mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-10-08 21:03:51 +02:00
5ee9e751fbb38787c055ee793c3412ecfc89da2e
6
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 DottaandPaperclip
|
f56aee5423 |
fix(evals): capture complete durable run event streams (#13977)
## Thinking Path > - Paperclip manages AI agents and records their durable work outcomes. > - Product E2E checks those outcomes through the browser and public API. > - A successful long run can emit more than 1,000 durable events. > - The harness read one page and missed the later completion evidence. > - This pull request reads every page before it checks runtime invariants. > - Invalid or incomplete capture still fails. A missing page cannot produce a pass. ## Linked Issues or Issue Description Related: #13882 (Grok qualification). A duplicate search found no existing event-pagination fix. **What happened?** The local structured-question case in [campaign 36071063537](https://github.com/paperclipai/paperclip/actions/runs/36071063537) completed the task and passed its six outcome matchers. It failed native runtime invariants because the capture contained exactly 1,000 events. The last captured event preceded the run's completion by more than a minute. The API caps each response at 1,000 rows. The harness did not request the next page. **Expected behavior** Read the complete durable event stream through the public API before checking semantic-result and terminal-event counts. Reject incomplete or malformed evidence. **Steps to reproduce** 1. Complete a native task that emits more than 1,000 durable events. 2. Place the semantic-result and terminal events after row 1,000. 3. Capture the run with the Product E2E harness. 4. Before this fix, the invariant checker sees only the first page. **Paperclip version or commit** Observed at `4196a4cd76db434854b679035e4146c7f69689ce`. The same single-page capture exists on master. The original failed result remains unchanged; missing historical tail evidence is not reconstructed or graded as a pass. ## What Changed - Add a bounded event collector that advances through the public `afterSeq` cursor. - Use it for task success/failure evidence and shared chat run evidence. - Reject invalid pages, missing or non-increasing sequence numbers, repeated cursors, failed later requests, and an exhausted page limit. - Test completion events beyond the first page, exact page boundaries, and malformed evidence. - Correct the existing Everyday catalog test from 38 to the maintained 47 cells. The suite stays explicit-only. - Document the complete-capture requirement and its bound. ## Verification - Product harness typecheck passes. - The full credential-free harness suite passed 477 tests. After adding the chat integration regression, all 34 chat evidence tests pass. - Seventeen pagination tests cover the valid tail and malformed-evidence cases. - `git diff --check` passes. - [Full repository CI](https://github.com/paperclipai/paperclip/actions/runs/36080683422) passes at `aed6f79c089226be79e75dcf390969561ec4f787`: 52 successful checks and two intentional skips, including Rust, typecheck, build, server tests, and browser shards. Greptile gives this exact head 5/5 with no findings. No local Docker, Rust build, browser suite, or model invocation was used for this change. - [Live Grok requalification](https://github.com/paperclipai/paperclip/actions/runs/36080870743) is running on combined source `1b0551bb7c8de3c54f4bee64dbe2c88328b3645e`, with the runner, scheduler, and evidence fixes. Its full credential-free harness passes all 493 tests and typecheck. Live results are pending; the original campaign remains a failed measurement. ## Risks Long runs need more read-only API requests and larger private evidence files. Capture stops with an explicit error after 100 full pages. This changes neither production APIs nor provider behavior. It does not relax an invariant or change a historical grade. ## Model Used OpenAI GPT-6 through Codex, with repository tools and code execution. The exact serving identifier 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 #` 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> |
||
|
|
3790ca2f13 |
fix(runner): repair approval and Stop races and eval infrastructure (#13750)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - Runner tasks must continue after approval and stop when the user presses Stop. > - Live evals found races at approval delivery and provider startup. > - Browser readiness and CI setup errors also hid the actual task results. > - This pull request fixes those races and the related test infrastructure. > - Regression tests and saved live reports show which cases now pass. ## Linked Issues or Issue Description Companion eval definitions PR: https://github.com/paperclipai/paperclip-evals/pull/25 (AgentCore paused and provider/environment infrastructure). Related: #13741 now supplies the late-startup Stop fence and warm-attachment recovery; this PR retains that fence and extends startup tracking and regression coverage to both native backend paths. #13539 introduced queued approvals during active runs. #13738 fixes child assignment, task replies, and warm process continuity and is already in the base. #13291 concerns automatic continuation of interrupted legacy sandbox runs; this PR fixes native startup cancellation and does not change that recovery policy. **What happened?