sandbox disk is full (#18705)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work > - Native runs on a remote sandbox save two things after each turn: the provider checkpoint and the workspace export > - An agent can fill the sandbox disk during its own work, for example with a package install > - The turn still finishes, but both saves need free sandbox disk, so they fail and the task becomes blocked with no clear reason > - This pull request makes the checkpoint work on a full disk and makes a full disk visible in the export failure > - The benefit is fewer stranded checkpoints, and an operator sees at once why an export failed ## Linked Issues or Issue Description Related: #18715 added the shared-principal response-recovery control synchronized below. #18644 reduced the scratch disk that the workspace capture uses. #18172 added the disk-free streaming fallback to the Daytona directory sync-out. Both were deployed when this failure occurred. Neither changes the checkpoint copy, and neither can help when the disk has no free space. **What happened?** A native agent ran a package install for a large monorepo in a 10 GB Daytona sandbox. `df` in the run showed `10G 10G 4.6M 100%`. The agent committed its work and reported `done`. Then: 1. The provider checkpoint made a filtered copy of the provider directory inside the sandbox (`tar -cf archive; tar -xf archive -C snapshot`) to apply its exclusions. On the full disk, tar failed. The run recorded `runner_remote_checkpoint_snapshot_failed` without the reason, and the partial copy stayed on the disk. 2. Workspace finalization failed three times, and file preservation failed six times. The capture builds its archive inside the sandbox, so each attempt failed. Every failure was recorded as `remote_workspace_operation_failed` / `export_interrupted`, so nothing named the full disk. 3. The task went to `blocked`. The only copy of the commit was in the retained sandbox. **Expected behavior** The checkpoint does not need free sandbox disk when the provider can stream. A workspace export that cannot fit says that the sandbox disk is full. **Steps to reproduce** 1. Run a native agent with a provider whose checkpoint has exclusions (for example ACPX) in a small remote sandbox. 2. In the turn, fill the sandbox disk (for example `pnpm install` of a large repository, or `fallocate` of the free space) and finish the task. 3. The checkpoint fails with `runner_remote_checkpoint_snapshot_failed`, and every workspace export attempt fails with a generic code. **Paperclip version or commit** Observed on `410b0e27015`. This change is based on `5955dd4eb4c`. ## What Changed - Synchronize the shared-principal response-recovery regression with its saved comments and empty final deadline. Cover an in-flight discovery invalidation. Keep every authority lock, exact publication, error, and no-retry assertion. This changes test synchronization only. - `syncRemoteRunnerDirectoryOut`: when the in-sandbox checkpoint copy fails with a full-disk tar diagnostic, remove the partial copy and pass the exclusions to the provider sync-out (`exclude` on the directory mapping). The Daytona sync-out streams the directory without sandbox disk when it cannot write its archive. Other copy failures and timeouts still throw `runner_remote_checkpoint_snapshot_failed`. They now also remove the partial copy. - `remote-workspace-transport.ts`: the helper install and the capture scripts report `remote_workspace_sandbox_disk_full: ...` when Node reports `ENOSPC` or `EDQUOT`, or a Git or tar child reports "No space left on device". `command()` passes only that exact message through, like the existing unfinished-Git-operation message. - `native-workspace-finalizer.ts`: a remote save that fails with that message adds it to the failed operation's stderr. The failure code, the retry budget, and the recovery flow do not change. ## Verification - Current-head infrastructure interruption: the [Build job](https://github.com/paperclipai/paperclip/actions/runs/41637668555/job/125923826158) completed the actual build and production bundle boot, then the runner received a shutdown signal during the clean public Grok npm-install check. GitHub marked that step cancelled. The exact host-stop cause was not retained. One bounded rerun of only the failed Build and dependent `ci / verify` jobs passed on the unchanged head. [Build attempt 2](https://github.com/paperclipai/paperclip/actions/runs/41637668555/job/125931400745) completed successfully, including the clean Grok npm-install check. No product code, assertion or workflow changed for this interruption. - CI failure retained: commit `6c5937fcae` failed [chat shard 1/3](https://github.com/paperclipai/paperclip/actions/runs/41634638570/job/125913957217) because the shared-principal fixture saw zero saved comments after `worker.ready` and zero active work. The other chat child passed. The exact case passed once locally and then 30 times on the same Node 24.22.2 version as CI; those passes alone do not explain the CI interleaving. - Controlled RED/GREEN: invalidate the real database discovery once while its read is in flight. The worker correctly retains a future reconstruction deadline, but the old initial-pass/active-count barrier reproduces the same zero-versus-one comment assertion. Waiting for both saved comments and final `active=0`/`nextWakeAt=null` passes. Six real-PostgreSQL controls pass on final commit `a60e5629e8`, including endpoint independence, source revocation, and external lock failure/recovery. No sleep, timeout increase, weakened assertion, or product scheduling/authorization change was added. - A local server typecheck first lacked the checkout's built Runner declarations. After the declared Runner TypeScript build, the server typecheck passes on `a60e5629e8`. Final-head remote verification is complete: 54 successful checks, 3 expected skips, no failures or pending checks, and successful `ci / verify` and `ci / e2e` aggregates. Greptile is 5/5 on this exact commit with no unresolved review threads. - Merge preparation: rebased onto master `b22ef22d56` and kept both test-import additions. Added the checkpoint fallback and capacity-failure contract to `doc/execution-semantics.md`. On rebased commit `6c5937fcae`, 47 focused checkpoint, transport, finalization and retained warm-projection tests passed across three files. The name filter skipped 1,040 unrelated cases. Map and whitespace checks passed. The final-head CI and Greptile results are recorded above. - `npx vitest run server/src/services/native-runtime/native-session-executor.test.ts server/src/services/native-runtime/native-workspace-finalizer.test.ts server/src/services/remote-workspace-snapshot-archive.test.ts` passes (1,152 tests). - `npx vitest run server/src/services/remote-workspace-transport.test.ts server/src/services/native-runtime/native-workspace-finalizer-failure.test.ts` passes (30 tests). - New tests: - Checkpoint, with real files: a full-disk snapshot leaves a partial copy. The provider then gets the exclusions, the session files are kept, the package caches are left out, and the partial copy is removed. - Checkpoint: a permission failure and a timeout still fail, remove the partial copy, and do not call the sync-out. - Transport: the real capture scripts run with an `ENOSPC` preload, for the helper install and for the archive write. Both report the full-disk message and leave no partial attempt. A message with extra text still collapses to `remote_workspace_operation_failed`. - Finalizer: the operation stderr names the full disk and keeps `workspace_sync_out_failed`. Other failures keep the bare code. - With the source change removed, the 3 checkpoint cases and the 2 transport ENOSPC cases fail. - `npx tsc --noEmit -p server/tsconfig.json` is clean. - Suites that refer to the changed error codes (`explicit-native-continuation`, `native-recovery-notice-work`, `native-run-finalizer-telemetry`, `native-workspace-export-retry`, `remote-workspaces.integration`) pass. In `chat-channels.integration`, 2 cases fail locally on unmodified `master` too (a bundled skill path outside the checkout, and a chat run timing assertion). They are not related to this change. - Manual recovery of the observed task: the retained sandbox was resumed with free space, file preservation was requested again, and the saved workspace contained the agent's commit. ## Risks - The fallback runs only after the snapshot fails with a full-disk diagnostic. The normal path does not change. - Provider exclusions are not anchored to the directory root. In the fallback, a nested entry that has an excluded name (for example a nested `tmp` or `config.toml` in a Codex home) is also left out. The fallback never copies more than the snapshot, so credentials stay out of the checkpoint. - The fallback depends on the provider honoring `exclude`, which the sync protocol requires. The Daytona and Kubernetes providers honor it. - A workspace export on a full disk still fails. This change only names the cause. A full fix needs one of these: free-space reserve for the export, removal of re-creatable caches before the export, or a capture that streams its archive. Each of these changes behavior, so this pull request does not include one. - No migration. ## Model Used - OpenAI Codex, GPT-6 family, assisted with merge preparation, conflict resolution, documentation, test synchronization and verification. The harness does not expose the exact serving model ID or context-window size. - Claude Opus 5.5 (`claude-opus-5-5`), Anthropic, in Claude Code with tool use (shell, file edits, test runs). It did the root-cause analysis from run events, tool output, and sandbox inspection, wrote the code and tests, and ran local verification. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> Co-authored-by: Dotta <bippadotta@protonmail.com> |
||
|
DottaandPaperclip
|
dc9f387edc |
fix(api): surface existing skill source import in discovery (#18757)
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - Agents discover supported API operations through the public OpenAPI
catalog.
> - Importing existing skill content must retain its source provenance.
> - A normal source-import query misses the applicable operation on its
first page because that operation has little search metadata.
> - This pull request describes the existing import operation and its
source boundaries.
> - The benefit is clearer discovery without changing search scoring or
import authority.
## Linked Issues or Issue Description
**What happened?**
`search_api` does not return `POST
/api/companies/{companyId}/skills/import` in the first ten results for
`import skill from source SKILL.md`. Source-management operations rank
ahead of the import operation. The import entry has only the summary
“Import a skill.”
**Expected behavior**
Ordinary source-import discovery should surface the existing operation.
Its documentation should explain that imported content retains
provenance and remains subject to company policy and source checks.
