mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-10-07 16:11:46 +02:00
Devin FoleyandPaperclip
228f0e2807
## Thinking Path > - Paperclip lets people supervise agent work and review its outputs. > - A task can wait for a human to review a pull request while an agent schedules checks. > - The PR work product and detected external links use separate display paths. > - A private PR can disappear from the useful sidebar view when its provider lookup fails. > - The monitor countdown says when the next check runs but omits the saved review request. > - This change shows the saved PR and requested review in the task's existing surfaces. ## Linked Issues or Issue Description **What happened?** A task kept checking whether a private GitHub PR had merged. Its work product requested board review, but the Properties sidebar used detected external objects instead. The PR URL was stored only in metadata, which the PR refresh path did not read. The composer showed a generic monitor countdown with no PR link or review action. **Expected behavior** Show saved PR links even if provider access fails. During a GitHub monitor wait, make outstanding review requests visible with the next check time and a status-check action. Stop requesting review when a PR merges or closes. **Steps to reproduce** 1. Register a PR work product with `reviewState: needs_board_review` and its URL in `metadata.url`. 2. Schedule an external-service monitor for GitHub. 3. Open the task with external-object lookup unavailable or unable to access the private repository. 4. Inspect Properties and the composer wait strip. Related work: #14469 added rich artifact cards. #8759 addresses attention on monitored task blockers. This change uses the existing work products and monitor action; it adds no blocker or approval mechanism. ## What Changed - Show saved PRs in Properties, including metadata-only links, independent of external-object availability. - Deduplicate equivalent GitHub PR URLs while preserving the saved navigation link, and put explicit review requests first. - Keep provider status and freshness on the combined PR row; keep private PR links usable when lookup fails. - Show the saved PR review request above the composer and in the existing monitor banner during a GitHub monitor wait. - Label the existing monitor action `Check status` and show check failures inline. - Suppress review prompts for merged, closed, or archived PRs even if their review flag is stale. - Refresh PR metadata using `metadata.url` and the existing `repository` alias. - Share saved work-product reads across the thread, Properties, and Artifacts. Refresh GitHub in a separate query and enrich only matching PR versions. - Start a fresh saved-row request on live invalidation so late provider responses cannot hide new artifacts or changed review requests. - Cover stalled GitHub lookups in both panels so refresh latency cannot hide saved work. - Scope monitor-check mutation state to the task so failures and late responses do not leak across navigation. - Refresh GitHub status when the displayed run finishes, including when saved PR rows have not changed. - Clarify that PR review and external release handoffs need a saved human-input interaction with an agent assigned for continuation; a flag, monitor, or handoff comment alone does not create that card. ## Verification - 386 tests pass across eleven affected UI and server suites on `184c7d8746`. Regressions cover saved links, provider status, cold-cache loading, panel reopening, monitor errors during navigation, and live updates during provider refresh. - Three live-update regressions and two run-completion regressions fail before their fixes and pass afterward. These tests use the real API client's GET coalescing and abort handling. - Full repository typecheck and build passed on the merged parent `1a49112dd2`. The latest UI changes pass UI typecheck and token gates; CI also verifies the latest build. Capability contract/inventory checks pass. - Storybook build and browser checks passed before the query race fix. Browser checks cover desktop and 390px mobile review waits, plus video, mixed-file, and empty artifact galleries after background refresh. - The earlier full local `pnpm test:run` was stopped under disk pressure. It reported failures outside the changed suites; an isolated skill-cache run reproduced three existing macOS permission failures. The full local suite was not rerun for this follow-up. - Latest-head CI passes on `184c7d8746`, including the browser shards and canary dry run. Apex is 5/5 with no actionable findings and no unresolved threads. All eight reported findings are addressed. ## Risks - This uses saved PR review state. When GitHub access fails, the saved state can remain stale until the agent updates it. The link remains visible and the status check remains available. - `Check status` wakes the existing monitor owner. It does not merge a PR, accept an approval, or mark the task done. - No schema, permissions, or scheduler behavior changes. ## Model Used OpenAI GPT-6 via Codex, with reasoning, repository analysis, code editing, and test execution. The exact deployment model ID and context window were 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 — 277 affected tests - [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>