** An accepted service approval could wait after its source run stopped. Stop could return success before the provider handle existed. Work could then start after Stop, or a cancelled run could be recorded as failed. Some E2E tests also failed on unloaded browser content or irrelevant reply wording. Runner CI could fail before model work because of dependency or sandbox setup. **Expected behavior** Deliver each settled approval once after its source run stops. Do not start work after an acknowledged Stop. Preserve the audited cancellation. Test the intended product behavior with a ready browser and verified runtime dependencies. **Steps to reproduce** 1. Approve a service request while its source run is active. Let the run finish. Check that its result starts one continuation. 2. Delay provider startup. Press Stop before its handle is available. Check cancellation, then submit `/new`. 3. Run the browser, warm-workspace, and Stop-and-redirect cases from the linked report. **Paperclip version or commit** The branch includes master at `9d19f98b5`. The report records the original source for each focused attempt. **Deployment mode** Isolated local development instances and disposable Daytona sandboxes. ## What Changed - Deliver settled tool-action results for the exact company and source run during final cleanup. Keep the existing idempotent receipt and periodic recovery sweep. - Wait for startup to hand off its provider handle before acknowledging Stop. Reject first-turn admission after cancellation. Preserve a matching audited pending or acknowledged cancellation. - Wait for mounted task history and connector controls in browser tests. Record failure evidence. Grade workspace contents and process continuity separately from exact reply wording. Require each warm-turn marker once and in order, allowing surrounding prose. - Stop-and-redirect now checks that the source file exists and work is active before Stop. - Resolve target dependency locks in an uncredentialed CI job. Verify the lock artifact hash. Keep orchestration and publication on the trusted workflow revision. - Materialize the pinned OpenCode executable and configure the exact Codex executable's user-namespace profile before provider credentials are available. - Compress Daytona directory uploads with gzip. Preserve files, executable modes, symlinks, empty directories, and confinement checks. - Classify file-transfer RPC deadlines as infrastructure. Keep unrelated runner RPC failures visible. ## Verification - [Focused live report with screenshots and original attempts](https://pages.paperclip.ing/runner-reliability-20260921/): 14 of 15 selected Product E2E cases pass across the recorded revisions. Claude and Codex Stop → `/new`, Claude service approval, delegation, both hiring/reuse cases, and native Daytona warm continuity pass. - Two credentialed Runner smoke cases pass. These are not full protocol coverage. - E2E harness after the master merge: 429 tests pass. E2E and server TypeScript checks pass. - Daytona plugin: 239 tests pass, 6 skipped. Plugin TypeScript build passes. The compression test fails against the old code and passes with the change. - Runner backend/runtime regression group: 161 tests pass. Cancellation/startup selection: 26 tests pass. Approval delivery: 34 real-database tests pass. - Workflow security: 7 tests pass. Both edited workflows pass actionlint. Runner TypeScript and Rust builds pass. - After merging master, all 389 native executor tests pass, including both native backend paths and late startup after the Stop deadline. - Post-merge `pnpm -r typecheck` and `pnpm build` pass. The monolithic local `pnpm test:run` was interrupted to integrate master and is inconclusive. The [hosted CI test partitions](https://github.com/paperclipai/paperclip/actions/runs/35620461738) pass on `50a3e43822bcba1e0d07b1b45b0be91cbf9312da`. An unchanged sandbox callback schema test initially received HTTP 503. It passed five isolated local runs, its full local test file, and one failed-job CI retry. No assertion was weakened. ## Risks - Stop can wait for the bounded startup handoff. If it cannot settle, the existing pending-recovery state remains instead of a false acknowledgement. - Immediate approval delivery must remain idempotent across cleanup and recovery sweeps. Tests cover duplicate delivery and company/run boundaries. - The workflow changes still need hosted Linux verification. They retain the trusted workflow and credential boundaries. - Gzip reduces the observed provider upload from about 1.8 GB to 663 MB. It does not yet fix the remaining Claude Daytona transfer timeout. That recovery test never reached Claude, so recovery remains unverified. Use a matching image with the verified provider package preinstalled for the next recovery test; retain cold-upload coverage separately. - The report preserves diagnostic runs with missing source metadata and marks them as such. It does not claim a new full-suite pass. - This PR adds no new prompt policy or historical status reconciliation. ## Model Used OpenAI GPT-6 through Codex performed the primary implementation and review. The exact primary backend model ID is not exposed in this session. OpenAI `gpt-5.6-luna` assisted with bounded infrastructure work and verification. The agents used repository tools, code execution, and browser tests. The exact backend revision 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> Co-authored-by: OpenAI GPT-6 <noreply@openai.com> |
