Jannik/PaperClipAI
1
0 Fork 0
mirror of https://github.com/paperclipai/paperclip.git synced 2026-10-06 10:48:12 +02:00
Code Issues Packages Projects Releases Wiki Activity
Files
25c422ba7e30f9e4bfb34ddeff88ce4bd6b2773e
PaperClipAI/packages/paperclip-runner/docs/tutorials/local-runner.md
T
Dotta 560e7e48b5 feat(runner): add SDK and developer tooling (#12608) 
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - The runner package already provides the production protocol and
execution spine.
> - Contributors still need stable SDK surfaces, deterministic test
tools, and local inspection tools.
> - Those surfaces share generated contracts and must change as one
package boundary.
> - This pull request adds the package-local SDK, labs, examples, and
drift checks.
> - The benefit is a reviewable developer platform that does not change
application execution selection.

## Linked Issues or Issue Description

**Subsystem affected**

`packages/paperclip-runner` — runner SDK, conformance tools, and
developer tooling.

**Problem or motivation**

The production runner spine is present, but package consumers cannot
build deterministic integrations, inspect sessions, or verify
provider-neutral behavior through supported surfaces.

**Proposed solution**

Add browser, React, standalone, live-session, scenario, conformance, and
evaluation surfaces. Add generated contract inventories and
package-local verification scripts. Keep production application routing
unchanged.

**Alternatives considered**

We considered splitting each generated catalog, SDK surface, and demo
into separate pull requests. Those changes share exports, fixtures, and
drift gates. Splitting them would create intermediate package states
that do not build.

**Roadmap alignment**

No overlapping item appears in `ROADMAP.md`. This work extends the
runner package that is already on `master`.

## What Changed

- Add browser, React, standalone, live-session, and issue-thread SDK
surfaces.
- Add deterministic mock control-plane, scenario, conformance, replay,
and evaluation tools.
- Add bounded Codex, OpenCode, and ACPX development transports and
fixtures.
- Keep deferred managed-provider execution fail-closed. Persisted
compatibility data remains readable.
- Add generated capability inventories with their source files and drift
checks.
- Add examples, package documentation, browser checks, and
clean-consumer checks.
- Preserve the reviewed protocol bounds, replay compatibility aliases,
process environment isolation, and semantic redaction limits.
- Update the ACPX package patch that the existing workspace patch
registry already tracks.
- Do not change `pnpm-lock.yaml`, repository workflows, server runtime
selection, or the application UI.

## Verification

GitHub Actions is the verification authority for this pull request. The
repository CI, package TypeScript and Rust checks, package tests,
generated-output drift checks, browser checks, security scans, and
Greptile review must pass on the exact head.

Local test suites were not run because this series uses parallel GitHub
Actions for verification.

## Risks

This is a large greenfield package change. The main risks are public
export drift, generated-output drift, and optional React consumer
compatibility. Package boundary checks, clean-consumer checks, and
browser tests cover those risks. Production adapter selection and server
execution are outside this pull request.

## Stack

1. **This PR:** runner SDK and developer tooling.
2. [Codex production server
integration](https://github.com/paperclipai/paperclip/pull/12616).
3. [Provider-neutral task-thread
UI](https://github.com/paperclipai/paperclip/pull/12617).

## Model Used

OpenAI Codex, GPT-5, high-reasoning mode, with tool use and code
execution.

## 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 described the issue in-PR following the feature request
template
- [x] I have not referenced internal/instance-local Paperclip issues or
links
- [x] My branch name describes the change and contains no internal
Paperclip ticket id
- [ ] 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 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
2026-08-31 21:33:11 -05:00

5.0 KiB
Raw Blame History

Local runner: Run the Local Runner and Fake Harness

What this phase is

Local runner is a local live-run path. A TypeScript mock core starts a Rust runner. The Rust runner starts a scripted fake harness. The processes use JSON lines over stdio.