**Steps to reproduce**
Call `searchRunnerApi({ query: "import skill from source SKILL.md",
limit: 10 })`. Inspect the operation IDs on the first page. The new
regression sends this result through the real bounded MCP serializer.
Related: #18717 clarifies the separate `create_skill` tool. This change
covers the existing source-import HTTP operation. Neither PR is merged
by this work.
## What Changed
- Expand only the source-import OpenAPI summary and description.
Describe allowed local skill directories, supported remote sources,
retained provenance and existing access checks.
- Add a first-page discovery regression through the actual authenticated
MCP bridge and serializer. Keep non-import lookup, company/mode refusal
and denial forwarding controls.
- Update the feature and access maps with the test scope and its limits.
## Verification
- Parent regression: the controlled loopback MCP response lacked the
applicable import operation on the first page.
- Focused current checks: 2,060 tests pass in
`skill-import-discovery.test.ts` and the existing `runner-api.test.ts`
owner.
- `node scripts/check-maps.mjs` passes. This checks map structure only.
- The first head had one server test job cancelled after a runner
shutdown signal. The reason for the shutdown is unknown. The reviewed
successor removes only two test comments required by contribution rules;
its four focused controls pass. The remaining obsolete CI run was
cancelled after that successor was published. No same-head retry was
used.
- The new test is under `server/src/__tests__`, which the production
server typecheck excludes and the server Vitest suite includes.
- No local app, database, native agent or provider model was run. Normal
PR CI on `a4aaae8e320fb1d56cdbff9f376ec832a634a5da` passed with 48
successful jobs and one skipped job, including the required `ci /
verify` and `ci / e2e` checks. No completed agent-import journey or
reliability result is claimed.
- Greptile is 5/5 on the current head with no unresolved threads. The
optional Paperclip Review check remains queued. This is not merge
authorization.
- Keep this core PR open and unmerged. Do not enable auto-merge.
## Risks
The expanded metadata can change where this operation ranks for related
queries. Search weights, request schemas, HTTP routes, source parsing,
authorization and company skill policy are unchanged. The synthetic 403
control proves denial forwarding; it does not qualify the real policy or
source-import route. A live import still needs separate acceptance.
## Model Used
OpenAI `gpt-6.1-sol`, through Codex with reasoning, source review and
code execution tools. The exact context-window size is not exposed by
this environment. A separate agent reviewed the source and contract
before publication.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
canary/v2026.1103.0-canary.9
|
||
|
DottaandPaperclip
|
ef5b98594f |
Explain the human primary-agent control in API discovery (#18747)
## Thinking Path > - Paperclip lets a person manage a company of AI agents. > - A primary agent is that person's company preference. > - Only the person's board session can change it. > - Native API discovery explains that restriction but omits the real human control. > - An agent can then suggest a Chat-picker setting that does not exist. > - This change describes the existing profile crown and replacement confirmation. > - It adds no agent authority and no new tool. ## Linked Issues or Issue Description **What happened?** API discovery for the personal primary-agent preference explains board-only access. It does not explain the supported profile control. A retained native response gave a correct board API alternative but also suggested a primary setting in the Chat picker. The picker only opens conversations. **Expected behavior** The discovered operation should direct the person to the chosen agent's profile, the crown control labeled Set as my primary, and the replacement confirmation. It must keep the current person's board authority and the agent denial. **Steps to reproduce** 1. Search the native API for set primary agent. 2. Read the primary-agent operation's description. 3. Compare its guidance with the profile crown and the Chat picker. 4. The current description has no supported UI path. **Paperclip version or commit** Master `7b8aa60b6d` and the retained evaluated application have this metadata gap. The profile control was checked in both sources. A scoped open-PR search found no matching primary-discovery fix. ## What Changed - Add general profile/crown/confirmation guidance to the existing GET and PUT operation description. - Clarify that the Chat picker does not change the preference. - Keep all other OpenAPI bytes, authority, schemas and handlers unchanged. - Update the two existing capability maps. ## Verification - The actual API search result passes through the existing authenticated OpenCode MCP bridge and its bounded JSON serializer. The parent fails because the delivered description has no profile control. The new result retains the full guidance, board-only actor and original schema. - The existing API client keeps its agent bearer/run identity and propagates a mocked board403. This is client propagation proof, not a new real-route claim. - Initial selected controls:997PASS,1061 intentionally unselected. After relocating the same test into the existing server `__tests__` directory, both complete owning files pass:2058 controls. These are static/protocol controls, not native journeys. - The first normal CI head failed with TS6059 because the test imported Runner source outside the server build root. The successor moves only that test to the already-excluded test directory and adjusts four relative imports. An independent source peer confirms identical test bodies; `--listFilesOnly` confirms build-program exclusion. The failed head is retained. Corrected-head CI41599277833 passed48 jobs/1skip on attempt2 at08:17:23UTC. Attempt1 had46passing jobs but a server4 runner-shutdown cancellation; exactly that job and dependent verify were rerun once on the unchanged head. - Maps and diff checks pass. The initial loopback listen EPERM was a setup failure; the controlled transport test then reached the actual assertion RED. - No Paperclip app, database, browser, model, external provider or real credential ran locally. Full typecheck/build passed in corrected-head normal PR CI. Fresh Greptile5/5 is tied to that exact head, with zero unresolved threads. ## Risks The added description can affect API search ranking. It changes guidance only. It does not impersonate a person, change permissions or automatically select an agent. Future model behavior remains unqualified. The prior mixed failed journey is retained; its separate versioned grading work is not part of this PR. Keep this core PR unmerged. ## Model Used OpenAI `gpt-6.1-sol` with reasoning, source inspection and tool use. Context-window size was not exposed by the host. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
a2a6129ed5 |
Allow a new human intent to request tool review after a decline (#18742)
## Thinking Path > - Paperclip manages work for companies of AI agents. > - App writes can require a human review for each invocation. > - Decline and Always allow are separate controls. > - The gateway currently reuses a declined action across later human requests with the same arguments. > - A new request can therefore receive the old decline instead of a new review. > - This change uses the saved run and human comment to distinguish a new intent. > - The new intent still follows the current policy and approval gates. ## Linked Issues or Issue Description **What happened?** After a person declines an Ask first app write, a later request from that person for the same write receives `action_declined`. The gateway matches company, task, agent, tool and arguments across runs. It does not distinguish the later human request. **Expected behavior** A distinct later request from the same responsible person can open a new Ask first review. It must not execute the write or reuse approval. A retry of the declined intent, including its rejection continuation, must remain declined. **Steps to reproduce** 1. Ask an agent to perform an app write that requires approval. 2. Decline the proposed action. 3. Send a new request for the same action. 4. Observe that the gateway reuses the old rejection and creates no new pending review. **Paperclip version or commit** The behavior is present on master `7b8aa60b6d` and in a retained native journey. The focused parent control reproduces the matching error. No related open fix was found in the scoped PR search. ## What Changed - Only a rejected match can be released for a later human intent. - Require the active run, company, task, agent and responsible person to match the saved direct-human comment. Reject derived, deleted, untrusted or system-authored input and invalid chronology. - Preserve all other replay states, signed argument checks, current policy, unresolved-effect guards, approval and execution owners. - Document the distinction between decline and a standing policy change. ## Verification - Actual matching-function control: parent fails; this change passes. The control uses the unchanged owner body with synthetic database rows. It does not claim real PostgreSQL or provider verification. - 52 focused pure controls pass. These cover same-intent retries, rejection continuations, foreign scope and actor, invalid authorship, chronology and other action states. - TypeScript 7.0.2 checks pass for the new helper and its owning tests. - `node scripts/check-maps.mjs` and `git diff --check` pass. - The existing gateway service suite has a new signed-action integration test. It covers ordinary and reopened human requests. It requires the old rejection plus one new pending request, a stable retry, and no completed call. Each leaf prints one fixed success label after its assertions so CI can distinguish execution from a platform skip. No local PostgreSQL, app, model or provider execution was run. Current-head normal CI41594912212 completed with48 successful jobs and1 expected skip. Both real PostgreSQL leaves executed and emitted their post-assertion markers; the owning job passed. Full typecheck/build checks passed in CI. ## Risks This adds one indexed run/comment read only after a rejected matching action. Missing identity or ambiguous provenance keeps the old rejection. Legacy and external comments without direct-human provenance do not reset it. A new human intent is not consent to execute. This core PR remains unmerged. Candidate native journey qualification is separate; the original failed journey is retained. ## Model Used OpenAI `gpt-6.1-sol` with reasoning, source inspection and tool use. The host did not expose the context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #123` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
8bd45b5b30 |
fix(ui): expose redacted arguments in pending tool reviews (#18734)
## Thinking Path
> - Paperclip helps people manage AI agents and review their actions.
> - Ask first tool actions require a person to inspect the proposed
operation.
> - The task and Apps review surfaces share the pending tool-action
card.
> - That card shows a short preview but omits the available redacted
argument body.
> - A valid preview can omit important text, such as the date at the end
of a proposal.
> - This pull request adds optional argument inspection to the shared
card.
> - The person can inspect the proposed content before using the
existing decision controls.
## Linked Issues or Issue Description
Refs: #16540.
