## Thinking Path
> - Paperclip manages agents and their assigned tasks.
> - A task pause controls execution without changing task status.
> - Resume can remove that pause and request work from eligible
assignees.
> - The preview counted blocked tasks and former executors as wake
targets.
> - The release route correctly skipped those targets, so the dialog
promised work that did not start.
> - Resume also trusted stale pause snapshots after tasks were reopened.
> - This PR aligns preview with execution eligibility and rechecks
current task state when releasing a pause.
## Linked Issues or Issue Description
**What happened?**
Pause an assigned blocked task, then open Resume. The dialog says that
one task will resume and offers to wake one agent. Applying it only
removes the pause. The route skips blocked and backlog tasks. A former
executor can also inflate the wake count after reassignment.
**Expected behavior**
The dialog must distinguish removing a pause from starting work. Its
wake count must use the current assignees and statuses accepted by the
release route. Blocked work must stay blocked.
**Steps to reproduce**
1. Open an assigned blocked task.
2. Choose Pause work, then Resume task.
3. Inspect the wake count and apply Resume work.
Related runtime recovery work is tracked separately in #17097 and
#17439. This PR does not change remote execution recovery.
## What Changed
- Share the resume status list between preview and release. Count only
current assignees on todo, in-progress, and in-review tasks for resume
wakes.
- Explain that Resume removes the pause. Show the existing preview
warnings, including tasks that remain idle.
- Disable and uncheck the wake option when no agents qualify. Send the
same choice to the server.
- Reevaluate previously skipped pause members after reopening or
releasing another hold. Retain active-pause and unsafe-execution guards.
- Add real PostgreSQL and HTTP-route regressions plus browser coverage
for both task interfaces. Keep draft, ancestor-pause, and
Stop/reconciliation coverage.
- Update the task coordination maps. Add an opt-in dedicated test
database URL for hosts where embedded PostgreSQL cannot start.
## Verification
- Final head `9548eb16736f632cc111a5c13a9a994ec1d95a13`: all 55
completed checks pass, with two optional Storybook checks skipped. [CI
40188094904](https://github.com/paperclipai/paperclip/actions/runs/40188094904)
passes, including typecheck, build, all general/serialized suites, and
all eight E2E shards. Greptile is 5/5 on this commit; the reopened-task
finding is fixed and resolved.
- Red: the real PostgreSQL regression receives three agents instead of
the one eligible current assignee on the original code.
- Green: all 27 tree-control service, route, and unit tests pass. They
cover current assignees, blocked/backlog exclusions, reopened tasks,
company boundaries, remaining pause holds, and execution blockers.
- The reopened-task route regression fails on the previous
implementation: skipped members never reach preflight/wake, and an
unsafe replay releases the hold instead of returning 409. All cases pass
with the fix.
- Browser: all five applicable cases pass together (90 seconds):
blocked-work resume in both interfaces, draft/ancestor-pause cases in
both interfaces, and the process
Stop/pause/reconciliation/restart/cancel lifecycle. The optional
native-runner fixture is skipped locally because its fixture binary is
not configured.
- All 157 task-page component tests pass on the prior commit; subsequent
changes affect the server route and Stop fixture only.
- `pnpm -r typecheck` and `pnpm build` pass before rebase, and CI
typecheck/build passed on the previous PR commit. Server typecheck
passes again after the route fix. The final-head CI run also passes
these full checks.
- `node scripts/check-maps.mjs`, `node scripts/check-token-gates.mjs`,
and `git diff --check` pass.
- Full `pnpm test:run` was attempted but did not complete: this host
cannot initialize embedded PostgreSQL, and unrelated workspace-runtime
tests timed out. The focused database tests and E2E suite use disposable
PostgreSQL 18 instead. All general-test and serialized-test CI shards
passed on the previous PR commit; the final-head CI run also passes
every test shard.
- The first CI run caught the existing Stop test trying to check a
disabled wake option for blocked legacy work. It now asserts zero wake
targets and retains the no-execution-before-reconciliation checks.
- Earlier E2E attempts exposed fixture problems: assignment started the
fixture before it was paused, and the classic interface had two matching
editor nodes. The fixture now enables dispatch after the hold exists and
scopes the final assertion to the task composer. No product assertion
was removed.
## Risks
Resume preview is a snapshot. The release route still checks current
status, assignment, and execution blockers. It does not bypass approval,
budget, or recovery gates. Pause, cancel, and restore keep their
existing affected-agent semantics. The test database override only
accepts a dedicated database name ending in `_test`; it runs migrations
before the suite and closes its clients afterward.
## Model Used
OpenAI GPT-6 through Codex, with reasoning, repository tools, browser
control, and code execution. 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
#` / `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>
Resolve the exact delivery through its primary key before looking up stale-head cancellations. Backport the predicate from #17664 while preserving the current worker lifecycle and receipt checks.
Verified 372 selected tests, workspace typechecking, the server build, green CI, and an Apex 5/5 review.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Include validated preserved monitor proof in the immutable assessment input so restored receipts or completion times can produce a new status decision at the same deadline. Preserve accepted results, earlier decisions, provider attempts, monitor deadlines, and idempotent replay.
Validate two baseline regressions, 201 targeted tests on current master, workspace typecheck, server build, full CI, and Apex 5/5.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read compact run identity during authentication and compact turn state during conversation recovery. Preserve fresh checks, durable steering and reply evidence, locking, and complete stored payloads.
Validate 184 targeted PostgreSQL and lifecycle tests, workspace typecheck, server build, full CI, and Apex 5/5.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Find the contiguous native cursor in PostgreSQL while preserving gaps, late arrivals, source isolation, and event replay. Verify a long-history response budget against current master.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Replace broad newer-run responses with bounded company-scoped existence probes. Read keyless history once per unique agent and preserve existing matching and millisecond rules. Keep all stored run data.
Verify with 31 real PostgreSQL attention tests, workspace typecheck, server build, response-budget counterfactuals, full green CI, and Apex 5/5 review.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Use current scoped database inventory fingerprints and the current on-disk expectation before reusing bounded snapshot verification. Preserve full validation, seeding and access checks when content, expectation or row existence changes.
Verified177 skill and authorization tests, workspace typechecking, server build, all green CI and exact-head Apex5/5. Prior staging release measured and saved Legacy data fingerprints preserved.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Guard the retained-evidence update with its existing JSONB comparison. Reuse the full first result when evidence is unchanged, while preserving new evidence delivery, source-bound retention, cleanup, and recovery holds.