What this phase proves

This phase proves that a native session can run from start to finish without Paperclip, a network provider, or a real model. It proves cleanup, request resolution, interruption, bounded logs, one terminal result, and live/replay parity.

Prerequisites

  • Node.js 24.11 or newer
  • pnpm 9 or newer
  • a stable Rust toolchain with cargo
  • a Chromium-compatible browser

Start from a clean repository checkout and run every command from the repository root. Install the package dependencies and browser:

pnpm install --filter @paperclipai/paperclip-runner --lockfile=false --ignore-scripts --dev
pnpm --filter @paperclipai/paperclip-runner exec playwright install chromium

On a minimal Linux host, install the browser libraries too:

pnpm --filter @paperclipai/paperclip-runner exec playwright install-deps chromium

playwright install-deps requires root. On a Debian or Ubuntu host where root is unavailable, use the package's rootless verification command instead. It downloads the required browser-library packages into a user-owned cache, extracts them without installing system packages, and scopes LD_LIBRARY_PATH to the verification process:

pnpm --filter @paperclipai/paperclip-runner verify:rootless

1. Run the package verification path

pnpm --filter @paperclipai/paperclip-runner verify

Expected: Rust, TypeScript, protocol, boundary, documentation, and browser checks pass. The final Local runner line reports one terminal event and a successful semantic result.

2. Run the happy path in the CLI

pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario happy-path --quiet

Expected summary facts:

  • eventCount is 19;
  • terminalCount is 1;
  • semanticResult is done;
  • harness and runner exit codes are 0.

Capture the live trace as a validated replay fixture:

# Recorded evidence generation is deferred from this release.

The command writes packages/paperclip-runner/.paperclip-local/evidence/local-runner-happy-path.json.

3. Run the scripted control flows

pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario permission-input --quiet
pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario interrupted --quiet
pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario error --quiet
pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario duplicate-terminal --quiet
pnpm --filter @paperclipai/paperclip-runner trace:local-runner -- --scenario happy-path --duplicate-turn-command --quiet

Expected:

  • permission and input finish with done;
  • interruption exits the harness with 130 and reports yielded;
  • scripted error exits the harness with 7 and reports yielded;
  • duplicate terminal still reports terminalCount: 1;
  • duplicate turn command does not start a second turn.

4. Use the live browser

Start the package-local server:

pnpm --filter @paperclipai/paperclip-runner browser:dev --host 127.0.0.1 --port 4179

Open http://127.0.0.1:4179, then follow these steps:

  1. Keep Happy path selected and choose Start local run.
  2. Confirm the terminal badge says succeeded.
  3. Confirm Harness process exit is 0 and Semantic result is done.
  4. Confirm Live and replay reducer output says Match.
  5. Select Permission and input, start the run, and choose Allow.
  6. Enter a trace name and choose Send input. Confirm the run completes. Each interactive request remains open for five minutes. If it expires, the page reports that the run finished and asks you to start a new run.
  7. Select Interruption, start the run, and choose Interrupt turn when the button becomes active.
  8. Confirm the terminal badge says cancelled, the semantic result says yielded, and the timeline contains one run.terminal event.
  9. Switch to Static replay and confirm the Replay fixture path still works.

Stop Vite with Ctrl+C.

The browser regression suite writes its temporary screenshots under the ignored packages/paperclip-runner/test-results/ directory. It never rewrites the committed protocol fixtures.

5. Inspect the fixtures

What this proves

  • the Rust runner owns and cleans the fake-harness process group;
  • the fake driver emits deterministic live events and requests;
  • logs are bounded;
  • process exit and semantic result stay separate;
  • duplicate commands and terminal proposals do not repeat effects;
  • every browser event passes the Replay validator and reducer;
  • replay after completion produces the same final snapshot.

Durable recovery must not begin until the Local runner human checkpoint is accepted.

Reference in new issue View Git Blame Copy Permalink