Related PRs: #16571 exposes reviewed arguments on formal board approvals
and scrolls the task review into view. #16716 routes tool actions to
their required formal board approvals. This change addresses the
separate shared pending tool-action card used by the task takeover and
linked Apps reviews. It does not close the other reported paths in
#16540.
## What Changed
- Add the existing Arguments and audit hash disclosure pattern to the
shared pending card.
- Show only the bounded argument summary already redacted by the server
and its audit hash. Show an unavailable message for missing, malformed,
or oversized fields.
- Keep the current approval, decline, and remembered-scope callbacks.
- Add a long-proposal story and update the tool-review feature recipe.
- Add React coverage for the real task review transition, the linked
Apps review, plain text rendering, invalid fields, and unchanged
decision calls.
## Verification
- The new task-review test failed on the parent because the disclosure
was absent. It opens Review request, inspects the proposal tail and
hash, then approves the same interaction.
- Oversized-whitespace controls fail on the previous guard ordering
because trimming runs before the length bound. The guard now checks
length first for both fields.
- Passed: 117 React tests across TaskChatInteractionCard,
IssueThreadInteractionCard, and ReviewQueueCard. The tests use the
repository UI config with the existing installed Vitest 5.0.3 dependency
tree.
- Passed: token gates, map checks, Storybook catalog checks, and git
diff checks.
- Local full UI typecheck did not pass. The linked dependency tree lacks
adapter and json-schema-to-ts declarations and has stale shared exports.
It reports no diagnostics in the changed files. This is not a successful
full typecheck.
- Full workspace build, browser, and provider checks were not run
locally. Current head
|
||
|
DottaandPaperclip
|
07721e33f7 |
fix(runner): distinguish new skill authoring from source imports (#18717)
## Thinking Path > - Paperclip lets people manage agents and shared company skills. > - Agents can author new skills or add existing skills from approved sources. > - Source import and catalog install retain the source used by update, audit and reset flows. > - The create_skill tool described saving complete Markdown, without explaining this choice. > - An agent can therefore read an existing source and create an unrelated managed copy. > - This PR explains the distinction in the existing tool contract and tests its actual model-facing projections. ## Linked Issues or Issue Description **What existing behavior does this improve?** The model-facing description of create_skill and the choice between authoring and source import. **Current behavior** The tool says it saves a complete single-file skill. It does not distinguish new authored content from adding an existing skill. The Company Skills reference already directs external sources to import and catalog skills to install. **Proposed behavior** The tool explains that it saves newly authored content. It directs existing-source work to the existing import or install APIs through search_api and call_api. It retains the existing assignment and retry guidance. **Reason and benefit** A managed copy does not retain the original source locator. Clear guidance helps an agent choose the source-aware operation without adding a tool or changing permissions. This is a guidance repair. It does not claim that the import backend failed. ## What Changed - Clarify the canonical documentation, live descriptor and scenario descriptor for create_skill. - Test the actual discovered contract and provider projection, including API claim and company-boundary negatives. - Refresh generated descriptions and the seeded catalog digest and manifest entry. ## Verification - A focused regression fails on the parent because the actual discovered contract lacks the authoring versus import distinction. - Three focused pure suites pass: 48 tests. They include existing skill creation, retry, schema and authorization controls. - The source-owned serializers and protocol fixture validator regenerated the inventories without an app or native build. - Both generated inventories change only create_skill description. Arguments, required claims, modes, idempotency and all other tools remain byte-equivalent. - Current head `bca441726a6cba80e6e7da0321db59e27183c5b8` passes required `ci / verify` and `ci / e2e`; the full main CI run has 49 successful jobs. Windows, Docker, security and review checks also pass or are intentionally skipped. Greptile and the final Paperclip review are 5/5, with zero unresolved threads. This PR remains deliberately unmerged for the separately reviewed evaluation candidate. - Original head CI preserved: one unrelated UI clipboard-confirmation assertion failed after the clipboard call passed; 47 other executions succeeded and the final verify check failed as its dependent. The two UI owners are unchanged from the base. Exact React scheduler/timer timing is unknown. No unchanged-head retry was used before the necessary master conflict sync. - The conflict sync preserves incoming task-assignment guidance and regenerates metadata. It is not a fix for the UI test failure. The synchronized current head passes CI and review. No same-head recovery was consumed. ## Risks The description changes model context and the catalog digest. It may affect tool choice, but actual same-request and same-model acceptance is still pending. The change does not alter authorization, API behavior or grading. Local checks did not run the app, database, native runner, full graph typecheck or build. Intermediate test-fixture claim and placement mistakes were corrected; the final focused run passes. Do not treat these pure controls as a feature pass. ## Model Used OpenAI GPT-6.1 Sol (`gpt-6.1-sol`) through Codex, with reasoning, source inspection and tool execution. The host does not report a context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs; narrow create_skill/import and skill-import-guidance searches found no duplicate. - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
|
a497bae83a |
test(e2e): verify Cloud workspace access isolation (#18655)
## Thinking Path > - Paperclip keeps agent work inside company boundaries. > - Cloud users switch workspaces through the tenant account menu. > - A company-bound credential must work inside its permitted company. > - A missing resource cannot prove an access denial. > - This journey uses real controls in two registered test workspaces. > - It checks reads and task creation in both directions and revokes both keys. ## Linked Issues or Issue Description Refs: #18728, #18729, #18652. **What existing behavior does this improve?** Product E2E coverage for Cloud workspace selection and company-bound agent API access. **Current behavior** The journey catalog lacks a two-workspace boundary check with real allowed and denied controls. **Proposed behavior** Use the visible Cloud menu to switch workspaces and return. Read a real task and create one task in each credential's own company. Require real foreign reads and the same task-creation operation to return 401 or 403 in both directions. Verify that no foreign task was created and that protected content was not returned. Revoke both credentials and require their earlier allowed reads to return 401. ## What Changed - Add a browser journey and a narrow target-provider fixture contract. - Require independent owner reads, exact agent authorship, zero-to-one task creation, stable run history, no unexpected work and no foreign creation. - Validate observed workspace entry links against the exact registered targets. - Add wrong, missing, leaked, foreign and invalid-entry evidence controls. Saved passes reject collected runs or nonempty model usage. - Register all seven required checks, definition-hash inputs, feature associations. The feature-map documents are unchanged, as requested. - Base the final tree on verified repository master `7b8aa60b6d79b6cd9e4d46ca1369515c33e2d5d4` plus only 15 access journey and catalog files. Preserve all other master files and modes. The held completion journey is outside this PR's diff. ## Verification - 851 catalog, journey, completion-helper, artifact and browser-action support tests passed across 40 files. - E2E TypeScript, credential-free discovery, feature-map validation, full workspace recursive typecheck, full build and `git diff --check` passed. - The broad local `pnpm test:run` attempt stopped in chat integration shard 1 before any assertion. Embedded Postgres initialization exited with code null. No chat tests or later unit shards ran. The startup cause is unresolved, the failure is retained, and the command was not retried. The database/chat/test-runner source is unchanged from the fixed master snapshot. The separate hosted full CI result is recorded below. - The first local E2E typecheck read stale compiled plugin SDK declarations after the merge. The SDK source already declared the field. Rebuilding its declarations cleared the diagnostic without a source edit. - The complete tree comparison proves exactly 17 changed paths before the documentation removal against the fixed master snapshot. Every other master file and mode is preserved. All four access flow, contract and test blobs match the previously passing source `fab9a1fb2fefb41a7248e4fc1aa239d77455f5e2` byte for byte. - The historical staging attempt on source `fab9a1fb2fefb41a7248e4fc1aa239d77455f5e2`, the earlier Core42 deployment and the earlier adapter passed all seven checks: owner resource controls, visible Cloud switch and return, scoped data, both restricted-agent positive controls, cross-workspace read/task-creation denials, credential revocation and no agent work. Both credentials were revoked, owned credential agents had terminal receipts, and evidence cleanup was complete. No agent or model work was recorded. - The earlier attempt failed before journey checks while reading the borrowed fixture agent's historical details. That failed result remains preserved. The provider then supplied a separately owned primary identity. The public history assertions remain unchanged. - The strict root catalog expectation first failed because it omitted the new access case. Adding only that expected entry made the test pass. The ordered list and default-selection assertions remain unchanged. - The first combined-head full CI run failed two unchanged attention-retention tests at their original 15-second limits. The catalog fix does not change those tests or their timing. The initial failures are retained, their cause is unresolved, and that failed run was not retried. - Normal full CI passed on the new commit in [run 41595755161](https://github.com/paperclipai/paperclip/actions/runs/41595755161), attempt 1: 48 successful jobs and one intended skip. The two unchanged attention-retention cases passed at about 12 seconds; the earlier timeout cause is still unknown. The fresh review on commit `4c0ea682444738ad6fa17017a6afd25a49037559` is 5/5 with no open findings. The fresh staging result is recorded below. This PR changes no daily plan, private configuration, model or live workspace. - Fresh staging qualification on commit `4c0ea682444738ad6fa17017a6afd25a49037559` passed all seven checks and all five phases. Both ordinary same-workspace task creations returned 201. Both cross-workspace task creations returned 401. Both credential agents reached terminal state. Both keys were revoked, and both revoked-key reads returned 401. The manifest records