||
|
DottaandPaperclip
|
64895f187b |
fix(runner): clarify completion errors and restart test failures (#13682)
## Thinking Path > - Paperclip manages work across persistent agent sessions. > - The runner validates completion calls before accepting their results. > - Generic validation errors can leave the agent unable to repair a rejected call. > - Restart tests also exempted every later failure on an intentionally interrupted run. > - This change gives bounded schema feedback and limits the test exemption to expected interruption outcomes. > - Failures become easier to repair and diagnose without changing authorization or task prompts. ## Linked Issues or Issue Description Refs #13674 and #13676. Related environment and Agent Chat fixes landed in #13677 and #13678. Those changes do not cover these diagnostics. **What happened?** A malformed completion call received a general field list without the failed schema location. The everyday restart test hid later adapter errors on an intentionally interrupted run until its deadline. A clean pnpm install also broke the shutdown test because it resolved an undeclared Playwright package. **Expected behavior** Return enough schema information to repair completion calls without returning submitted values. Fail promptly on an unexpected recovery error. Resolve the declared test package's CLI. **Steps to reproduce** Run the new completion-validation and everyday lifecycle regressions against the parent commit. The new assertions fail there. Run the shutdown test in a clean workspace installation. **Paperclip version or commit** Based on master |
||
|
DottaandPaperclip
|
84fe89906d |
fix: complete native agent review handoffs (#13581)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work > - Native execution uses durable runs, issue locks, wake requests, and typed tool authority > - A child can finish with a native agent review request while its original assignee stays responsible for the work > - The reviewer then needs a bounded execution path that can inspect the child, record one decision, and finish safely > - Before this change, assignee-only gates rejected the reviewer or left the parent waiting after the child review ended > - This pull request adds typed reviewer admission, scoped reviewer tools, durable wake and recovery handling, and parent continuation evidence > - The benefit is that native review handoffs complete without changing child ownership or granting broad mutation access ## Linked Issues or Issue Description Refs: #13314 Refs: #13574 **What happened?** A native child run could report `needs_review` for an agent reviewer. The reviewer wake then failed assignee and execution-lock checks. The child remained in review and the parent remained waiting. **Expected behavior** The named reviewer should receive one durable wake. The reviewer should inspect the child and resolve the exact review card. The child assignee should stay unchanged. The parent should receive the recorded review outcome after the child reaches its terminal state. **Steps to reproduce** 1. Run a native task with a different named agent reviewer. 2. Keep the child assigned to its original worker. 3. Let the worker finish with a native completion review request. 4. Start the durable reviewer wake. 5. Resolve the review and finish the reviewer run. 6. Observe the child and parent state. **Paperclip version or commit** Base: `e926b1301`. PR head: `b31ad9ab8`. Live reviewer verification source: `eea171aae`. **Deployment mode** Built from source. **Installation method** Built from source (pnpm build). **Agent adapter(s) involved** Not adapter-specific (core bug). **Access context** Both. **Database mode** Embedded PostgreSQL in the isolated live test fixtures. ## What Changed - Add server-validated native review assignment facts. - Admit only the exact company, issue, source run, decision, revision, addressee, and resolver policy. - Give reviewer runs a narrow set of Paperclip read and resolve tools. File and shell access follow the configured agent and environment policy, so reviewers can run tests. - Separate server-owned reviewer instructions from untrusted persisted review data. Escape the data boundary; retain server-enforced authorization. - Keep the child assignee unchanged. Atomically claim the reviewer run, wake request, and issue execution lock. A competing lock prevents provider startup. - Require the exact running reviewer session and current issue lock to resolve its assigned card. Reject missing, unrelated, or terminal reviewer runs. - Add durable reviewer wake, lock, stale-card, and abandoned-run recovery handling. - Prevent duplicate native wake dispatches during deferred admission and recovery. - Carry accepted or rejected child review outcomes into parent task context and continuation evidence. - Add focused server, runner, and native protocol coverage. - Preserve upstream continuation rules. Add child review decisions as separate evidence, while keeping real human answers in their own field. - Return actionable completion validation feedback to both providers. Permit a corrected completion after rejection. Keep strict terminal acknowledgment validation. - Apply exclusive shared-workspace locks to sandbox environments. Local and SSH folders can run concurrently, including when old settings request serialization. - Repair test timing, native event parsing, and the review artifact