Files
PaperClipAI/docker/daytona-runner
06bc8db75c fix(runner): install verified API artifacts in execution workspaces (#18494)
## Thinking Path

> - Paperclip manages AI agents and their work.
> - Native agents use authorized API tools to read saved results.
> - Copying a large response as model-generated text can lose bytes.
> - The execution workspace needs an exact copy of the saved artifact.
> - This change copies the authorized bytes without shell API
credentials.
> - The runtime verifies size and SHA-256 before it reports success.
> - A live evaluation checks capture origin, installation and delivered
bytes.

## Linked Issues or Issue Description

**What happened?**

A native agent lost one byte while copying a 24576-byte API text page.
Existing capture and paging controls passed. The runtime had no verified
saved-artifact-to-workspace copy tool. A separate historical screening
refusal does not establish capture corruption.

**Expected behavior**

An authorized native run can install the complete saved artifact. The
stored and installed bytes must match in size and SHA-256. A conflicting
file must remain unchanged. A lost publication receipt must not cause an
unsafe repeated write.

**Steps to reproduce**

Run
`server/src/services/native-runtime/artifact-transfer.integration.test.ts`
on baseline `c46b8f35add9255fc71be294d27f5a7865bcb47f`. Its capability
assertion fails because `download_artifact` is absent. Apply this PR and
run the same regression with the filesystem and API controls. It
verifies 388131 synthetic bytes through the authority and installation
boundary.

Refs #18429. Related #17393, #17086 and #17285 cover paging, transfer
foundations and remote output reads. This PR adds the inbound
installation contract.

## What Changed

- Add `download_artifact` for supported standard-mode native runs, with
a 10 MiB limit.
- Reuse current company, task, run, asset-content and responsible-user
access checks. Check authority again before publication.
- Bind installation to the current execution lease. Stage through
confined Linux descriptors and publish with an atomic create-only link.
Keep API credentials out of execution commands.
- Verify bytes at storage and execution boundaries. Reject unsafe paths,
symlinks, different existing files, truncation and corruption. Reconcile
an unknown publication through verification only; an identical explicit
retry remains safe.
- Add API documentation, access-map entries and content-free audit
records.
- Preserve only the pinned Daytona SDK numeric HTTP status when failed
creation cleanup is unconfirmed. Keep provider exception messages and
headers private.
- Keep only the finished runner binary and portable provider pack in
their builder layers. Remove Rust targets/downloads and the Node install
workspace/store before export. Report tree sizes.
- Add one explicit Codex/Daytona Product E2E cell. Its independent
oracle checks the server capture audit, redacted tool receipts, stored
source, delivered attachment and browser link after reload. Uploaded
substitutes and one-byte loss fail.

## Verification

- Merge update at `6766b46553cab5b9787ade87a002eccaf50e68fc`: merge
current master and preserve both explicit suites in the catalog. Focused
catalog, image and oracle checks: 122 pass. Filesystem and authority
integration: 20 pass. Maps, module boundaries and whitespace checks
pass. New-head CI and Greptile are running. Full local build and
typecheck remain unavailable because cargo is absent.