complete cleanup and no active operations. No model or task run was recorded. All three workflow jobs passed. - The coordinator verified TestOps case 719 and result 3761 as passed, then closed launch 1987. Its report shows model none and the current qualification run. The automatic Slack report recorded one pass and zero failures. - The original normalized result was passed. The first evidence collector left its own qualification false because its allowlist omitted the standard `dashboard.html` artifact. The adapter already declared that file, and the public launcher generated it. An exact-object read matched the 118,791-byte object, ETag and SHA256 archive receipt. This completed the three-artifact inventory. The original result, collector proof and seal remain unchanged. No HTML was executed or rendered, and no test was rerun or regraded. | Fresh staging provenance | Verified value | | --- | --- | | Public commit | `4c0ea682444738ad6fa17017a6afd25a49037559` | | Qualification workflow | `41597774599`, attempt 1 | | Serving Core commit | `09162fe893b464dd2ccdafd14bd3ef1611202a86` | | Target adapter commit | `7367cf6fd0ee1cf52c2651a8a6671e087728a9c1` | | Actual workflow bootstrap commit | `55351ffba2704a4e13c1ead3e50665acd51aef87` | | Exact 117-file QA contract SHA256 | `14d95c29059c82563ab0e1d65a01aa4ac6c4f60f77a236803cf06ad5d9e7d27b` | | Provider-pack configuration version | `9` | - At the user’s request, follow-up commit `7e0dbc4bdba8a25ff6486648b0ca0d0db2616a4c` removes both feature-map additions. The final diff contains 15 paths and no `maps/feature-map/` changes. An entire-tree comparison against the QA-passed commit proves these two Markdown files are the only changes; every executable file, test, fixture, registration and file mode is unchanged. `node scripts/check-maps.mjs` and `git diff --check` pass after the removal. The live QA run above was not repeated for this documentation-only change. Fresh current-head full CI passed in [run 41633385961](https://github.com/paperclipai/paperclip/actions/runs/41633385961), attempt 1. The current-head Greptile review is 5/5 with zero unresolved threads. The normal merge preserves every newer master file outside the 15 PR paths and keeps all 15 PR files byte-for-byte intact. ## Risks The private target provider must register two exact test workspaces and preserve their separate ownership and cleanup. The case verifies company-bound agent credentials. It does not certify human membership roles. It tests task creation because ordinary agent task edits require a real execution context. No execution context is forged and no product authorization guard changes. The local broad unit run could not start embedded Postgres, so it does not establish a full local unit pass. The earlier hosted attention-retention timeouts also remain unexplained. Passing a later run would not prove the source of those timeouts. The four case files retain their historical passing bytes. The fresh qualification verifies the executable test code retained in this PR against the deployed staging stack. Daily admission remains a coordinator action after the normal source merge. The case does not cover future deployments without another run. ## Model Used OpenAI GPT-6 through Codex with reasoning and code execution. The runtime does not expose a more specific model identifier or context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> Co-authored-by: Codex <noreply@openai.com> |
||
|
DottaandPaperclip
|
b22ef22d56 |
fix(runner): show provider failure details on tasks and runs (#18622)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Native Runner starts providers and records their run outcomes. > - Provider failures can contain useful details below a generic protocol error. > - Those details were lost at the sidecar, command receipt, and run projection boundaries. > - A blocked task therefore showed a generic failure even when the provider gave a clear reason. > - This pull request carries a bounded, sanitized provider failure through those boundaries. > - Operators can read the actual cause on the task and run pages. Existing recovery policy still controls retries. ## Linked Issues or Issue Description Refs #17543 for the related closed JSON-RPC classification work. Refs #17272 for provider recovery in the legacy OpenCode adapter. This change adds display evidence for native Runner providers and keeps recovery policy separate. **What happened?** A native Claude session returned an HTTP 429 message below an ACP `Internal error`. The task and run showed only a generic startup failure. Operators had to inspect the provider response to learn that the account was rate limited. **Expected behavior** The task and run show the sanitized provider message, provider name, startup or turn phase, and available HTTP status. Unknown categories keep the message. Display evidence does not grant a retry or replace a recovery decision. **Steps to reproduce** 1. Start a native Claude run whose account is rate limited during model selection, or return a structured provider failure from a Codex or OpenCode turn. 2. Let the run fail and open its task page. 3. Before this change, the page shows a generic failure. With this change, the page shows the provider explanation. The run page shows the same explanation after reload. **Paperclip version or commit** Source checkout based on `f8a888a072`. ## What Changed - Add one JSON Schema for `ProviderFailure` and generate matching Runner and shared TypeScript readers. Keep the Runner package independent of the app workspace at runtime. - Normalize known ACPX, Codex, and OpenCode error fields. Preserve unknown messages. Redact known credential values and credential patterns before applying the 4,000-character limit. Do not copy response bodies, headers, stacks, or prompts. - Carry startup failures through the sidecar, Rust Runner, durable command receipts, and private transport error details. Keep provider text out of generic transport error strings and Sentry failure messages. - Keep turn failures in authenticated run events. Recover the same evidence from persisted events during server finalization. Store the cause in the company-scoped run result. - Show one shared notice on failed tasks and run details. Keep existing retry permissions, scheduled retry state, and resolved-task behavior. - Reapply registered run-secret and current-user redaction to task display data. Batch secret lookups for only the visible provider causes. - Document the contract and update the capability maps. Add regressions for nested Claude 429 errors, OpenAI quota and rate limits, OpenCode errors, redaction, durable receipt replay, restart finalization, and UI rendering. ## Verification - `pnpm -r typecheck` passed. `pnpm build` passed. All GitHub Actions build, test, package, and validation jobs passed on `d49c79ab4a`. The security checks passed. Greptile gave this head 5/5 with no open review threads. Paperclip Review remains pending. Its review worker exhausted six workspace recovery attempts with export_interrupted. A fresh review request was accepted and is waiting for that saved workspace. The ordinary Retry action rejected reuse of the stopped request. This is an external review-worker blocker; no check was overridden. The duplicate local `pnpm test:run` was stopped after the CI test shards passed. - Focused server settlement, provider evidence, and restart tests: 93 passed. Runner admission and normalization tests: 103 passed. Focused Runner wire and terminal tests: 29 passed. The real Rust/sidecar command receipt replay test passed. - The new projection test failed before implementation because the task had no provider cause. The startup producer regression also failed before extraction moved ahead of the generic error-code replacement. Both pass with this change. - Live check: a rebuilt local sidecar received an actual Claude HTTP 429 during startup. The new cause kept the full provider message. A synthetic task in an isolated local app then reached the same provider error through the native Runner path. The dedicated run result, task API, task banner, and run detail all showed the same message after reload. The later reporting fix keeps run.error generic; a regression proves the display cause still survives restart and stays out of Sentry capture. - The first isolated app attempt failed earlier in `admission_credential`. An ordinary manual Retry reached the provider and produced the 429 above. This PR does not claim to fix that separate admission failure. An initial parallel local database test setup also failed; the serial restart suite passed. - Security review found that copying display text into run.error could forward it to Sentry. A regression failed with that behavior. The fix keeps the cause in its display field. All 38 settlement/restart tests and the server typecheck pass. - All 29 protocol contract tests pass, including bounds and extra-field rejection. An earlier full Rust core run passed. The later full run hit one 5-second receipt-limit timing assertion; its isolated rerun passed. The full Linux and Windows Rust CI jobs passed on this head. - Review regressions pass for explicit Codex driver identity, native Rust startup and resume receipt propagation, oversized OpenCode/ACPX errors, task read-time registered-secret redaction, and the standalone source closure. The focused Runner run passed 103 tests; projection/restart tests passed 16. - Read-time redaction edge cases passed: a missing decryption key omits only the unsafe optional explanation; expanded redacted messages remain visible within the Unicode character limit. All 14 task projection tests and server typecheck pass. - The package Build check found an old Pi assertion against Error.message. The probe now checks the exact closed classification in the authenticated receipt and requires the new provider cause. Both polling transport cases pass. A local credential-free Pi probe stopped at admission_installation because this checkout lacks the Pi distribution; the installed Linux package probe passed in CI on this head. - UI coverage includes startup and turn text, unknown categories, literal rendering of provider text, read-only retry behavior, and resolved-task suppression. Token gates and map checks passed. - The live check covers Claude startup on macOS. Codex, OpenCode, terminal failures, and restart paths have deterministic regression coverage. The change is not deployed. ## Risks - Provider messages are untrusted text. The boundary selects specific fields, redacts credential patterns and known environment secret values, bounds the message, and renders plain text. New provider-specific secret formats may need added redaction patterns. - Optional fields preserve compatibility with older producers and stored runs. Old runs without the field keep their current fallback text. No database migration is required. - The cause stays in company-scoped run data. This change does not send provider text to first-party Telemetry or add it to OpenTelemetry span attributes. - Normalized categories and retryability are display evidence. Existing recovery classifiers, retry limits, ownership checks, and approval gates keep their authority. ## Model Used - OpenAI GPT-6 through Codex, with reasoning, repository tools, code execution, and browser verification. The host does not expose the exact serving variant or context window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
718336eb04 |