assertion. Allow a valid reject, correct, and accept review sequence. Check the accepted card against its reviewer run and decision. Keep polling within the existing deadline when review acceptance precedes the parent wake projection; report a specific missing-continuation error at timeout. - Apply the ACPX pending-call limit to reserved finish/block calls, with capacity-release and cancellation tests. ## Verification - `pnpm build`: passed on `eea171aae`. - `pnpm -r typecheck`: passed on `eea171aae`. - `pnpm test:e2e:runner:unit`: 359 tests passed in 30 files on `b31ad9ab8`; runner E2E typecheck also passed. - `pnpm check:token-gates`: passed. - Focused DB review, reviewer authority, and prompt-boundary checks: 31 tests passed. They cover invalid reviewer runs, competing locks, atomic admission, duplicate claims, and valid resolution. - Heartbeat, workspace, and recovery checks: 30 tests passed. - ACPX sidecar suite: 27 tests passed. Moving the capacity guard back below reserved handling makes both new regression cases fail. - Four focused live continuation checks passed on their first attempt at `f15f55e0a`: answer updates scope (6/6 each on Codex and Claude) and question tool guidance (12/12 each). These cases do not use the reviewer prompt path changed afterward. - Fresh Codex and Claude review-handoff checks passed all 29 native checks each on their first attempt at `eea171aae`. Both runs received the expected fixed prompt and completed cleanup. Only the six selected live flows were tested; no full paid provider catalog run. - The final commit only extracts the existing test-harness timeout diagnostic into a shared helper and adds positive and negative coverage. Removing the accepted-review guard makes two regression assertions fail; restoring it passes all six timeout tests. Production runtime code, prompts, deadlines, and grading criteria are unchanged by this final commit. - Deadline regressions: a valid continuation delayed 20 seconds succeeds within its 30-second unit-test deadline; an absent wake returns a specific candidate-failure diagnostic at that same deadline. Both assertions failed before the fix. Production E2E deadlines remain unchanged. - Historical native failures remain recorded: Docker availability failures; a valid reject/correct/accept sequence that the first-card grader misread; and a test that rejected the gap between accepted child review and parent wake projection. No failed result was regraded. The latest tests use a protected reference to the pinned Docker image and the unchanged artifact oracle and time limits. - Full repository verification runs in GitHub CI. Local verification uses the focused suites above, full build, and full typecheck. An unchanged Codex shutdown timing test failed once in CI, passed in isolation, and its full shard passed on the final commit without changes to that test or its causal code path. The original failure is retained in the verification record. Greptile reviewed `b31ad9ab8` at 5/5 with no outstanding actionable findings. All review threads are resolved. All current-head CI gates passed, including the isolated native runner Docker build (55 successful checks; two skipped by the workflow). ## Risks - Reviewer admission depends on exact persisted decision and interaction bindings. A stale or changed card is rejected. - Paperclip control-plane tools are limited to inspection and review resolution. This is not a filesystem permission boundary; provider file and shell access retain the configured policy. - Deferred wake recovery changes dispatch receipt coalescing. A scheduler regression could delay a continuation if the receipt state is wrong. - Parent review outcomes are evidence for the model. They do not grant tool authority or change issue ownership. - This change does not address legacy lease-hold handoff behavior. > Roadmap review: native execution, review gates, and durable recovery are existing roadmap capabilities. This PR completes a narrow reliability path for those capabilities. ## Model Used OpenAI `gpt-6-astra` with reasoning, tool use, and code execution. OpenAI `gpt-5.6-luna` assisted with bounded implementation, review, and journal work. Context window size is not exposed by this session. Live test subjects use `gpt-5.6-sol` and `claude-sonnet-5`; they are not the PR authors. ## 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
|
9fd2e50310 |
feat: create company skills from runner tasks (#13538)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The Runner gives agents tools to change company resources. > - Users need agents to save reusable skills during a task. > - A saved skill needs a visible result that users can inspect and edit. > - This pull request adds `create_skill` and a task feed card linked to Skill Studio. > - Users can open the saved skill from the task and edit the same resource. ## Linked Issues or Issue Description **Subsystem affected** Runner tools, company skill storage, task feed, and Skill Studio. **Problem or motivation** The Runner has no dedicated tool to create a company skill. A user cannot follow a creation result from the task feed to the saved skill. **Proposed solution** Add a company-scoped `create_skill` tool. Save the skill with the existing company policy. Add one creation card to the task. Open a named sidebar tab from that card. Let the user open the same skill in Skill Studio. **Alternatives considered** An agent can