- Behavioral red/green: the absent-tool regression failed before
implementation and passed afterward. The initial missing-package attempt
is recorded as setup failure.
- Linux filesystem and real PostgreSQL/authority integration: 19
filesystem tests and 1 authority test pass. Existing API authority,
capture and paging controls: 70 pass, 2 skipped. Provider transport in
these integration controls is simulated.
- Qualification calibration: 118 focused tests pass. The real
persisted-event redaction case and a byte-identical uploaded substitute
each failed before their oracle repair. The unchanged controls then
passed. Harness typecheck passes.
- Full recursive typecheck and build pass locally on the integrated
production source at `086f7a2901`. Later commits change only
qualification tests and documentation.
- Structural map/module checks and whitespace checks pass.
- Full local `pnpm test:run` was attempted by the implementation worker
and stopped under its 8 GiB memory limit, with no full-suite verdict.
Current-head CI must supply the full test result.
- The complete local Product E2E support suite has 13 failures. The
unchanged pre-qualification baseline reproduces all 13: host PostgreSQL
semaphore capacity, stale timeout/profile/workflow inventory
expectations, and expired fixed-date allowance fixtures. This is not a
full local suite pass.
- Earlier CI at `086f7a2901` failed a browser test during cleanup
cancellation with a database lock timeout. It remains recorded. All
browser shards passed on both later fully checked heads.
- Two qualification campaigns were canceled during setup before provider
execution while correcting the oracle. They provide no live product
verdict.
- Previous-head CI at `28fa001457f59df1c300c016dc569049e72d0061`: 53
checks pass, 3 intentional selection skips; no failed or pending checks.
- Real image/export regression: campaign 41272351780 compiled the Linux
runner and passed root/unprivileged CLI and transport checks, then
failed BuildKit export with `no space left on device` (exit 102,
filesystem 99.99%). No provider cell ran. Removing the Rust build trees
alone was insufficient: campaign 41273604815 also failed during
provider-pack export with exit 102. Commit `5bbb549c54` also
consolidates dependency installation and pack construction so the
disposable workspace/store can be removed in the same layer. All 122
image/oracle/catalog checks pass. Image and remote-pack publication then
passed in campaign 41275196741 after cleanup changes plus a scoped
60-to-120 GiB expansion of only its delete-on-termination builder disk.
The VM terminated and the expanded volume was deleted. This does not
establish that cleanup alone makes the default 60 GiB builder
sufficient.
- The Daytona admission probe then failed before agent creation, with
unconfirmed provider cleanup. The queued continuity case was canceled
because it shares this prerequisite. No artifact-transfer or continuity
success is claimed.
- Status-diagnostic regression: the original generic wrapper failed the
HTTP-status assertion. After the repair, the full plugin suite passes
501 tests, with 7 skips; its standalone TypeScript check passes.
Controls reject invalid status values, avoid accessors, preserve the
original create status and exclude credential canaries.
- Previous-head Greptile: 5/5; its summary names
`28fa001457f59df1c300c016dc569049e72d0061`. All review threads are
resolved.
- Live artifact transfer and three-turn warm continuity: NOT VERIFIED.
The diagnostic case in
https://github.com/paperclipai/paperclip/actions/runs/41277598289
confirmed Daytona HTTP 401 before agent creation. The configured
`DAYTONA_API_KEY` in the existing `runner-e2e-paid` environment needs
verification or refresh. The queued continuity control was not run after
its shared admission prerequisite failed. No further provider attempt is
planned until the credential changes. Unconfirmed provider cleanup
remains recorded.

## Risks

- Source-provenance changes can repeat the provider dependency install.
The combined build layer trades install-cache granularity for lower
retained disk use.
- Linux descriptor semantics and Node are required on the execution
host. Unsupported providers and restricted runtime profiles fail closed.
- Storage reads buffer at most 10 MiB. A provider outage can leave a
reserved staging file. Partial bytes never become the destination.
- A lost publication receipt can leave a complete file with an unknown
result. Recovery verifies the existing file without installation.
Conflicts are never overwritten.
- Artifact contents remain untrusted. The copy does not bypass screening
or authorize downstream execution.
- Live qualification uses the governed Codex API profile. It does not
establish every model, subscription profile, platform or historical
failure cause.

## Model Used

OpenAI GPT-6 in Codex assisted with diagnosis, code, tests and
qualification. The desktop runtime does not expose a more specific model
deployment ID or context-window size. The model used reasoning, source
inspection, tools and code execution. The native implementation agent is
configured with OpenAI `gpt-6.1-sol`.

## Checklist

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

---------

Co-authored-by: devinfoley <139239+devinfoley@users.noreply.github.com>
Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-02 08:27:09 -06:00
..

Paperclip Daytona runner image

Ordinary provider packs now include pinned Cursor assets for Linux x64; no --candidate-providers=cursor flag is needed. Cursor admission is enabled after the production readiness gates passed. The public CLI installation command is paperclipai runtime setup cursor; it runs explicitly on the execution host and is never an npm installation hook.

This image is the Paperclip Cloud fleet sandbox image plus a source-built paperclip-runnerd and immutable provider pack. The pack contains Node 24.11, OpenCode 1.18.34, the compiled OpenCode proxy, ACPX 0.13.1 sidecar, qualified ACP agents, and the production lockfile. Its manifest digests each executable bridge and records the runner source revision as provenance. Its interface declaration allows compatible app releases to reuse it without uploading or installing a pack.

The fleet pins are intentionally copied from paperclip-cloud/fleet-sandbox-image/Dockerfile. Update both definitions together until the fleet base is published as a stable image that this Dockerfile can extend directly.