fix(native): refresh prior warm controller projections (#18718)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Native remote sessions keep their controller state in a database checkpoint. > - An admission can also leave a full local copy of a prior run's state. > - The prior owner can commit more state without changing its run ID. > - Warm admission skipped a refresh when the local and saved run IDs matched. > - This pull request refreshes that prior-run copy when no retained warm owner exists. > - The next turn can use the committed state while existing ownership checks remain effective. ## Linked Issues or Issue Description **What happened?** A prior warm run can complete after an earlier admission reads its controller checkpoint. The local copy can still contain a pending command or lack the final terminal event. A later admission sees the same run ID and skips restoration. It can then hold work against that old copy even though the database checkpoint has settled. **Expected behavior** Read the canonical prior-run checkpoint before reuse when no retained in-memory owner exists. Keep current-run state with its controller. Keep unresolved canonical work and invalid checkpoint bindings blocked. **Steps to reproduce** 1. Create a full prior-run local projection while a command is pending. 2. Complete that command and record the terminal in the canonical checkpoint under the same run ID. 3. Start an ordinary remote warm continuation without a retained in-memory owner. 4. Before this change, the local command remains pending. With this change, the completed command and terminal replace the stale copy. **Paperclip version or commit** Reproduced on master `36df8933f0` before changing production code. Related work: #18018 restored per-turn checkpoints; #18343 added controller replacement recovery; #18554 restored missing remote suspension receipts. This change handles a full prior-run projection whose run ID already matches the saved supervisor. It does not change the settlement guard or MCP compatibility policy. ## What Changed - Refresh a prior-run warm controller projection even when its run ID has not changed. - Keep current-run and retained warm owner authority on the existing path. - Add five executor cases for a settled source, a pending canonical command, a current run, legacy file-backed state, and an invalid checkpoint binding. - Add an opt-in Linux continuity case with a stale full local projection and two follow-up turns. - Update execution documentation and the paired session verification recipes. ## Verification - RED before the production edit: the executor fixture retained `pending` in its local command record when the canonical record was `completed`. - GREEN on the final commit: 48 focused tests passed across executor restoration, retained warm owner rotation, per-turn restoration, and remote settlement guards. - `pnpm --filter @paperclipai/server typecheck` passed. The server TypeScript check also passed after the integration fixture was added. - `node scripts/check-maps.mjs` and `git diff --check` passed. - The unit fixture uses the real durable controller serializer and the real executor entry. It stops at current tool authority before provider/backend creation. It proves restoration and guard behavior, not full task acceptance. - Actual protocol RED/GREEN: the same opt-in continuity fixture uses real PostgreSQL, TLS, the current Linux Rust Runner, and a deterministic fake Codex backend. With only the executor restored to its parent revision, continuation fails with `prior_owner_terminal_state_indeterminate` in the settlement guard after one provider turn (20.16 seconds). The PR source completes all three turns with one provider thread and final suspension (34.77 seconds). Hard recovery permits one provider-process replacement; the healthy final follow-up starts no additional process. Both runs use the same Runner binary and owned container. - The exact PR executor source was restored after the baseline comparison, its Git blob was verified, and the worktree is clean. The disposable local container was removed. This fixture proves the controlled stale-projection recovery path; it does not establish a live model or Daytona API result. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on the current commit, with no unresolved review threads. No original task was retried and no staging sandbox lifecycle was changed. ## Risks - Admission without a retained owner performs one additional scoped checkpoint read when a full prior-run projection already exists. - The canonical checkpoint must still pass its company, task, session and record validation. A pending canonical command stays held. Current-run and retained owner state are not replaced. - The regression proves a stale local projection defect. It does not establish that a historical incident had those same local bytes. ## Model Used OpenAI Codex, GPT-6 family. The harness does not expose the exact serving model ID or context-window size. Capabilities: repository inspection, code generation, tools, and test execution. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
db8a273562 |
fix(native): publish fresh retry sessions for later turns (#18713)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Native tasks save a provider session for later comments and monitor checks. > - The goal projection also records the run that owns the session event sequence. > - A fresh retry uses a new native session, but the projection accepted only the old session. > - The accepted checkpoint then rejected session publication because the goal source still named the old run. > - This pull request lets the current task executor advance that source after a terminal predecessor with no unfinished goal. > - Later turns can then use the fresh retry's saved session. ## Linked Issues or Issue Description Related: #12669, #18343. I searched open and closed PRs for source epochs and native task-session publication. I found no duplicate repair. **What happened?** A successful fresh native retry could retain its predecessor's task-session pointer. A later comment or monitor selected the old provider session. The goal source transition required both runs to have the same native session. The terminal checkpoint then saved a `suppressed_newer` publication result. **Expected behavior** The current task executor can publish its accepted fresh session. A later turn resumes that session. Delayed events cannot restore the old source or revive a cleared goal. **Steps to reproduce** 1. Complete a native turn and retain its task-session row. 2. Admit a fresh session on the same task without deleting the row. 3. Complete that turn, then send an ordinary follow-up. 4. Before this change, the saved pointer can still name the first session. **Paperclip version or commit** Reproduced on source commit `36df8933f0` with a PostgreSQL regression before the production change. **Deployment mode** Source checkout. The affected producer runs for local and remote native sessions. ## What Changed - Permit a different native session to advance the goal source only when its run holds the task execution lock and assignee, and the task is in progress. - Keep the terminal predecessor, native runtime, runner identity, absent or completed goal, source sequence, and conversation reset checks. - Test accepted session publication, a fresh heartbeat followed by an ordinary resumed turn, stale events, current-owner loss, and unfinished or unreadable goals. - Add an opt-in live Codex file-persistence check. Update execution documentation and both recovery maps. ## Verification - Red: the new PostgreSQL publication regression failed before the producer change. The saved session, display ID and goal source still named the predecessor. - Green: 72 distinct tests passed across the final focused runs of `runner-goals.test.ts`, `native-task-session-publication.test.ts`, `native-session-resumption.test.ts`, and `remote-warm-session-recovery.test.ts`. - The assembled heartbeat test admits three turns, opens two provider sessions, resumes the second session once, and stores three results. It uses a deterministic provider backend. - `node scripts/check-maps.mjs` and `git diff --check` passed. - Real local Codex acceptance passed on macOS arm64 with `gpt-6.1-sol` and explicit warm mode. Three heartbeat turns saved three exact file lines. The fresh second turn published its new normalized and provider session IDs. The ordinary third turn retained both IDs. The fixture closed its warm owners and removed its private credential copy and workspace. The production code was at `450e15e`; the final test-only commit selects explicit warm mode and uses the existing owned-home cleanup helper. The production change is identical in the final head. - The earlier per-turn fixture preserved all three file steps and selected the correct normalized session, but provider-thread reuse failed. `run.attach` reported that the Codex provider had pending events; the safe fallback used another provider thread. This remains a separate failure. The first fixture cleanup also failed on immutable instruction assets. The owned-home helper fixed cleanup, and the second per-turn attempt exposed the provider identity assertion directly. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on the current commit, with no unresolved review threads. Broad local workspace checks were not duplicated; the focused regressions and actual provider acceptance above were run locally. No merge, deployment, or recovery of an original task is included. ## Risks - Source epoch changes affect saved goal state. The fresh-session path requires current task authority and refuses unfinished or unreadable saved goals. - The change starts no provider work and changes no prior-owner recovery guard. Unknown external outcomes retain their reconciliation hold. - A prior session remains held when it lacks the required authority or state proof. - Per-turn provider-thread reuse remains an open coverage gap described above. This change does not alter that attachment guard. ## Model Used OpenAI Codex, GPT-6 family. The exact model ID and context window are unavailable in this session. The agent used code inspection, tools, and executable tests. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
89e3c23062 |