Verified with 199 recovery and retention tests, full workspace typechecking, the server build, green CI, and exact-head Apex5/5.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select only runtime mode, status, complete result evidence and the delivery token from the locked legacy terminal-state read. Keep filters, row lock, guards, writes, full returned rows, source retention and recovery holds.
The synthetic regression reduces a 699,454-byte read to under 4 KB while preserving full stored and returned data. All CI checks, 236 combined tests, workspace typecheck and server build pass. Apex 5/5 after correcting a documentation contradiction, with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Project unused launch context and native execution input out of two preliminary recovery reads. Preserve full results, unknown controls, complete legacy profiles, current filters, native rejection and full locked terminalization.
Apex 5/5 with no unresolved comments and all CI checks green. Ten combined local suites pass (370 tests), plus workspace typecheck and server build. Hosted traffic effects require staging verification.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Preserve a valid saved monitor after its successful source releases execution and checkout. Recheck accepted result identity, receipts, ownership and governance under the task lock. Notify the existing scheduler once and preserve accepted live and recovered replies.
Apex 5/5 after fixing its reply-presentation finding. All CI checks, 191 focused tests, 293 broader native tests, workspace typecheck and server build pass. Hosted effects require staging verification.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Keep locked process and terminal-state evidence plus JavaScript-compatible checkpoint presence. Preserve every host, provider, lease, coordinator and stop-receipt check without transferring saved prompts or checkpoint bodies.
53 focused tests, server typecheck, build, all 55 CI checks, and exact-head Apex 5/5 pass.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Project native input to the complete schema, binding, provider, and workspace evidence used by execution-blocker checks. Preserve absent and malformed fields and retain all saved inputs and recovery holds.
149 focused tests and 123 workspace-recovery tests pass. Server typecheck, build, all 55 CI checks, and exact-head Apex 5/5 pass.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read native wake-release control context without saved launch payloads. Keep stored input, full legacy context, cancellation evidence, and release decisions. PostgreSQL response-budget regression, 785 lifecycle tests, typecheck, build, and all 55 CI checks pass. Apex reviewed the exact head at 5/5 with no findings.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Native runner sessions keep durable state under
`<instance>/runtime/paperclip-runner/durable-sessions/`
> - When a session starts a fresh provider session, the retired state is
moved into `<session>/continuity-breaks/<epoch>-<uuid>/`; when state
loses authority, it is moved into
`quarantine/<session>.<reason>.<epoch>.<pid>.<uuid>/`
> - Nothing ever removes either, and each can hold a full Codex home of
several gigabytes
> - A sleep-aware sweep already expires failover backup copies in the
same tree
> - This pull request extends that sweep to remove both kinds of
evidence after a retention window, while keeping anything recovery may
still read
> - The benefit is that native runner disk use stops growing without
bound
## Linked Issues or Issue Description
Bug report described in-PR (bug report template).
**What happened?**
On one instance, one day of native runs produced 20 continuity-break
archives totaling 4.8 GB, one session alone holding 4 GB. The quarantine
folder held 12 entries totaling 2.1 GB, the oldest a month old. Neither
is read again by normal operation (continuity-break archives are written
by `archiveContinuityState` and never read), and nothing deletes them.
They were removed by hand to free space and will regrow.
**Expected behavior**
Evidence older than a bounded window is removed automatically, as
failover backup copies already are, without removing anything a live
session or recovery path still needs.
**Steps to reproduce**
1. Run a native Codex session that hits a continuity break (fresh
provider session), or whose state is quarantined (for example
`identity_indeterminate`).
2. Inspect `durable-sessions/<session>/continuity-breaks/` and
`durable-sessions/quarantine/`.
3. Wait any length of time: the entries stay forever.
**Paperclip version or commit:** `master` at e335265568.
**Deployment mode:** any with native runner sessions.
**Agent adapter(s) involved:** native runner (Codex, and any provider
whose state lands in these folders).
## What Changed
-
`server/src/services/native-runtime/native-harness-backup-expiration.ts`:
the existing tick also retires evidence.
- **Continuity-break archives** (`<epoch ms>-<uuid>` under each
session): removed when older than the window. Age is the epoch in the
name, or the folder mtime when that epoch is in the future.
`archiveContinuityState` refuses to archive state with warm-transition
evidence, and no code path reads `continuity-breaks`.
- **Quarantine entries**
(`<session>.<identity_indeterminate|identity_mismatch>.<epoch
ms>.<pid>.<uuid>`): removed by the same age rule. Native cleanup source
archives (`<session>.identity_indeterminate.cleanup.<uuid>`) are
referenced from recovery history, have no epoch, and keep their original
inode's mtime, so they are always kept.
- Every entry is skipped while its session is executing, initializing,
warm, or closing in this process; this is checked again just before
removal. Symlinked entries and symlinked
`continuity-breaks`/`quarantine` folders are reported as `unsafe_path`
and never followed.
- Removal renames to `.expired-<uuid>` in the same folder and then
deletes; later sweeps finish interrupted deletions. An empty
`continuity-breaks` folder is removed.
- Archives, quarantine entries, and leftover deletions count toward
`maxEntries` and `maxRemovals`, and stop on `shouldContinue`. Paging
resumes inside a session's archives and then in quarantine, so a bounded
or dry-run batch always makes progress and never recounts a session's
backup.
- New result fields: `evidenceCandidates`, `evidenceCandidateBytes`,
`evidenceRemoved`, `evidenceReclaimedBytes`, and skip reasons
`recent_evidence` and `cleanup_archive`. The service also exposes
`noteEvidence(writtenAt)`.
- `server/src/services/native-runtime/native-runner-evidence-signal.ts`
(new): a process-local signal. `native-session-executor.ts` raises it
with the entry's epoch right after `archiveContinuityState` creates an
archive and after `quarantineRunnerdStateRoot` moves state into
quarantine; `server/src/index.ts` subscribes the sweep and unsubscribes
on shutdown.
- `server/src/config.ts`:
`PAPERCLIP_NATIVE_RUNNER_EVIDENCE_RETENTION_DAYS` (default `7`, minimum
`1`).
- `server/src/index.ts`: passes the window and logs a batch when it
found evidence candidates.
- `doc/recoverable-disk-cleanup.md`, `doc/DEVELOPING.md`: document the
variable and behavior.