write a local file, but that file is not a company skill. A second document copy in the task would become stale after a Studio edit. The sidebar therefore reads the saved skill directly. **Roadmap alignment** This extends the shipped Skills Manager, Skill Studio, and Skills Store milestone. The maintainer requested and approved this scope. Search found no duplicate `create_skill` PR or issue. Related UI validation work: #8715. This PR does not change that validation display. ## What Changed - Add the real Runner tool, its contract, and its mock implementation. - Validate the complete SKILL.md and derive company, task, agent, and run identity from authentication. - Apply the existing company skill policy. Do not assign the skill to an agent. - Make keyed retries return one skill and one creation event. Reject conflicting retries. - Make concurrent file creation safe. Never replace an existing published skill during creation. - Add a creation card, a named sidebar tab, and an Open in Skill Studio action. - Show saved Studio edits when the user returns to the task. - Add storage, policy, mode, retry, UI, and Product E2E tests. Document the tool. - Fix deleted-name reuse, onboarding panel persistence, immediate feed refresh, and mock validation parity from review. - Serialize Studio file edits and renames with skill deletion and recreation. Reject stale editor requests before they can change a replacement skill. - Generate the standalone mock parser and validator from the production contract. Use portable UUIDs so the browser scenario bundle builds. ## Verification - All latest-head PR checks pass on `145dd76a5`, including all server shards, browser E2E, Runner verification, build, typecheck, and release dry run. Greptile: 5/5 with no open findings. An interrupted CI runner was retried successfully. - `pnpm -r typecheck`: passed. - `pnpm build`: passed. - `pnpm check:token-gates`: passed. - Review regressions: 73 storage tests, 6 real API tests, 63 UI tests, and 61 semantic runtime tests passed. Parser synchronization passed. - CI exposed existing fire-and-forget Sentry test races. Reproduced the resumption race locally, then synchronized the related sweep and finalizer assertions on the actual report; all 27 tests across the three affected files pass. - Runner scenario browser build and strict content-security-policy check: passed. - Runner suite: 2,012 tests passed; 10 skipped. - `pnpm test:run`: the general-server batch had 12,416 passes and two failures. The old tool-count assertion was fixed; all 16 authority tests then passed. The chat webhook test had a socket error; it passed four isolated reruns. - Both workspace test groups passed. The isolated route suites completed. Two socket failures in the initial route batches passed on individual reruns; all remaining 61 files passed. - Product E2E `create-skill-studio`: passed with local Codex and local ACPX Claude. - Manual browser test: submit a task, observe the real tool call and creation card, open the sidebar, edit in Studio, save, and return. The task reached Done. The saved second revision and sidebar tab survived a server restart. - The new companion headless Runner Eval passed. Companion coverage PR: https://github.com/paperclipai/paperclip-evals/pull/23. Daytona was not run because no immutable runner image was configured. ## Risks - Database writes and local file writes cannot share one transaction. Recovery accepts only an exact file-for-file retry after a database rollback. Conflicting files remain untouched. - The sidebar displays the current skill. The feed card remains the historical creation receipt. - No database migration, dependency, or workflow change is included. - Remote Daytona behavior still needs a run with a configured immutable image. ## Model Used OpenAI GPT-6 (`gpt-6-astra`) handled design, integration, review, and browser verification. OpenAI `gpt-5.6-luna` assisted with bounded implementation and eval work. Both used code execution and tool access. 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 #` / `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> |
||
|
|
cceeb0aa66 |
test(runner): add everyday workflow evaluation harness (#13474)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The native runner must support project work, delegation, hiring, and service access. > - Browser tests exposed lost connection access, rejected helper events, and stalled recovery. > - Some eval failures also came from incorrect fixtures and decision controls. > - This pull request fixes those paths and adds eight everyday workflow stories. > - The tests retain observed failures and verify delivered files independently. > - The benefit is repeatable evidence for common user tasks and their remaining gaps. ## Linked Issues or Issue Description Related work: #13404 contains earlier workflow fixes. #13300 and #13470 changed the CI contracts used by the harness security tests. Merged companion: [paperclip-evals#22](https://github.com/paperclipai/paperclip-evals/pull/22). **What happened?