fix(daytona): stream files when export copies cannot fit (#18712)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - A native task must save its files before finalization can release its writer. > - Remote workspace capture can produce an archive that fits on the sandbox disk. > - Daytona then makes another copy before it downloads that archive. > - The second copy can run out of space and leave completed work behind a save hold. > - This pull request uses the existing bounded stream when that copy cannot fit. > - The benefit is a saved archive without a second archive-sized sandbox allocation. ## Linked Issues or Issue Description Related: #18644 reduced capture scratch space. #18172 added the disk-free directory export. #18705 preserves provider checkpoints on a full disk and reports capture failures. This change handles the later single-file download copy. It does not depend on #18705 and does not replace its checkpoint fix. **What happened?** A native provider turn returned a successful result and captured its workspace archive. The Daytona outbound file guard then failed with exit 43 and `No space left on device` while it copied the archive to a reserved download file. Finalization and file-preservation retries could not pass this boundary. The workspace kept its writer hold. **Expected behavior** Download an existing archive through a confined sandbox process when its second copy cannot fit. Keep the source and all finalization holds until the normal save commits. **Steps to reproduce** 1. Capture a workspace archive in a Daytona sandbox. 2. Leave less free space than the archive needs for another copy. 3. Finalize the task. The outbound single-file guard fails before download. **Paperclip version or commit** Reproduced against `6d72243d8a`. The incident used the native sandbox runtime. This branch is based on current master. **Deployment mode** Cloud with native Paperclip Runner execution and remote workspace storage. ## What Changed - Remove partial outbound file copies after a failed snapshot command. Only copy capacity failures select the stream path. Permission, confinement, and transport errors still fail. - Use the existing acknowledged terminal stream to read each file as the sandbox user. Recheck canonical confinement and regular-file status. Use a no-follow open. Keep the sandbox disk limit, host free-space reserve, shared transfer deadline, and bounded teardown. - Keep host files in private temporary files until the complete batch has transfer receipts. Never pass a mutable live source to the privileged download API. - Add real-file and real-terminal regressions, a disposable Daytona acceptance case, and workspace verification notes. ## Verification - Red before repair: a real 5 MiB copy under a controlled file-size limit failed with exit 43. The same test passes after repair and compares every source and destination byte. - The ten capacity cases pass. They cover producer failure after output, source replacement, stream size limits, permission errors, timeout teardown, cleanup, and a successful retry. A two-file case completes the first stream and fails the second. Both old host files stay intact, both source files survive, and retry saves both files. An isolated early-publication mutation makes that test fail. - The initial complete Daytona package suite passes: 537 tests passed, 8 skipped. After the test-only review follow-up, all 60 focused transfer, stream, and diagnostic tests pass. The package typecheck also passes. - Real Daytona acceptance passes: an owned Ubuntu 24.04 sandbox stops and resumes, a copy to `/dev/full` reports ENOSPC, production export saves the exact archive hash, and a second save after reconnect passes. The test deletes only its disposable sandbox. This is provider-boundary acceptance, not a UI journey. - Assembled finalization acceptance passes with real Postgres and a live Linux x86_64 Daytona sandbox. The production control-plane port accepts one fixture result. The production workspace finalizer encounters the failed copy, saves the files, and commits the task as `done`. A new database client reads the committed result and exact file hash. The workspace has no writer or running save, and there is still one run and one result. No model turn runs. The target resolver is bound to the live provider fixture. Object storage uses the production encryption and local-disk service with an S3 admission label; this does not prove an S3 network transfer or a UI journey. - The live tests used the fixed transfer source with SHA-256 `857b3d8c9ad75170707d95277c17c051ce4582fa79a9c38331d51d2800b36af5`. That source is identical in this PR head. - `pnpm -r typecheck` passes locally. The duplicate full local test run passed all six chat shards before it was stopped during the server lane in favor of completed current-head CI; a duplicate local build was skipped. Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. - Greptile is 5/5 on the current commit with no unresolved review threads. - `node scripts/check-maps.mjs` and `git diff --check` pass. Map checks prove structure only. ## Risks - The fallback requires the Bash and GNU `dd`/`base64` tools already required by the existing stream path. - Terminal streaming is slower than the normal bulk download. It runs only after a copy capacity failure. - The workspace capture itself still needs sandbox space. This change does not fix a full disk before an archive exists. - The test injects a failed copy sink. It does not exhaust a shared host disk. No original task was retried, and no deployment or merge is part of this PR. - No schema, API permission, writer ownership, or retry-budget change. ## Model Used OpenAI GPT-6 through Codex, with reasoning, repository inspection, code editing, shell tools, browser diagnostics, and automated tests. The exact serving variant and context-window size are not exposed in this session. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
1bc3ed6dd5 |
fix(opencode): keep native tool calls pending until Runner settlement (#18711)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Native OpenCode sends semantic tool calls through the durable Runner protocol. > - The Runner retains a call until its result is known. > - OpenCode and its HTTP bridge each stopped waiting after 30 seconds. > - A slow response could leave a call pending after the provider had moved on. > - A later run then failed the pending-call guard during attachment. > - This pull request keeps native calls under their Runner owner until settlement or cancellation. > - The benefit is one delivered result and a safe next run. ## Linked Issues or Issue Description Related: #18343 applies the durable wait policy to native ACPX calls. This change applies that policy to native OpenCode. **What happened?** An assigned tool call took more than 30 seconds. OpenCode reported an MCP timeout while the Runner still retained the request. The source turn could finish, but suspension failed. A later `run.attach` failed because provider tool calls were still pending. **Expected behavior** The native proxy must wait for the authoritative result under the Runner owner. Stop, bridge close, and lease expiry must still end the wait. Unknown effects must still block attachment. **Steps to reproduce** 1. Start a native OpenCode run with an assigned semantic tool. 2. Hold the controller's result for 31 seconds. 3. Observe the provider's MCP result and attach a later run to the session. **Paperclip version or commit** Reproduced against `6d72243d8af0362d73fbcb5331fd42a384800d4e`. The branch is based on `36df8933f0`. ## What Changed - The native OpenCode proxy enables a driver capability for durable tool requests. - That capability removes the bridge's separate request timer and sets OpenCode's fixed MCP client ceiling to one hour, consistent with the native ACPX sidecar. - Ordinary driver calls keep their 30-second deadline. Assigned external MCP connections keep their existing budget. - Cancellation, result replay, tool authorization, and the pending-call attachment guard keep their existing contracts. - The native regression holds a result for 31 seconds, then completes a second run with the same Runner process and no pending provider calls. - The deterministic provider fixture honors the emitted MCP budget and retains the initial completion contract across ordinary follow-ups. - Execution semantics and the session verification maps describe the wait and its safety checks. ## Verification - Red before production edits: the bridge aborted a durable request at the ordinary 30-second boundary. The driver emitted `30000` instead of the expected `3600000` client budget. - The final combined bridge, driver, proxy, and native protocol run passed: 95 tests in five files (71.17 seconds). - The native delayed-result probe passed with the current Rust Runner and current bundled proxy. It completed the later run with the same Runner PID and an empty pending-call ledger. - Actual provider check passed on macOS arm64 with `opencode-ai@1.18.34`, the changed source, and a local deterministic OpenAI-compatible model. One assigned write waited 31 seconds and saved one file. Both turns completed. The second turn retained the saved tool receipt. The write ran once across five model requests. - Bridge safety cases cover result replay without another action, ordinary timeout behavior, MCP cancellation, and teardown of an unanswered durable request. - Both runner TypeScript projects passed typecheck. The first attempt lacked the eval-kernel build; it passed after building that declared dependency. - Map alignment, module boundaries, and `git diff --check` passed. - The first native follow-up probe failed because its synthetic provider forgot the initial completion-contract revision. This caused a repair turn and another discovery call. The fixture now retains the contract, and the same probe passed. The failure is retained here. - Runtime acceptance used the implementation at `583d316ad4`. The later clarification commit changes comments, documentation, and a test name only. It does not change runtime behavior. - Full current-head CI passes, including typecheck, build, general and serialized tests, Rust/Runner checks, all eight E2E shards, and the `ci / e2e` and `ci / verify` aggregate gates. Greptile is 5/5 on this commit, with no unresolved review threads. No deployment or retry of the affected historical task was performed. The actual provider check used a local model; it does not qualify a deployed Linux sandbox or a live model service. ## Risks - A native call can wait longer while its Runner owner remains valid. Stop, teardown, and lease expiry still end the wait. The MCP client has a separate fixed one-hour ceiling. Runner ownership can renew beyond that ceiling. If the client deadline expires first, the operation outcome remains unknown and the existing hold must still block replay and run attachment. - This prevents new premature timeouts. It does not classify or replay an old request whose effect remains unknown. - No schema, migration, credential, or permission change is required. ## Model Used - OpenAI Codex, GPT-6 family. The harness does not expose the exact model ID or context window. Capabilities used: code generation, tool-assisted diagnosis, local execution, and test analysis. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
dd02595390 |
