Files
PaperClipAI/packages/plugins/sandbox-providers/daytona
Devin FoleyandPaperclip 0d3e7bf6ac fix(daytona): keep commands alive after log stream closure (#14799)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - Sandbox providers run agent processes and send their output to the
host.
> - The Daytona SDK can close a log socket while the remote command
still runs.
> - The driver treated a clean socket close as completion before it had
a command exit code.
> - This pull request recovers log observation for the same command and
waits for a recorded exit.
> - The host keeps receiving new output without a second command
dispatch.

## Linked Issues or Issue Description

**What happened?**

A clean close of the Daytona session log WebSocket resolves the SDK
callback promise. If the command still runs, the driver returned
`exitCode: null` with `timedOut: false` after a short status check. A
streamed ACP bridge can then report a process disconnect.

**Expected behavior**

A log socket close must not complete a running command. Recovery must
preserve new output, the caller's lifetime controls, and one command
dispatch.

**Steps to reproduce**

Use the callback form of `getSessionCommandLogs`. Let that promise
resolve while `getSessionCommand` still has no exit code. Keep the
command running, then expose its final logs and exit code. The new
regressions exercise this sequence, including streams that run longer
than the provider operation timeout.

Related: #11021, #11049. #14485 covers input delivery retries, which are
a separate transport path.

## What Changed

- Require a recorded command exit after a clean log-stream close.
Reconnect once, then read status and full log snapshots at most once per
second.
- Forward new output during recovery. Remove replayed prefixes and
reconcile a final snapshot so bytes written after socket close are
retained.
- Hold a trailing UTF-8 replacement suffix until replay or completion
resolves it. This handles the SDK decoder flush when a socket closes in
the middle of a character.
- Preserve healthy initial and reconnected stream lifetimes. Bound each
recovery read. Preserve the existing fallback timeout budget after
rejected stream attempts.
- Retain partial output on timeout. A log-observation timeout reports an
unconfirmed result and preserves any observed exit code in metadata; it
does not synthesize successful completion.

## Verification

- `pnpm exec vitest run --config
packages/plugins/sandbox-providers/daytona/vitest.config.ts`: 335
passed, 14 gated tests skipped.
- `pnpm exec vitest run --project @paperclipai/plugin-daytona`: 335
passed, 14 gated tests skipped.
- `pnpm exec tsc --noEmit -p
packages/plugins/sandbox-providers/daytona/tsconfig.json`: passed.
- `pnpm exec tsc -p
packages/plugins/sandbox-providers/daytona/tsconfig.json`: passed.
- `pnpm --workspace-concurrency=1 -r typecheck`: passed.
- `CARGO_BUILD_JOBS=2 pnpm --workspace-concurrency=1 -r build`: passed.
- `pnpm test:run`: exited with failure after 845.77 seconds. The server
phase reported 43 failed files, 529 passed, and 163 skipped; 14 failed
tests, 9,122 passed, and 5,599 skipped. All failure entries were traced
to local PostgreSQL startup/cleanup errors or ten macOS skill-cache
rename errors. The package helper restored 17 missing PostgreSQL library
links; the four sequencing/migration tests and fourteen
native-workspace-finalizer tests then passed. The ten cache failures
match clean-base evidence with identical source/test blobs. This is not
a full-suite pass; later local test groups did not run. PR CI supplies
the complete check result.
- Regressions cover clean and rejected stream recovery, two hour-long
streams with a five-minute operation budget, live fallback output, one
dispatch, delayed final output, stale snapshots, SDK UTF-8 decoding,
bounded observation, and timer cleanup.
- `git diff --check` and a local diff scan for secrets and private
identifiers passed.
- Greptile reviewed head `293c4dbe67` at 5/5. Both prior findings are
fixed, both threads are resolved, and no new actionable findings remain.

## Risks

- The SDK returns full snapshots with no offset API. After both stream
attempts end, polling bandwidth grows with retained output. The
one-second cadence limits request frequency.
- The SDK callback stream has no cancellation handle. Existing caller
stop logic and provider/session teardown still own its lifetime. Late
callbacks from a settled stream are ignored.
- A successful status read with no exit code keeps recovery active under
the existing caller guard. A failed read stops recovery; it is not
retried indefinitely.
- No command replay, provider API change, schema migration, or runtime
timeout policy change is included.