** Native ACPX sessions did not receive the assigned connection gateway. Codex helper events could arrive before their spawn receipt and fail thread validation. A parent continuation could take a shared workspace before its child retried. A failed native continuation could leave the task status without a clear recovery blocker. The eval harness also confused tool approvals with new connection requests and could reject a valid delegated download. **Expected behavior** Keep assigned gateway access and its approval checks. Verify helper lineage before accepting helper progress. Let a waiting child proceed before automatic parent recovery. Preserve a failed task's recovery ownership. Grade the actual requested workflow and its delivered files. **Steps to reproduce** Run the everyday workflow suite with the native Codex and Claude profiles. Exercise service approval, connection refusal, delegated project work, and teammate reuse. The commands and case requirements are in `tests/runner-e2e/EVERYDAY-WORKFLOWS.md`. Use `pnpm test:runner-recovery` for controlled crash and replacement cases. ## What Changed - Pass the scoped connection gateway binding through the native ACPX host and sidecar. - Recognize Codex helper lineage from parent metadata and spawn receipts. Verify early helper events with `thread/read`. Keep helper events separate from root completion authority. - Guide agents to use persistent hiring, child tasks, dependency records, and a blocked handoff while waiting for a child. - Defer automatic parent recovery while a child has an active execution path in the same shared workspace. Allow parent recovery when the child needs review. - Record Blocked status and recovery evidence when a failed native continuation needs reconciliation, including existing active or escalated incidents. Preserve their owner and retry budget. - Add eight browser-driven workflow cases. Use real decision controls, explicit child feedback delivery, managed hiring credentials, and independent ZIP checks inside a bounded Docker sandbox. Verify sandbox availability before task creation. Record screenshot SHA-256 at capture. - Keep runner crash probes in controlled recovery tests. Preserve the original failure when cleanup also fails. - Display missing accounting and replay revisions as unavailable. Align harness security assertions with the approved CI changes. - Make the channel-rejection browser fixture bind its file after the send captures its payload. This prevents live refresh from removing the file before the simulated race. ## Verification - Full workspace `pnpm -r typecheck` passed after merging current master. - Runner E2E typecheck passed. Harness unit tests passed: 216/216. - Wake-queue database tests passed: 55/55. The two added existing-incident tests failed before the fix and pass after it. - Docker artifact calibration passed: 12/12. Host-file and host-loopback isolation tests failed before the fix and pass after it. Read-only delivery and output limits are also verified. - Full `pnpm build` passed. Targeted recovery tests passed: 83/83. - The channel-rejection browser test passed five consecutive runs after fixing the fixture race found in CI. - Local general-server (12,351 tests), UI (6,250), CLI (485), and workspace package groups passed. The monolithic run stopped at an unchanged lock-heartbeat fixture race; the isolated workspace group passed on rerun (shared: 747/747). A separate local serialized run passed 97 files before two socket errors in the unchanged issue-list route suite; that suite passed 15/15 on isolated rerun. These local full commands did not finish uninterrupted; the complete CI matrix below covers the remaining suites. - Final head `0fb293733fe307be7e6667ae8f1364077d0c6455`: **34 successful checks, 2 expected skips**, including every server/workspace shard, browser shard, native runner verification, build, and typecheck. [Final CI run](https://github.com/paperclipai/paperclip/actions/runs/34989136700). - Greptile reviewed this exact head at **5/5**; all review threads are resolved. Both Superagent security checks are successful. - ACPX credential-boundary tests passed: 118/118. Superagent accepted the runner/sidecar versus provider-environment trace and cleared its finding. - The latest paid local campaign on source `f6a2fdf7ac2af859826a2ae627ff4125a5478529` passed 22/24 cases: Sol 8/8, Claude 7/8, Mini 7/8. These results predate the merge with current master. - The two remaining failures are in `hire-reuse`: Claude exceeded the attempt deadline during final review; Mini made invalid deliverable tool calls and remained Blocked. - Six Daytona cases were not run because the matching immutable runner image was unavailable. This PR does not claim new remote model results. ## Risks The changes affect connection admission, helper identity, and recovery scheduling. Assigned gateway grants and user approval still govern service calls. The workspace admission gate still exists; the broader folder-sync design is separate work. Provider behavior can still cause the two recorded hiring failures. No database migration is required. Paid cases are opt-in and have bounded attempt deadlines. Project stories now require Docker and the documented pinned Python image on the harness host. ## Model Used OpenAI `gpt-6-astra` performed implementation, diagnosis, and substantive review. OpenAI `gpt-5.6-luna` assisted with verification, PR preparation, and review tracking. Both used repository tools and code execution. Context-window sizes were not recorded. ## 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 (focused checks and isolated reruns; full-run limitations are documented above) - [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: OpenAI GPT-5.6 Luna <noreply@openai.com> Co-authored-by: Paperclip <noreply@paperclip.ing> |