**Sleep-aware background work**
(`rules/sleep-aware-background-work.mdc`):
- *Trigger:* each entry's own expiry deadline (its epoch plus the
window), plus a direct call from the executor when it writes new
evidence. No timer, interval, or polling loop is added; the pass runs
inside the existing sweep's tick and keeps its drain, warm-standby, and
shutdown handling.
- *Source of truth:* the filesystem entries themselves. After a restart
the first sweep reconstructs the next deadline from them.
- *Pending-work predicate / next deadline:* after a full pass the sweep
keeps the earliest due time of the entries it retained; ticks before
that time skip the evidence pass entirely (no readdir, no SQL). A busy
session, a failed removal, or a full batch makes it due on the next
sweep while that obligation remains.
- *Quiet condition:* no evidence, or only evidence not yet due. The pass
never queries the database, so retained evidence never causes recurring
SQL. Dry run reports each candidate once rather than every hour.
No code comments were added.
## Verification
- `pnpm exec vitest run src/__tests__/native-harness-backup-expiration`
(in `server/`, embedded Postgres): 52 passed. New cases:
- old archives and quarantine entries are removed, recent ones kept; the
empty `continuity-breaks` folder is removed;
- age comes from the name, with mtime used for a future epoch;
non-archive folders are ignored;
- a busy session keeps its evidence, including when it becomes busy
after the first check, and it is retried on the next tick;
- dry run reports candidates and bytes, changes nothing (including
leftovers), and reports each candidate once until new evidence comes
due;
- symlinked entries and symlinked evidence folders are skipped without
touching the target;
- an old quarantine entry is removed even when its session's issue is
still open; cleanup source archives are kept;
- the pass stays quiet until retained evidence is due, and a
`noteEvidence` signal schedules newly written evidence;
- inspected archives and leftovers count against `maxEntries`, and
leftover removal stops on `shouldContinue` and resumes next tick;
- a due pass stopped by drain before any session stays pending;
one-entry batches finish without recounting a resumed session;
- interrupted `.expired-*` deletions are finished; paging across ticks
with `maxRemovals`, in normal and dry run, with no recount of the backup
when resuming.
- `native-session-executor.test.ts`: quarantining a legacy root raises
the evidence signal with the epoch in the entry name.
- `native-harness-backup-expiration-config`: default, explicit, minimum,
and fallback parsing.
- **Mutation checks**, each failing at least one test: removing the age
check, the future-epoch fallback, either busy check, the busy retry
deadline, the recent-entry deadline, the due gate, the `noteEvidence`
update, merging the pass deadline, the entry and both folder symlink
checks, the dry-run return, the dry-run leftover guard, leftover
removal, the leftover budget check and its retry deadline, counting
archives against `maxEntries`, the cleanup-archive skip, the
empty-folder removal, the resume cursor, the per-archive budget check,
the resume skip of the backup pass, not charging a resumed session,
keeping the pass active across drain, and the executor's quarantine
signal.
- `pnpm exec vitest run config native-harness
native-workspace-sync-retention native-session-executor server-startup`
(in `server/`): 1272 passed on the previous commit; `config
native-harness native-workspace-sync-retention server-startup`: 450
passed on the head commit.
- `pnpm run typecheck` (in `server/`): pass.
- Not run: server build and boot smoke. The continuity-archive signal
call is exercised only by a runner integration test that needs the
runnerd binary.
## Risks
- **Quarantine as a fail-closed fence:** recovery
(`recoverQuiescentRunnerdState`) and failed-run retry
(`nativeFailedRunRetryStateIsSafe`) treat a quarantined entry for a
session as a reason not to bootstrap fresh. Once such an entry ages out,
a later run of that session can start a fresh provider session instead
of failing with `quarantine_owner_unverified`. This applies only after
the window, matching what backup expiry already accepts for closed work.
- **Signal coverage:** the signal is process-local. Evidence written by
another process sharing the same state directory is picked up when the
next retained deadline is due or after a restart, not immediately.
- **Legacy-path quarantine:** entries quarantined from legacy roots
carry a legacy digest, so the in-process busy check cannot match them;
they are removed by age only.
- **Remote sandboxes:** archives moved aside on a remote runner
(`<root>.continuity-break-<token>`) are outside this sweep.
## Model Used
- Claude Opus 5.5 (`claude-opus-5-5`), 1M context, through Claude Code
with tool use (shell, file edits, 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
#` / `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>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Native runner sessions on managed sandboxes keep a failover backup
of provider state, so a replacement sandbox can resume the session
> - Each backup copies the provider's runtime directory, with a
per-directory sha256 manifest, minus `excludeEntries`
> - Codex already excludes credentials and re-downloadable plugin
caches; OpenCode excluded nothing, and ACPX excluded only per-agent
credentials
> - OpenCode also leaked a temporary HOME on every launch, because
removing it failed on the read-only skill snapshot inside and the error
was swallowed
> - So every backup carried every earlier launch's HOME (with its npm
cache) plus re-installable plugin `node_modules`
> - This pull request removes those homes correctly, sweeps homes left
by earlier versions, and excludes the throwaway entries from OpenCode
and ACPX backups
> - The benefit is much smaller backups and live session directories,
while session history is still backed up and restored
## Linked Issues or Issue Description
Bug report described in-PR (bug report template).
**What happened?**
On one instance, current native failover backups totalled 24 GB across
about 300 sessions. For OpenCode sessions:
- `opencode/<session>/home-XXXXXX/`: 170 per-launch HOME directories
across 67 sessions, 9.3 GB. Each held `.npm` (about 111 MB of npm
cache), `.claude` (the skill snapshot) and `.local`.
- `opencode/<session>/config/opencode/node_modules`: 68 sessions, 4.3 GB
of OpenCode plugin installs.
One ACPX session also carried `acpx/<session>/data/pnpm/store` (229 MB
in the current backup, several GB in older copies).
Cause of the leaked homes: `startRuntime` in the OpenCode driver creates
a fresh `mkdtemp` HOME for each launch and seals the assigned skills
inside it (directories `0555`, files `0444`). `close()` and the
startup-failure paths removed it with `rm(..., { recursive: true, force:
true }).catch(() => undefined)`. That fails with `EACCES` on the sealed
directories, and the failure was ignored, so each launch left its HOME
behind.
**Expected behavior**
Per-launch homes are removed when the runtime closes, and backups hold
only state needed to resume a session. Caches that the provider
re-creates are left out.
**Steps to reproduce**
1. Run an OpenCode native session with at least one assigned skill on a
managed sandbox, over several turns or runs.
2. Inspect `<session>/` in the OpenCode runtime directory: one
`home-XXXXXX` per launch remains.