Harness versions

The October 2026 refreshes pin Codex 0.160.0, Claude Agent SDK 0.3.286 (Claude Code 2.1.286), and OpenCode 1.18.34 in the shared provider pack. Claude Code 2.1.280 is the minimum for Opus 5.5 and 2.1.284 is the minimum for Sonnet 5.5; the pinned runtime also supports Fable 5.1. Codex 0.160.0 ships bundled metadata for the current GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna model IDs. The sandbox layer pins Grok CLI 1.0.70, Gemini CLI 0.65.0, Kimi Code 2.3.3, Cursor CLI 2026.10.22-27dcf5f, and GitHub CLI 2.103.3 (October 2026 refresh). The native Cursor provider uses its separately verified distribution, pinned to 2026.09.26-dd393fe; it does not use the sandbox layer’s global Cursor executable. Grok CLI 1.0.70 supports the current Grok 4.7 model family. Hermes stays at 0.19.0, the newest release on PyPI.

Keep the patched ACP bridge versions separate from their CLI runtime pins. Their executable digests do not change when only the runtime dependency changes. Refresh the runtime executable digests from integrity-verified npm release archives for every supported platform, and keep the native runner, provider manifest, and remote controller version checks aligned.

Build and verify

Run pnpm --filter @paperclipai/paperclip-runner test:opencode:qualification after installing dependencies to exercise the actual pinned OpenCode executable. It checks health/version, session creation and retrieval, SSE messages, an async prompt, and session deletion against a loopback mock provider. It uses an isolated home, starts no paid model request, and retires its process group. Set PAPERCLIP_TEST_OPENCODE_BINARY to the materialized Linux executable when qualifying an assembled provider pack.

The fleet image is currently amd64-only because the pinned Cursor and GitHub CLI checksums cover amd64.

content_id="$(pnpm --silent test:e2e:runner:image-id)"
docker buildx build \
  --platform linux/amd64 \
  --build-arg PAPERCLIP_RUNNER_CONTENT_ID="${content_id}" \
  --build-arg PAPERCLIP_RUNNER_SOURCE_REVISION="$(git rev-parse HEAD)" \
  --tag "paperclip-daytona-runner:e2e-content-${content_id}" \
  --load \
  --file docker/daytona-runner/Dockerfile \
  .

docker run --rm --platform linux/amd64 \
  --entrypoint paperclip-runnerd \
  "paperclip-daytona-runner:e2e-content-${content_id}" \
  --build-metadata

The metadata must advertise dial_ws_loopback, dial_wss, and listen_ws. The explicit entrypoint is needed only for this local probe because Daytona's base image uses its own long-running sandbox entrypoint.

test:e2e:runner:image-id hashes the audited Docker build dependency closure, target platform, the immutable Dockerfile syntax-frontend digest, and every immutable FROM reference. It fails before the paid workflow can build when the frontend or a base is not pinned to a sha256 digest. When updating the syntax version, resolve and review its registry digest and update both values in the first Dockerfile line. Git commits that do not change those inputs reuse the same content tag.

PAPERCLIP_RUNNER_SOURCE_REVISION remains the full Git SHA that built the first published copy and is retained as provenance rather than cache identity.

Use in Paperclip

Publish the image to a registry Daytona can pull, or use the environment editor's Configure image flow to produce a Daytona snapshot. Set the environment image to that immutable tag or snapshot. Paperclip probes the sandbox user's PATH for paperclip-runnerd and codex and checks /opt/paperclip-runner/provider-pack for OpenCode and ACPX. It uses the pack when its selected provider interface and platform are compatible. App source revisions and unrelated provider pins do not force an update. Startup reads the small manifest in the existing runner metadata query. It does not scan the pack, run a version command, or transfer files on this path. Otherwise it stages the pack configured by PAPERCLIP_RUNNER_REMOTE_PROVIDER_PACK_PATH. Remote OpenCode and ACPX never fall back to host-local processes.

src/live/provider-pack-contract.json in the Runner package declares the OpenCode and ACPX interface revisions. Change a revision only for an incompatible runner/pack boundary change, and qualify the new pack before publishing it. Bumping a revision also retires every older image pack. The pack build also records a contractDigest, computed by scripts/contracts/contract-fingerprints.mjs from the JSON contracts under protocol/ and generated/, from Rust serde shapes, payload keys, match literals and env keys, and from the Runner's ACPX patch, rather than from source bytes. An image pack is reused only when its digest equals the controller's bundled pack; comments and refactors leave it unchanged, so a mismatch means a contract changed and the controller uploads its bundled pack. Packs without a digest are not reused. Define a new runner/pack contract in a schema under protocol/ and generate the TypeScript from it, as the sidecar's session.open parameter allowlist is generated from acpx-sidecar.schema.json; TypeScript-only definitions are not part of the digest. Dependency versions and Git SHAs remain build provenance. ACPX retains its selected agent's qualified profile checks.