## Model Used

OpenAI GPT-6 through Codex, with tool use and independent code review.
The exact serving model identifier is not exposed in this session.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [ ] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-09-30 18:37:54 -07:00
..

@paperclipai/plugin-daytona

Published Daytona sandbox provider plugin for Paperclip.

This package lives in the Paperclip monorepo, but it is intentionally excluded from the root pnpm workspace and shaped to publish and install like a standalone npm package. That lets operators install it from the Plugins page by package name without introducing root lockfile churn for Daytona's SDK dependencies.

Install

From a Paperclip instance, install:

@paperclipai/plugin-daytona

The host plugin installer runs npm install into the managed plugin directory, so transitive dependencies such as @daytonaio/sdk are pulled in during installation.

Configuration

Configure Daytona from Instance Settings -> Environments, not from the plugin's plugin page.

  • Put the Daytona API key on the sandbox environment itself.
  • When you save an environment, Paperclip stores pasted API keys as company secrets.
  • DAYTONA_API_KEY remains an optional host-level fallback when an environment omits the key.
  • Optional apiUrl and target settings map directly to the Daytona SDK/client configuration. If apiUrl is omitted, the Daytona SDK uses its default endpoint.

Notes:

  • The current published Daytona SDK package is @daytonaio/sdk.
  • The driver supports both snapshot-based and image-based sandbox creation. If both are set, validation rejects the config as ambiguous.
  • Each cold create uses a unique provider name and ownership labels. If creation fails after Daytona has allocated a sandbox, the driver looks up that exact name, checks every ownership label, and waits for deletion. Failed, missing, or timed-out lookups and failed deletions report unconfirmed cleanup with the provider name; they never return a usable lease. A missing name lookup after an uncertain create is not proof that a delayed provider request cannot create a resource.
  • For plugin-backed sandbox lease acquisition, unresolved creation cleanup crosses the worker RPC as a validated ownership envelope. The host records pending_cleanup before retrying deletion. If only a creation name is known, the provider first returns an ownership-verified sandbox ID without deleting it. The host saves that observation before authorizing deletion. Its existing cleanup sweep and durable spool preserve retries across controller restarts and environment deletion. A missing observed ID confirms cleanup after a lost deletion reply or database update; a name that has never been observed remains unresolved. Provider exceptions and resolved credentials are excluded from that envelope. This requires the matching host and plugin SDK update; probe and custom-image interactive-setup calls still report unconfirmed immediate cleanup without this lease-recovery path.
  • Reusable leases map to Daytona stop/start semantics. Non-reusable leases are deleted on release. A provider-resolved target does not change the identity of an existing sandbox. Release closes the same scoped lease that a later sentinel-verified resume reopens.
  • A session log socket can close while its command still runs. The driver requires a recorded command exit before returning a completed execution. It reconnects the log stream once, then polls status and log snapshots at most once per second. Replayed output is removed by byte offset, new output continues to reach the host, and the command is never dispatched again. Healthy initial and reconnected streams retain their existing lifetime under the caller’s RPC/run guard. After a clean close, each status or snapshot read gets a fresh operation timeout; successful recovery polling does not impose a new command lifetime. When both stream attempts fail without a clean close, fallback retains its original timeout budget starting after the stream attempts.
  • The SDK returns full log snapshots, so fallback bandwidth grows with retained output; it has no paged snapshot or stream cancellation API. A recovery-read timeout keeps partial output and explicitly reports whether command exit remains unconfirmed; it does not prove the remote command stopped. Provider/session teardown remains responsible for closing an outstanding socket. A final snapshot reconciles bytes missed by the stream before returning a recorded exit.
  • A sandbox record can survive the loss of its underlying container. Resume treats it as expired only when a fresh provider read confirms the exact missing-container error for that sandbox and marks it unrecoverable. Unknown errors and failed confirmation reads preserve the lease. The host still requires a verified native-runner backup before replacement.

Local development

cd packages/plugins/sandbox-providers/daytona
pnpm install --ignore-workspace --no-lockfile
pnpm build
pnpm test
pnpm typecheck

These commands assume the repo root has already been installed once so the local @paperclipai/plugin-sdk workspace package is available to the compiler during development.

Package layout

  • src/manifest.ts declares the sandbox-provider driver metadata
  • src/plugin.ts implements the environment lifecycle hooks
  • paperclipPlugin.manifest and paperclipPlugin.worker point the host at the built plugin entrypoints in dist/