3. Inspect `failover-backups/current/opencode/<session>/`: the same
homes and `config/opencode/node_modules` are in the backup.
**Paperclip version or commit:** `master` at e335265568.
**Deployment mode:** any with native OpenCode/ACPX sessions on managed
sandboxes.
**Agent adapter(s) involved:** OpenCode (native), ACPX.
## What Changed
- OpenCode driver
(`packages/paperclip-runner/src/drivers/opencode/opencode-server-driver.ts`):
- `removeIsolatedHome` unseals the skill snapshot with the existing
`releaseMaterializedNativeRuntimeSkills`, then removes the HOME. Close
and every startup-failure path use it. It only unseals when `.claude` is
a real directory, not a symlink.
- Per-launch homes are now created under `<session>/homes/home-XXXXXX`,
so one exact path can be excluded.
- When a session opens, `home-*` directories left at the session root by
earlier versions are removed. History is not in them: OpenCode keeps it
in `XDG_DATA_HOME` (`<session>/data/opencode/opencode.db`), and HOME
only holds caches, state locks and the skill copy.
- Exports `opencodeRuntimeSessionDirectoryName` (the existing
session-directory naming) and `OPENCODE_ISOLATED_HOMES_DIRECTORY` for
the server.
- `resolveNativeHarnessPersistenceProfile`
(`server/src/services/native-runtime/native-session-executor.ts`):
- OpenCode excludes `<session>/homes` and
`<session>/config/opencode/node_modules`.
- ACPX (every agent) excludes `acpx/<session>/home/.npm` and
`acpx/<session>/data/pnpm/store`, in addition to the existing per-agent
entries.
- `excludeEntries` are matched as exact relative paths (wildcards are
rejected), so the session directory is computed in the same way as the
existing ACPX entries.
- Exposes the two new exports through the server's runner vendor shim.
No code comments were added.
## Verification
- `pnpm exec vitest run src/drivers/opencode
src/drivers/runtime-context-materializer` (in
`packages/paperclip-runner/`): 108 passed. The 1 failure, `uses provider
structure rather than prose length to select a post-tool answer`, also
fails on unmodified `master`.
- The existing skill-tree test, with and without a startup retry, now
asserts that homes are created under `homes/` and that `homes/` is empty
after close.
- New test: a sealed legacy `home-*` at the session root is removed when
the session reopens, and `data/opencode/opencode.db` is kept.
- `pnpm exec vitest run src/services/native-runtime/
src/__tests__/native-harness-backup-expiration.test.ts` (in `server/`):
5080 passed. `plugin-runner-probe` could not run locally because of a
cargo manifest-path problem in this environment; it is unrelated to this
change.
- Profile tests cover the new OpenCode and ACPX exclusions, including a
session id that needs sanitizing.
- New checkpoint/restore round-trip tests for OpenCode and ACPX. They
use `syncRemoteRunnerDirectoryOut` for the backup and
`stageRemoteRunnerDirectory` for the replacement. Excluded caches are
absent from both the backup and the replacement, while history, config
and workspace files are restored and the live directory is unchanged.
- **Mutation checks**, each failing its test:
- removing the skill unsealing (homes leak again; 3 runner tests fail);
- removing the legacy sweep (1 test fails);
- emptying the OpenCode/ACPX exclusion lists (4 server tests fail).
- Checked by hand against the bundled OpenCode 1.18.34 binary, with the
driver's environment:
- launching with a fresh HOME and no `config/opencode/node_modules`
still lists earlier sessions;
- OpenCode reinstalls the plugin dependencies on launch.
- `pnpm run typecheck` in `server/` and `packages/paperclip-runner/`:
pass.
- Not run: server build and boot smoke.
## Risks
- **Rollout:** existing backups still contain legacy `home-*`
directories. After a restore, the sweep removes them when the session
opens, and the next backup no longer carries them.
- **First launch after restore:** OpenCode reinstalls its plugin
dependencies and npm repopulates its cache. This needs network access to
the npm registry, the same as a brand-new session.
- **Crash leftovers:** a HOME left behind by a hard crash (the runner
killed before `close()`) stays under `homes/` until the session
directory is removed. It is still excluded from backups. Homes under
`homes/` are not swept on open, to avoid touching a HOME that a
concurrently running runtime might still use.
- **pnpm store:** dropping the pnpm store from ACPX backups means the
next `pnpm install` in a replacement sandbox downloads packages again.
## Model Used
- Claude Opus 5.5 (`claude-opus-5-5`), 1M context, through Claude Code
with tool use (shell, file edits, tests, and running the real OpenCode
binary to check what state it reads on resume).
## 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
- [ ] I will address all Greptile and reviewer comments before
requesting merge
Co-authored-by: Paperclip <noreply@paperclip.ing>
Retain pending status deliveries when Native or legacy terminal recovery projects the same state. Preserve completion time on unchanged Native terminal status. Real state, metadata, error, explicit patch, and retained-file changes still notify. The publisher acknowledgement, accepted results, retry deadlines, recovery holds, ownership checks, and workspace cleanup remain intact.
Correct the queued Retry browser fixture race by releasing its old-process guard at the exact Retry request, requiring a successful response, and retaining fixture cleanup after failure.
Validated by 159 focused tests, 653 wider Native recovery tests, six real Chromium/PostgreSQL browser cases, five queued Retry repeats, build/typecheck, exact-head Apex 5/5, and green required CI. The original detached-button CI timeout is documented in the PR. The additional full local repository Vitest command is still running and is not claimed complete.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Apply the existing typed task identity filter in PostgreSQL before returning queued-comment wake controls. Preserve queue selection, ordering, payloads, and current run ownership checks.
All 84 route tests, build, typecheck, exact-head Greptile 5/5, and required CI pass. The old source fails the new deferred and queued byte budgets; corrected reads preserve the complete response and all saved rows.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Reuse one complete company roster for list responses and narrow other-agent eligibility checks to their five consumed fields. Preserve full selected-agent configuration, terminated-manager evidence, current spend, authorization, and fresh reads.
Verified with 170 affected tests, four rejected regression mutations, build and typecheck, exact-head Greptile 5/5, and green required CI.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Keep remote native work under its authenticated connection lease through controller outages. Hold durable ACPX tools through reconnection, align the native Claude HTTP MCP client budget, and wait for the recovered handler before dispatch. Preserve Stop, cancellation, exact identity and artifact checks.