The image build makes the pack root-owned and non-writable to the execution user. Fast admission requires protected ancestors and manifest, with no writable path alias. Task-owned packs and older packs without an interface declaration keep full content verification. An attached session retains the exact pack and daemon identity admitted at launch. Publish one new image with the interface declaration to enable reuse across later compatible app releases.

Do not promote paperclip-runner-e2e-20260826-v2 for OpenCode or ACPX. Build a new immutable image or snapshot from a clean committed revision and pass that full Git SHA as PAPERCLIP_RUNNER_SOURCE_REVISION.

Do not bake provider credentials, Paperclip bootstrap tickets, or Daytona preview tokens into this image. They remain per-run secret material.

The provider-pack build pins the official Linux x64 Node 24.21.0 image by manifest digest. Its bundled Undici is 7.29.1, which fixes GHSA-3wwx-pv8p-q78v. Pi separately verifies its private Node executable and nested npm dependency; changing the outer interpreter does not replace either provider-owned pin.

Provider CLI updates are manifest-only changes: repository CI owns the root lockfile. Resolve the complete workspace manifest graph in the build context before invoking Docker, matching CI when a source commit precedes the lockfile bot. The trusted workflow supplies this resolved lockfile as an immutable artifact. The complete resolved lockfile must match PAPERCLIP_RUNNER_LOCK_SHA256 before package installation or lifecycle execution. Review and refresh that digest with source dependency changes; registry-time resolution drift fails closed. The Product E2E workflow resolves one lockfile before the image build. It verifies the downloaded artifact, then passes that artifact's SHA-256 as the PAPERCLIP_RUNNER_LOCK_SHA256 build argument. The Dockerfile checks the resolved lock against this value before installation. The fixed Dockerfile default is for standalone builds; it must not replace a campaign's verified lock digest. Refresh the default from the clean tracked lockfile using the exact pnpm install --resolution-only --ignore-scripts --no-frozen-lockfile command, and verify a second resolution preserves the digest. A lockfile left by a filtered or incremental install can retain stale importer patch identities. Refresh exact runtime versions and qualification digests together; do not download dependencies when a task starts.

Candidate ACP qualification assets

Provider branches can build their pinned assets with node packages/paperclip-runner/scripts/build-provider-pack.mjs /absolute/pack --candidate-providers=cursor (or copilot or pi). The source revision must include the named provider's builder. Assets are installed at build time under provider-assets/<provider>/<platform>-<architecture>. The pack manifest binds each complete asset tree. Runtime admission separately checks the provider's source-owned closure pins and copies a verified launch snapshot. A pack with candidate assets does not qualify or enable that provider.

For an isolated Linux x64 Daytona qualification image, pass --build-arg PAPERCLIP_RUNNER_CANDIDATE_PROVIDERS=cursor with the normal build arguments. Compute its content ID with the same selection: pnpm --silent test:e2e:runner:image-id --candidate-providers=cursor. Candidate assets and the default pack have distinct image identities. Never reuse the default image's content ID for a candidate build. Use each provider branch's recorded version and explicit model. Keep credentials out of images. Paid qualification requires bound provider and Daytona credentials, inspectable spend, and the shared $100 ceiling recorded in the capability report.

The pack builder tests its copied Node interpreter after relocation. Use a standalone Node distribution if the host interpreter depends on a package manager's private shared libraries. Pi additionally pins its complete portable interpreter and npm dependency graph. Refresh exact runtime versions and qualification digests together; never download dependencies when a task starts. Grok's additive native ACP profile keeps its qualified 1.0.13 executable at the verified sandbox prerequisite path. It does not replace the legacy adapter's grok command on PATH.

Native Grok is an image prerequisite at /opt/paperclip/providers/grok/1.0.13/grok. Its checksum-verified provisioning is separate from the provider pack, which ships only the built-in launcher. Public npm installation never downloads this binary.