fix(db): retain hosted census construction callers (#18725)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Operators use a bounded database census to find expensive read paths. > - The hosted server combines application code and capture wrappers in generated files. > - Generic HTTP owners suppressed caller capture. Generated paths also failed the caller allowlist. > - This PR retains trusted query construction locations without source maps. > - Operators can distinguish application query sites while exact owners, privacy limits and capture bounds stay intact. ## Linked Issues or Issue Description Refs #17522, #17564. **What happened?** A real map-free bundle with two application query sites produces one callerless census group. This affects unowned queries and queries under the generic `http.request` owner. **Expected behavior** The census should distinguish trusted query construction sites. It must keep unknown calls unknown and preserve specific semantic owners without stack capture. **Steps to reproduce** Run the committed bundle test and pinned comparison helper. The helper bundles the actual capture module and real Drizzle, with a fake postgres-js client. It does not assign a fake path to an isolated mock caller. **Paperclip version or commit** Control: `09162fe893b464dd2ccdafd14bd3ef1611202a86`. The branch is rebased onto `1e3ab690b072bbf6a1c1f1d0454ea49315b2074c`; neither intervening commit changes this owner. ## What Changed - Keep the exact generic HTTP owner and add bounded caller capture only during the active census window. - Exclude capture wrapper frames with a local V8 stack boundary. Keep one static generated construction anchor for trusted root builders. - Keep unscoped bundled driver and relational paths unknown. Their first frame can belong to the ORM. - Cover exact application lines, alternate parameters, prepared builders, transactions, nested savepoints, privacy and inactive windows in a real bundled scaffold. - Add a portable pinned comparison helper and retained measurements. Existing group, page, byte, expiry and output-rate caps remain in force. ## Verification - PASS: 54 focused unit, real-bundle, existing runtime bundler and real-PostgreSQL tests. - PASS: database TypeScript check. - PASS at final head `b270b92821`: full local `pnpm -r typecheck` and `pnpm build`. - The first local `pnpm test:run` was stopped after two chat partitions passed (668 cases), because source changed during the run. This is partial evidence only. The fresh final-head full local command passed all six chat partitions (1,829 cases). Its general phase reported one failure in the unchanged workspace port-conflict test. The exact case then passed alone (1 pass, 178 skipped), and the unchanged full file passed 179/179. The original failure did not reproduce; its cause is unknown. The general phase was stopped after its remaining serial queue was measured at roughly three hours. Cancellation did not print the original assertion diff. The full command is failed/incomplete, not passed. Remaining general files, workspace-project phases and 164 serialized files were not reached. Fresh final-head CI passes all partitions. - Final-head code CI passes. Greptile is 5/5 with its thread resolved. Paperclip Review is delayed by an external review workspace recovery failure; it has not completed on this head. - Five serial alternating rounds per mode, 1,000 queries per sample, 40 retained samples, Node v24.22.2. Generic and unowned attribution rises from 0/1,000 to 1,000/1,000 across two exact application anchors. Specific owners stay stack-free. - HTTP median wall cost is 10.528 → 15.169 µs/query; process CPU is 15.461 → 17.662 µs/query, with noisy overlapping ranges. This is census overhead, not a product speedup. - Review failure retained: the existing runtime bundler test first failed because the public parser accepted arbitrary generated frames. The fix restores default rejection and permits generated frames only from the trusted builder boundary. The unchanged existing test now passes. - Full ranges, raw samples, proof boundaries and reproduction commands: https://github.com/paperclipai/paperclip/blob/b270b92821ae47c456a1b61d78c1593c5348ede7/doc/performance/2026-11-03-db-capture-hosted-callers.md ## Risks - Active generic HTTP census now captures stacks. More groups can reach the existing 2,048-group cap sooner. - Anchors identify construction sites in one exact generated build. They do not identify later async consumers, TypeScript lines or CPU owners. - Unscoped driver calls can remain unknown. This is intentional. - The no-network local fixture excludes deployed Sentry/OTel stack formatting and traffic effects. It does not establish exact hosted overhead. - Wrappers remain installed after expiry. Disabled, future and expired windows do not capture stacks. No schema, authorization, provider, or query behavior changes. ## Model Used OpenAI GPT-6.1 Sol (`gpt-6.1-sol`) with reasoning, code execution and repository tools. The host does not expose the context-window size. Independent review used the same model family. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [ ] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
bf7f77d8cd |
fix: avoid response recovery self-contention (#18715)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The server must present accepted agent responses as comments under current chat authority. > - The recovery worker can check four tasks at the same time. > - Different tasks can share a chat endpoint or external principal. Their authorization transactions then compete for endpoint or identity locks. > - A failed transaction leaves the accepted response pending and repeats its source queries on a later retry. > - This pull request groups checks that share a task, endpoint or principal before dispatch. It keeps every transactional authority check and lock. > - The benefit is less failed recovery work and prompt, exactly-once presentation of valid accepted responses. ## Linked Issues or Issue Description Refs #17714 for a separate publication-commit lock-order fix. Refs #18686 for separate run-list JSON work. Neither changes this recovery admission path. **What happened?** A response-presentation worker reported `could not obtain lock on row in relation "chat_endpoints"`. Four real PostgreSQL recovery fixtures reproduced three such failures in one pass with no external ingress. The worker grouped by task, while authorization locks shared endpoints and principals. Review also exposed two valid Slack endpoints in one workspace that share one principal and compete for its try-advisory lock. **Expected behavior** Recovery checks that share an endpoint or principal should retain task order without competing with one another. Unrelated endpoints should make progress. The server must still reject stale authority and retry legitimate external contention. **Steps to reproduce** Run the shared-endpoint regression or the portable comparison helper linked below. The fixture creates four accepted historical responses for distinct tasks and principals on one actual endpoint. ## What Changed - Read optional endpoint and principal scheduling hints through a company-scoped join to the canonical wake action. The join does not change eligibility or grant authority. - Extend existing bounded check groups to merge intersecting resource keys. Preserve task order, transitive conflicts, independent-group progress, shutdown, drain and pending scopes. - Add real PostgreSQL controls for shared endpoints, a principal shared across two endpoints, an observed blocked transaction with independent endpoint progress, external locks, exact selected output, and endpoint revocation despite a forged hint. - Add a portable helper that materializes the pinned baseline with `git show`, runs the existing real fixture, retains results, and restores temporary files. ## Verification - Final principal-extension head `57047441b28765ed1cf49b8924efc69f5c03dcb4`: 34 focused scheduler/chat controls passed on Node 24.22.2 / pnpm 9.15.4. Independent exact-head source review is green. Full local `pnpm -r typecheck` and `pnpm build` pass on this head. Fresh Greptile and Paperclip reviews are both 5/5 with no findings. All 59 final-head contexts are complete: 55 success, 3 skipped, and the neutral Paperclip review at 5/5. There are no failures or pending checks. - Earlier endpoint-only head `eea437e7edbe7eb41eb880d4d8fd6d99a9c247da`: full local typecheck/build and all remote CI passed. The redundant local full suite was interrupted for this review fix after 956 passing chat cases; it is not a full local pass. - Final shared-principal counts-only comparison completed the same two exact responses: product calls 165 → 147, rows 114 → 100, outer contention failures 1 → 0. SQL query errors stayed zero; advisory-lock refusal is not a failed SQL query. - The earlier endpoint-only study completed five alternating pairs / ten measured runs: median calls 443 → 293, rows 314 → 200, query errors 6 → 0 for four exact responses. Its instrumented shared-host CPU data predates the principal extension and is not a final-head or hosted CPU claim. Retry clocks were accelerated; wall time is not hosted latency. - [Raw samples, final principal work counts, portable commands and failure accounting](https://github.com/paperclipai/paperclip/blob/57047441b28765ed1cf49b8924efc69f5c03dcb4/doc/performance/2026-11-03-response-presentation-contention.md). - Initial server4 CI reported three provider calls instead of two in the unchanged Anthropic identity fixture. The exact local case passed; the full 88-case file passed with observations showing the two expected identity-route calls. The third CI call's origin remains unknown. One bounded rerun of the failed job and its dependent CI aggregate passed. No assertion or source change was made for that failure. - The causal benchmark baseline worker/query is `82c033ed127d07862e6765aa4b69a3cf339bde81`. The PR base is `36df8933f0`. They are not claims about the currently serving revision. ## Risks - Connected resource groups can reduce parallelism when tasks link several endpoints. This preserves the existing task ordering contract. - Missing or stale endpoint/principal hints retain current transactional checks. Other workers and ingress can still cause legitimate NOWAIT retries. - There is no schema, API, permission or provider-transport change. Endpoint revocation and company isolation remain enforced by their existing owners. ## Model Used OpenAI GPT-6.1 Sol (`gpt-6.1-sol`), with tool use and code execution, produced and reviewed this change. The runtime does not expose the context-window size. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
76ee5d0886 |
fix(evals): preserve singleton reports and pin reviewed chat repairs (#18749)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The chat eval lane runs selected cases from one immutable candidate. > - The finalizer must authenticate every original projection and publication. > - The pinned artifact action places one matched artifact at the download root. > - The finalizer expects named child directories, so a valid one-case run fails before final coverage verification. > - This PR supports both valid layouts and binds one reviewed repair candidate. > - The lane keeps exact selection, source, digest and publication checks. ## Linked Issues or Issue Description Refs: #18740, #18736, #18742, #18734, #18717. This PR is a proposed replacement for the source pin in #18740. That PR remains open. The evaluated candidate contains all five repairs from that prior cut and five further reviewed repair conditions. The product PRs remain unmerged. Merge authority for this replacement is pending; this PR does not dispatch a run. A completed one-case run has a verified original PASS and a verified publication. Its final aggregate fails because the pinned artifact action flattens a single matched artifact. The same path assumption also affects the later publication verifier. The final checks must still run; the repair must preserve their identity and digest checks. ## What Changed - Bind the finite candidate `ab2019e976076ec0eb60b026844f3d9211cb0208` through the manifest and exact descriptor snapshot. - Retain the original 244 case identities and the explicit cumulative definition overrides. Keep the historical definition source separate from the current candidate. - Add the reviewed source dependencies for real hire-review readiness and the strict later-human-intent review overlay from unmerged #18742. Preserve all original clocks, queries and model settings. Keep the earlier Studio144 accounting change explicit. - Document one current target and identify the prior Source5 cut as unmeasured. - Preserve named singleton download destinations for both bundles and publications. Authenticate every bundle before gathering its safe projection. Reject mixed, foreign, missing and symbolic-link layouts. Accept only the producer's exact closed once-marker schema. Do not export or upload results again. ## Verification - Candidate controls: 217 pure checks in 13 files; 242 real factory declarations with zero fixture callbacks; 42 admission controls; unchanged 