Verified with real-process regressions and a real Claude Opus 5.5 Daytona workflow: 125 seconds offline during useful work, original process and lease retained, no failed native tool calls, queued follow-up completed, independently verified ZIP and committed workspace receipts. All current-head CI gates pass and Greptile is 5/5 with all findings resolved.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Narrow saved-run provenance reads to current identity, scope, and typed legacy bindings. Preserve binding precedence, current viewer checks, and complete stored runs. Real PostgreSQL and route tests pass 23 cases; legacy full-context and type-coercion regressions are caught. Exact-head required CI is green; Greptile is 5/5 with Low impact and no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Omit saved wake payloads from ordinary unlocked Native authority reads. Load the current full wake for mail restrictions. Preserve stored rows, locked mutation reads, and current authorization predicates.
Verified with full build and typecheck, 2,360 affected tests, real PostgreSQL byte and response checks, counterfactual performance and mail authorization failures, complete exact-head CI, and Greptile 5/5 Medium with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Omit nine unused named launch values from the unlocked Native authority query. Preserve complete wake payloads, other context, result receipts, stored input, and locked execution and mutation readers.
Verified with full build and typecheck, 2,359 affected tests across 20 suites, real PostgreSQL byte/retention/cancellation checks, a counterfactual, complete exact-head CI, and Greptile 5/5 Medium with no unresolved comments. An existing cleanup scheduling test passed on CI retry after its suite passed locally with both base and branch source.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read current company/agent profile bindings and rules matching the exact catalog rows being evaluated. Preserve all five selector kinds, null application matches, exclusions, and the existing access evaluator without a cross-call cache.
Verified with full build and typecheck, 652 affected tests across ten suites, all 18 strengthened effective-assignment tests, exclusion mutation and byte counterfactuals, complete exact-head CI, and Greptile 5/5 Medium with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read current ownership, startup, cancellation, task-binding, and adapter evidence without transferring unused saved profiles, launch context, or output. Preserve normal continuation admission, recovery holds, and locked mutation reads.
Verified with full build and typecheck, 537 affected tests across 12 suites, a counterfactual real PostgreSQL byte regression, complete exact-head CI, and Greptile 5/5 Medium with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select the current run fields used by Native tool authority without transferring unused saved profiles or output. Preserve full context, receipts, authorization predicates, and locked mutation reads.
Verified with full build and typecheck, 2,358 affected tests across 20 suites, a counterfactual real PostgreSQL byte regression, complete exact-head CI, and Greptile 5/5 Medium with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Add a separate schema-only provider lookup with current plugin identity and SHA-256 content validation in one statement. Retain bounded schema definitions, return independent objects, and preserve current full driver and secret access paths.
Build, repository typecheck, 485 affected tests, the supported-host database tests, and the complete required CI matrix passed. Greptile scored this exact head 5/5 at medium impact with no unresolved findings.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Use one fresh authorization batch for each live event task and run binding. Preserve current membership, grant, company, and payload disclosure checks.
Build, typecheck, 70 affected local tests, and the complete required CI matrix passed. Greptile scored the exact head 5/5 at medium impact with no unresolved findings. The separate full local suite remains in progress.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read only typed queue display inputs from saved wake payloads and only runtime mode from queued runs. Preserve complete rows for locked edits, discard, steering, and interruption.
Validation: 93 focused route and policy tests, full build and repository typecheck. PostgreSQL fixtures preserve identical queue responses and saved input with less than 4 KB of read values; restoring the full wake read fails at about 1.8 MB. Exact-head Greptile 5/5 (medium), all CI checks green, and zero unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Reuse the existing authorization batch owner for current run-history read decisions after company-scoped existence checks. Preserve missing-task denial, per-request freshness, and the decide-only adapter path.
Validation: 97 focused tests, full build and repository typecheck. A real PostgreSQL fixture returns identical decisions with 4 queries instead of 327; the original path fails the query-bound regression. Exact-head Greptile 5/5 (medium), all CI checks green, and no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select only the run context used by GitHub operation credential checks. Preserve the default complete identity capture for Muse and other consumers, current trust checks, grants, and saved run evidence.
Validation: 72 focused GitHub, identity, Slack, chat-project, and project-route tests; full build and typecheck; exact-head Greptile 5/5 (medium), all CI checks green, and zero unresolved comments. Real PostgreSQL tests prove current revocation and response-size bounds. No live external GitHub push was performed.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read only the current run access fields and typed workspace binding for workspace-operation polling. Keep the current privacy decisions, operation checks, and complete run storage.
Verified focused run-state, live-event, and real database privacy-route tests, full build and typecheck, all CI checks, and exact-head Greptile 5/5 with no unresolved threads.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Apply the existing exact GitHub source marker predicate in the company-scoped connection query. Keep the JavaScript guard and every current install, grant, consent, identity, and credential check unchanged.
Validation: 74 focused tests, full build and typecheck, exact-head Greptile 5/5 (medium), full green CI, and no unresolved threads. Both marker locations, malformed JSON types, changed/disabled connections, and retained unrelated configuration are covered. The unfiltered counterfactual fails both row-count regressions.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Project current chat authorization fields for external review-wait discovery instead of saved execution input. Preserve typed source bindings, nested question receipts, candidate eligibility, live revocation checks, and stored context.
Validation: 181 native chat integration tests, full build and typecheck, exact-head Greptile 5/5 (medium), full green CI, and no unresolved review threads. The full-context counterfactual fails both direct-chat size regressions.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select only the five run fields used by usage receipt persistence under the existing company and row locks. Keep receipt history, replay checks, budget updates, and stored execution evidence intact.
Validation: 37 focused tests, full build and typecheck, exact-head Greptile 5/5 (medium), and full green CI with no unresolved threads. The first CI attempt failed an unchanged Runner shutdown test; its qualified local case and same-head failed-lane rerun passed.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read current control facts and typed issue/adapter bindings for queued-comment polling. Keep full mutation and execution reads, native preparation and steering, authorization, and stored evidence.
Validation: full build and typecheck, 160 focused tests plus two assembled route regressions; exact-head Greptile 5/5 (medium), full green CI, zero unresolved threads. Old-route counterfactual transfers 2.5 MB versus less than 2 KB.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip connects external work to an agent and its ordinary tasks.
> - A dedicated GitHub App must prove signed webhook delivery during
setup.
> - Cloud queues the original bytes until the tenant can authenticate
them.
> - An older personal connection can leave a long empty-queue backoff in
that tenant.
> - Active App setup can then time out before the queued ping is read.
> - This change lets active setup request the existing event drain with
a short rate limit.
> - The same webhook handler still owns signature verification and setup
state.
## Linked Issues or Issue Description
Refs: #17457, #17513.
**What happened?**
Cloud accepted a new App's ping, but the tenant read it only after the
setup wait timed out and cleanup archived the draft. An existing
personal GitHub connection had left an empty-queue backoff. The poller
checked that deadline before it read the new App state.
**Expected behavior**
An active manager setup request must be able to read the queued ping
promptly. It must not bypass signature checks or create another App.
**Steps to reproduce**
1. Let a personal GitHub connection build its empty-queue backoff.
2. Start a dedicated App and complete its installation.
3. Return to the waiting setup screen while the signed ping is queued.
**Paperclip version or commit**
This behavior is present on the base commit of this PR.
## What Changed
- Let the authenticated wizard request a queue drain only while its
Cloud App still lacks signed-delivery verification. Existing manager,
credential, and installation checks run first.
- Register a narrow callback to the server-owned GitHub poller. Keep
provider handling in the GitHub services. Remove the callback on server
close.
- Share one in-flight attempt between setup and background requests.
Limit setup attempts to one per five seconds, including after errors.
- Retain the existing passive backoff, sealed queue, acknowledgement
rules, and canonical webhook signature handler.
- Add regressions for the old backoff, concurrent requests, failed
requests, and wizard signature verification. Update the setup
documentation and map.
## Verification
- Observed the backoff regression fail against the old service: active
setup made no additional Cloud lease request. It passes with this fix.
- Three focused event-poller tests pass. They use the real
admission/backoff code with stubbed SQL and Cloud boundaries.
- All 39 startup tests pass. Server TypeScript, map checks, and diff
checks pass.
- The Postgres integration suites could not start locally because this
host exhausted its semaphore limit. They remain required CI checks. This
is not a passing local integration result.
- Full current-head CI passes, including the Postgres and chat
integration suites. Greptile scored this exact head 5/5 with no
unresolved findings.
- The additional Paperclip review check ended neutral because its
execution did not complete. It did not publish a validated assessment.
- A fresh deployed setup attempt remains required before live
qualification.
## Risks
A waiting setup screen now makes a bounded readiness request to Cloud.
The existing connector request timeout still applies. Concurrent calls
share the same drain, and failures release it for a later retry. The
request stops after persisted verification or when the browser stops
advancing setup. No timer, durable queue, fleet setting, or sleep
blocker is added or removed. The existing background reconciliation path
still has its existing sleep restrictions. Live setup-to-review behavior
is not established by these local tests.
## Model Used
OpenAI Codex, based on GPT-6. The exact runtime model ID and
context-window size are not exposed in this session. Used for code
execution and regression 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 <paperclip@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work.
> - The native runtime keeps a durable runner alive across controller
restarts.
> - A restart can begin while a provider session is still opening.
> - The old controller fenced read-only run admission before it could
publish that prepared session.
> - Admission failure then closed the runner that shutdown needed to
detach.
> - This pull request lets only the existing startup finish admission
within the bounded shutdown wait.
> - The controller detaches that exact session before it can submit a
turn. New work and state changes stay fenced.
## Linked Issues or Issue Description
**What happened?**
A controller restart during native provider startup could close the
prepared runner with `native control-plane run admission failed`.
Recovery then found no running process and placed the task in a
reconciliation hold. The same admission path follows both fresh startup
and replacement after a failed provider resume.
**Expected behavior**
Shutdown must detach the existing prepared session so the next
controller can recover it through the normal identity checks. Shutdown
must not admit a new executor, submit a turn, or override an explicit
Stop.
**Steps to reproduce**
1. Delay a native backend while it opens its provider session.
2. Request controller restart detachment for that run.
3. Let backend startup finish before the 30-second deadline.
4. Observe that read-only admission fails and shutdown reports the run
as inactive instead of detached.
**Additional context**
Related work: #17073 handles remote process-monitor classification;
#17459 handles macOS process fingerprints; #17097 covers broader hot
upgrade recovery. This change fixes the startup admission fence. It does
not relax process identity or provider launch-profile checks.
## What Changed
- Separate the read-only admission guard from the guard on control-plane
state changes.
- Let shutdown await only the startup already registered for the run.
Close its admission window when the wait ends.
- Reject a new execution after its run is fenced for shutdown.
- Detach the prepared session and exit before runtime-question
attachment or provider turn submission.
- If Stop is pending before or after publication, retain its exact
cancellation acknowledgment and cleanup owner across authority reads and
runtime-question attachment.
- Share one 30-second admission/Stop-settlement deadline across the
restart sweep. Report an unsettled run as pending while healthy runs
continue hot-restart adoption.
- Cover the real runtime and admission path, deadline expiry, new-start
rejection, persisted binding checks, and write fences.
- Document the bounded startup handoff in the restart ownership
contract.
## Verification
- Red: the new actual-runtime/admission regression failed on the
previous code. Shutdown reported the run as inactive and did not detach
it.
- Red: review regressions reproduced detach before a post-publication
Stop acknowledgment and an actual hot-restart preparation error when
startup Stop cleanup timed out.
- Green on `e10d0cb238`: 18 startup/restart tests passed, including Stop
during authority reads and runtime-question attachment, the real
recovery preparation and shutdown coordinator with a stalled Stop plus
healthy peer, and the shared sweep deadline. The earlier 36-case
Stop/restart selection also passed before this follow-up.
- Both complete changed test files previously passed on head
`8e9c2610fe`: 857 tests. They used real PostgreSQL 18 with a fresh
database per fixture. An earlier local harness attempt reused one
database and hit duplicate fixture keys; that harness was corrected
before this complete pass.
- Real PostgreSQL: 7 admission and revocation tests passed. These cover
wrong runner identity, persisted binding mismatch, checkpoint/result
rejection, and in-flight write revocation. The local test harness used a
fresh Docker PostgreSQL 18 instance because macOS embedded Postgres
exhausted the host semaphore limit. Only database provisioning was
replaced; assertions and production port code were unchanged.
- Transport safety acceptance: the current source transport controlled a
real Linux ARM runner and a deterministic Codex test provider in a
disposable Debian container. After startup, detach and generic cleanup
preserved the process. A new controller authenticated the same PID and
provider session. The provider log had exactly one session start and no
turn start. Duplicate launch was forbidden, the retained workspace was
unchanged, and no stop or suspend command was issued.