506-input SDK registry; focused actual database-row type check. - Combined trusted pin and layout controls: 124 Node checks in seven real files, plus source-byte closure for all 244 singleton evidence policies. The unchanged parent reproduces both layout failures. Actual producer-marker, multi-case and malformed-layout controls pass. - The two actual PostgreSQL route regressions on core #18742 head `61996a96dd09e90a5788e8711a94fc5ae631f425` ran and passed. That core PR remains unmerged. This is not chat-case acceptance. - The independent layout logic peer and the whole final eight-path PR source peer passed. Current-head CI completed with 48 successful jobs and one skip. Both required checks passed. All 60 exact-head checks completed: 55 success, two neutral and three skipped. The separate Windows Rust check passed. Greptile and Paperclip Review both scored 5/5, with no unresolved threads. - The optional raw-order note is [disposed with source evidence](https://github.com/paperclipai/paperclip/pull/18749#issuecomment-6598680442). The admitted selector writes canonical order before hashing. A changed raw order fails closed before copying. The checked code remains unchanged. - Full local candidate typecheck and build are not claimed. Installed workspace dependency gaps prevent that full local check. The existing remote source and build preflight must pass before a native query. - No new model run or TestOps upload was used to prepare this PR. Prior attempts and grades stay unchanged. ## Risks A permissive layout repair could hide missing or foreign publications. Exact selected names, singleton counts, source identities, hashes and final readback checks must remain enforced. The layout helper must reject symbolic links and mixed shapes. The active default-branch ruleset requires CODEOWNERS review for the workflow. Normal merge is currently blocked on that review. This PR does not enable auto-merge, change branch rules or dispatch an eval. The candidate contains explicit unmerged product overlays. It is not a core release or a full-catalog qualification. The other five SDK cases remain held. SDK115 completion remains unknown. Provider quota failures are not fixed by the skill observer change. No credentials, caps, runner permissions or workflow admission references are added. ## Model Used OpenAI `gpt-6.1-sol`, with code execution, tool use and independent agent source review. The host does not expose the exact context window. No model eval was run to prepare this PR. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes:` / `Closes:` / `Refs:` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub links) - [x] My branch name describes the change and contains no internal ticket id - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing>canary/v2026.1103.0-canary.8 |
||
|
Dotta
|
ab2019e976 | test(evals): bind reviewed source6 dependencies | ||
|
DottaandPaperclip
|
f1230ec76a |
test(evals): observe team install reply within original clock
Bind the actual human and canonical native source before independent installed-runtime readback. Preserve installer, partial-outcome, repeat, runtime, query and original480s contract. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
Dotta
|
f2b94c57d4 | test(evals): apply reviewed human-intent replay candidate overlay | ||
|
DottaandPaperclip
|
2b6a17c158 |
test(evals): observe skill updates at semantic checkpoints
Poll local native readiness before reading GitHub. Preserve real provider reads at preview and installation, reusing only an unchanged exact immutable-source checkpoint during grade settling. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
c75fba8e77 |
test(evals): observe personal-secret reply within original clock
Bind the current person's canonical source and native reply before secure human entry. Preserve the actual original660s definition and all independent secret/privacy phases. Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
ad1e11fca6 |
test(evals): await canonical hire review link within original deadline
Co-Authored-By: Paperclip <noreply@paperclip.ing> |
||
|
DottaandPaperclip
|
7b8aa60b6d |
test(e2e): add repository continuation journey (#18729)
## Thinking Path > - Paperclip keeps agent work in project workspaces. > - A later user request must continue from the saved repository state. > - A successful reply alone cannot prove file or Git history preservation. > - The repository journey in #18652 passed against QA staging with independent byte and commit checks. > - Its separate completion journey remains unpassed and must stay open. > - This PR delivers only repository continuation and reuses the merged shared support from #18728. ## Linked Issues or Issue Description Refs: #18652. Builds on merged #18728. This is the repository-only delivery from #18652. The original completion implementation and all failed attempts remain on that open PR. ## What Changed - Add `repository-continuation` with the exact measured source fixture, two native turns, and eight required checks. - Start from measured source `ac4efb7c6f4782ea806661683c656fd2b4cc3189`. Keep the repository flow and Git/byte oracle unchanged. Clarify the two file-content instructions to require the word first or verified followed by exactly one LF newline; retain the exact expected bytes and the original random-reference contract. - Document the two-turn proof boundary: final saved hidden Git metadata is not independently read. Keep all eight success assertions. - Use the original-deadline and routine failure-reporting repairs from merged #18728. Integrate its actual master squash `e5713d95c6d9a258882649789ea5eb071cc17dc0` through one normal merge. - Use a fixed public read-only repository commit. Verify downloaded Git objects, parent commits, changed paths, saved workspace identity, and exact source bytes across both turns. Reopen the registered outputs through the browser. - Keep the shared artifact reader behavior unchanged and narrow its type to the API capabilities it uses. - Add only this case to the catalog, feature association, dispatcher, source hash, types, run/check contract, strict catalog expectation, and journey documentation. ## Verification - Current head `86808b7ec17e1727bd55a3fd46bc7e0409ca0e31` passed fresh [QA staging run 41589522430](https://github.com/paperclipai/paperclip-cloud/actions/runs/41589522430) and its [full report](https://pages.paperclip.ing/evals-qa-smoke-41589522430-1/). All eight checks and five phases passed. Both native turns succeeded and committed in the same workspace. The independent same-buffer Git audit verified that the second commit has the first commit as its parent. The saved original files and browser downloads matched their expected bytes. This does not add a final hidden `.git`/HEAD assertion. - The fresh run used OpenCode with `openrouter/deepseek/deepseek-v4.1-flash-1017` on QA staging/Daytona. It used Core `09162fe893b464dd2ccdafd14bd3ef1611202a86`, Cloud `cloud-41570439002-1`, and environment configuration version 9. Actual checkouts matched public source `86808b7ec17e1727bd55a3fd46bc7e0409ca0e31`, adapter `7367cf6fd0ee1cf52c2651a8a6671e087728a9c1`, and protected workflow bootstrap `a9ad6877de62fe3f65aacadc8a9dfba7d90729cf`. The bootstrap matched the accepted 117-file QA contract. - Normal cleanup completed. The original manifest is complete, evidence cleanup is complete, no operations remain active, and the owned agent is terminal. The coordinator verified ACTIVE TestOps case 685 / result 3695 and closed [launch 1921](https://paperclip.testops.cloud/launch/1921). This records normal cleanup closure; no separate physical provider-stop acknowledgement was collected. - Measured source `ac4efb7c6f4782ea806661683c656fd2b4cc3189` passed [QA staging run 41573331189](https://pages.paperclip.ing/evals-qa-smoke-41573331189-1/) against Core `09162fe893b464dd2ccdafd14bd3ef1611202a86`. TestOps launch 1723 / result 3501 maps to case 685. All eight checks and five phases passed. The two run bindings, independent saved-byte/Git audit, and completed cleanup were verified. - Initial extraction `54103639850d08d4d125ebe0493d8d3d49b58211` preserved the measured case files. The current head also clarifies the two LF instructions, documents the final Git-metadata limit, and inherits the reviewed helper repairs. This changed the fixture conditions. The current head now has its own live pass, listed below. The original ac4 result remains a pass under its original conditions; it was not used to qualify this head. - The emitted-prompt control failed on the old quoted backslash+n instructions, while 14 sibling controls passed. All 47 focused repository/artifact controls then passed after the two-sentence repair. Local: 891 related support tests, runner E2E typecheck, recursive typecheck, build, maps, token gates, and `git diff --check` passed. The inherited routine source has 62 causal read/cleanup controls and 26 unchanged real component controls. - The initial repository head's normal full CI run 41578686494 passed. Its 4/5 review and all original findings remain recorded. The later stacked head `037b979477161881a4877230500e6f3de8faea5c` also passed full CI run 41581342503 and a fresh 5/5 review. Current integration head `86808b7ec17e1727bd55a3fd46bc7e0409ca0e31` passed normal full [CI run 41584981651](https://github.com/paperclipai/paperclip/actions/runs/41584981651), attempt 1. Its fresh exact-head Greptile review is 5/5, with zero unresolved threads and no open findings. The broad local suite was not repeated. The earlier full local run recorded in #18643 had four failures and did not reach 164 serialized groups; this PR does not claim an all-local pass. - Earlier repository failures remain recorded: provider-pack transfer timed out in one attempt; a later attempt could not read saved workspace files. The later pass does not regrade those failures. The product repairs are separate from this unchanged case delivery. - The fixed actual #18728 squash is now integrated. The PR targets master. Its aggregate diff has exactly 14 repository-only paths, each identical to `037b979477161881a4877230500e6f3de8faea5c`; all other master changes are preserved. No later master head was chased. The current CI, review, and fresh staging qualification all passed for this integration head. ## Risks - Git bundles and workspace bytes are private evidence. Public reports retain closed checks; the fixture never pushes to the public source repository. - The second committed output must retain the first commit as its parent, the same workspace binding, and original saved file bytes. Final hidden `.git`/HEAD after the second save is not independently exposed by the supported remote workspace API; this test does not claim third-turn Git metadata durability. - Only JSON metadata observations retry transient transport failures. Binary attachment and browser downloads remain single attempts. Transport failures and digest mismatches remain visible failures. - The existing artifact reader is shared. Its runtime code is unchanged and its regression coverage remains included. - Daily source pins and plan selection stay with the coordinator. This PR does not register or bypass the unpassed completion case. ## Model Used OpenAI GPT-6 through Codex, with tool use and code execution. The exact deployment ID and context window are not exposed by this session. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] All Paperclip CI gates are green - [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing>canary/v2026.1103.0-canary.7 nightly/v2026.1103.0-nightly.0 |