- The Linux acceptance check tests the shared native transport safety
contract. It is not a live Claude response, a hosted Daytona run, or
recovery of the original task. Those remain unverified. No original
task, sandbox, credentials, or permissions were changed.
- `pnpm -r typecheck` and `pnpm build` passed before the review
follow-up. Server typecheck passed on current head `e10d0cb238`;
current-head CI also passed the full typecheck and build gates.
- The full local `pnpm test:run` command was attempted, then stopped
after repeated embedded PostgreSQL bootstrap setup failures. That broad
local run is incomplete. The current-head four-file Docker PostgreSQL
run passed 879 tests, with one existing warm-copy `claim-uncertain` case
timing out and recovery failing to import because the temporary
provisioning mock omitted a support export. The harness is corrected and
affected cases are being rechecked. The timeout is unexplained; it is
not being labeled an infrastructure failure. No assertions are skipped
by the database provisioning replacement.
- All 55 clean-host CI checks passed on current head `e10d0cb238` (2
conditional skips), including all general test shards and native runner
verification. Greptile reviewed that exact head at 5/5 with no findings.
The previous Paperclip review findings are addressed in [the review
response](https://github.com/paperclipai/paperclip/pull/17511#issuecomment-6391520242);
its new advisory review remains queued.
## Risks
- Admission must stay read-only. All writes retain the original
synchronous controller fence, and both admission checks still validate
the exact persisted binding.
- Startup that exceeds the 30-second shutdown wait still fails closed.
It cannot claim a successful detach or start a turn.
- No schema migration or provider configuration change is included.
## Model Used
OpenAI Codex, based on GPT-6, with reasoning, repository tools, code
execution, and test execution. The exact runtime model ID and context
window 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>
Skip repeated successful native release checks for the same unchanged source within one recovery pass. Preserve every saved request, failure retry, cursor update, source change, legacy path, and current admission gate.
Verified with 201 focused tests, 32 full-control queue cases, build and typecheck. Exact-head CI is green, Greptile is 5/5 with Low impact, and no review threads remain unresolved.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Read current native Stop identity, acknowledgement, and process evidence without saved execution inputs, profiles, excerpts, or unrelated result data. Preserve the complete nested acknowledgement and every existing lease, remote-termination, process, ownership, and admission check. The physical-stop helper body is unchanged.
Verified with 101 tests across the affected suites, the full build, and typecheck. The review fix passed 62 focused tests. Large valid saved inputs stay below 2 KB at the raw driver boundary; large malformed string and array results stay below 1 KB and remain stored. CI is green and exact-head Greptile is 5/5 at Medium impact with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Connection readiness reads current permitted connection identities without unrelated catalog rows or complete profile-rule payloads. Reuse profile precedence and selector logic, preserve full public/runtime summaries, and keep installed connections, identity checks, and gateway authorization current.
Verified with 106 focused tests, the full build, and typecheck. The direct catalog inventory regression fails at 102 rows when the scope filter is removed and passes at two rows with the filter. One unrelated CI timing assertion differed by 1 ms; the exact local test and the single failed shard rerun passed on unchanged source. CI is green and exact-head Greptile is 5/5 at Medium impact with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Decline native legacy-continuation checks before loading saved execution input. Preserve exact nonblank task bindings, skip reasons, complete legacy recovery inputs, company scope, and retry accounting.
Verified with 105 recovery tests, the full build, and typecheck; the binding review fix passed 73 affected tests and server typecheck. CI is green and exact-head Greptile is 5/5 at Medium impact with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Native wake-queue release reads omit saved profiles and excerpts. Legacy release reads keep their complete inputs. Ownership, cleanup, and file retention checks remain unchanged.
Verified with 275 wake-queue tests, the full build, and typecheck. All CI checks pass. Greptile scored the exact head 5/5 with Medium impact and no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Nested project repositories are now worktrees of one shared clone
per project repository, and existing standalone clones are converted in
place
> - Conversion keeps the clone's remote-tracking refs as task-owned
branches (`paperclip/<id>/remotes/...`) so commits only they hold
survive
> - Expiration then required every task-owned branch, including those,
to be on the remote
> - Remote-tracking refs for branches deleted or force-pushed upstream
never are, so converted checkouts were kept forever
> - This pull request treats those refs the way the standalone-clone
path already does: commits missing upstream go into the restore point
instead of blocking
> - The benefit is that converted workspaces can expire and free disk
without losing any history
## Linked Issues or Issue Description
Refs #17407, #17382, #17289. Bug report described in-PR (bug report
template).
**What happened?**
On one instance, 17 closed workspaces in a two-repository project were
kept on every expiration sweep with
`unpushed_managed_checkout_branches`. Each nested checkout is a
converted worktree whose task-owned refs include
`paperclip/<id>/remotes/origin/<branch>` for upstream branches since
deleted or force-pushed (dependabot, squashed feature branches). Their
commits are on no remote, so `provenNestedWorktreeHistory` refused. The
task's own branch was fully pushed.
**Expected behavior**
As in the standalone-clone path (#17289), remote-tracking commits
missing upstream are preserved in the restore point, and only genuinely
local branches block expiration.
**Steps to reproduce**
1. Fetch an upstream branch into a standalone nested clone, then delete
that branch upstream.
2. Convert the clone into a worktree (nested clone migration).
3. Close the issue and let expiration inspect it: kept with
`unpushed_managed_checkout_branches`.
**Paperclip version or commit:** `master` at cb47dbb2da.
**Deployment mode:** any.
**Agent adapter(s) involved:** not adapter-specific (core).
## What Changed
- `provenNestedWorktreeHistory` in
`server/src/services/git-workspace-restoration.ts`:
- splits task-owned refs into those under `<prefix>remotes/` and the
rest;
- still requires the rest (and their reflogs) to be on the remote;
- returns commits under `remotes/` that are missing upstream, and their
reflogs, as `preserve`, which the restore point bundles. This matches
the standalone-clone path.
- `restorePointHistoryDigest` in
`server/src/services/git-restore-point.ts`: for a linked worktree, it
fingerprints the task's own refs and their reflogs instead of the shared
clone's `refs/remotes`. Other tasks change those whenever they fetch,
and they never covered the task-owned remote refs. It loads
`nested-repository-worktrees` lazily to avoid an import cycle.
No code comments were added.
## Verification
- `pnpm exec vitest run nested-clone-migration` (in `server/`). New
cases:
- A converted checkout whose only extra history is a remote-tracking
commit deleted upstream expires with a restore point. After its refs are
removed and the shared clone is garbage-collected, reopening restores
the commit.
- The digest ignores the shared clone's remote refs and changes when the
task's own refs move.
- **Mutation checks**, each failing its test:
- reverting the prover split;
- dropping the preserve list;
- reverting the digest change.
- The existing case where a converted checkout with a genuinely unpushed
branch is kept still passes.
- `pnpm exec vitest run git-restore-point nested-repository-worktrees
nested-clone-migration git-workspace-expiration
git-reproducible-managed-checkouts isolated-task-directory
isolated-repository-rehome git-credentials execution-workspace` (in
`server/`): 591 passed.
- `pnpm run typecheck` (in `server/`): pass.
- Not run: server build and boot smoke.
## Risks
- **Misclassified branches:** a local branch literally named
`remotes/...` in a standalone clone becomes `<prefix>remotes/...` after
conversion and is now preserved in the restore point instead of blocking
expiration. Its commits are still saved, not lost.
- **Leftover refs:** after expiry, the task's other owned refs stay in
the shared clone, as they did before this change, so their commits stay
in two places.
## Model Used
- Claude Opus 5.5 (`claude-opus-5-5`), 1M context, through Claude Code
with tool use (shell, file edits, tests, read-only diagnosis on a live
instance).
## 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
- [ ] I will address all Greptile and reviewer comments before
requesting merge
---------
Co-authored-by: Paperclip <noreply@paperclip.ing>
Select the complete workspace-sync record and native status wake identifier for reconciliation candidates. Preserve candidate policy, JSON types, full later evidence, and stored data. The committed replay regression includes the real reconciliation scan. Full build, typecheck, 114 focused tests, complete CI, and exact-head Greptile5/5 pass.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select current chat authorization fields for committed responses under the existing scoped share lock. Preserve complete evidence for actual retries, source revocation, JSON types, and saved run data. Full build, typecheck, 113 focused tests, complete CI, and exact-head Greptile5/5 pass.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Select fresh metadata for committed finalization and cleanup. Preserve full saved evidence for new finalization, cancelled workspace retention, and legacy Stop. PostgreSQL byte-budget and unchanged-row regressions pass, full CI is green, and Greptile scored the exact head 5/5.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
Keep more caller groups in one bounded database response capture and verify public caller paths in the production bundle. All local checks and CI passed; Greptile scored the exact head 5/5 with no unresolved comments.
Co-Authored-By: Paperclip <noreply@paperclip.ing>
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - Git cache expiration, isolated source re-home, nested clone
migration and workspace reopen fetch from the workspace's remotes to
prove history is on the remote or to rebuild workspaces
> - They built git credentials from the issue id alone, but the managed
GitHub identity is selected from the run's responsible user and agent
> - So the credential never resolved, they fell back to anonymous
access, and private repositories were unreadable
> - This pull request resolves credentials with the issue's responsible
user and assigned agent, the same identity agent runs use
> - The benefit is that workspaces with private repositories can expire,
re-home, migrate and reopen like public ones
## Linked Issues or Issue Description
Refs #17464, #17407, #17369. Bug report described in-PR (bug report
template).
**What happened?**
On one instance, 18 long-closed workspaces in one project were kept on
every expiration sweep with `managed_checkout_upstream_unavailable`.
Their nested checkout is a private GitHub repository. Expiration called
`createGitRemoteAuthProvider(db, companyId, { issueId })`. With that
context, `resolveManagedGitHubCredential` returned "No managed GitHub
identity is available for this run". `withAnonymousGitFallback` then
tried anonymous access, which cannot read the private repository. Called
with the issue's `responsibleUserId` and `agentId`, the same lookup
returned a credential.
**Expected behavior**
Background workspace jobs authenticate as the issue's identity, as agent
runs do, and can verify private repositories.
**Steps to reproduce**
1. Configure a project with a private GitHub repository and a managed
GitHub identity tied to the issue's responsible user.
2. Close an issue whose workspace includes that repository and let git
cache expiration inspect it.
3. It is kept with `managed_checkout_upstream_unavailable`.
**Paperclip version or commit:** `master` at fe85fd7ac4.
**Deployment mode:** any with managed GitHub identities.
**Agent adapter(s) involved:** not adapter-specific (core).
## What Changed
- `createIssueGitRemoteAuthProvider(db, companyId, issueId)` in
`server/src/services/git-credentials.ts`:
- looks up the issue's `responsibleUserId` and `assigneeAgentId` once,
lazily, scoped to the company;
- builds `createGitRemoteAuthProvider` with `{ issueId,
responsibleUserId, agentId }`;
- falls back to the issue-only context when the issue is missing.
- Callers switched from the issue-only context:
- git cache expiration (`expirationGitAuth`);
- isolated source re-home;
- nested clone migration;
- both credential sites in
`reopenClosedIsolatedExecutionWorkspaceForIssue`.
- Anonymous fallback for public repositories is unchanged.
No code comments were added.
## Verification
- `pnpm exec vitest run src/__tests__/git-credentials.test.ts
src/__tests__/git-expiration-auth-fallback.test.ts` (in `server/`): 32
passed. New cases:
- the helper passes the issue's responsible user and agent to the
credential lookup;
- it looks the issue up once across calls;
- it falls back to the issue-only context when the issue is missing.
- **Mutation check:** dropping the identity fields from the context
fails two cases.
- **Live instance:** for an affected workspace, the managed credential
resolved only with the responsible user and agent; with the issue id
alone, or with standing delegation, it did not.
- `pnpm exec vitest run git-credentials git-expiration-auth-fallback
git-workspace-expiration git-restore-point isolated-repository-rehome
nested-clone-migration` (in `server/`): 296 passed.
- `pnpm run typecheck` (in `server/`): pass.
- Not run: server build and boot smoke.
## Risks
- **Whose identity is used:** background jobs now use the issue's
responsible user's managed GitHub identity, which agent runs on that
issue already use. They only read: fetch, ls-remote and clone for
verification and restore.
- **Unchanged failure behavior:** if the identity is unavailable (for
example, the user is no longer a member), the existing anonymous
fallback applies, as before.
## Model Used
- Claude Opus 5.5 (`claude-opus-5-5`), 1M context, through Claude Code
with tool use (shell, file edits, tests, read-only diagnosis on a live
instance).
## 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
- [ ] I will address all Greptile and reviewer comments before
requesting merge
Co-authored-by: Paperclip <noreply@paperclip.ing>
