npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

plugin-workflow-git-review

v1.0.5

Published

Native Workflow integration for Git Manager code reviews.

Readme

Git Review Workflows

This plugin integrates Git Manager reviews with native NocoBase Workflow, Workflow Manual and AI plugins. It requires NocoBase 2.2.40 or newer in the 2.x series, collection triggers, workflow timeouts, AI provider services and native manual nodes, and plugin-git-manager 1.3.0 or newer. The setup page reports missing runtime capabilities and does not provision a graph when they are unavailable.

Setup

Install and enable Workflow, AI, Git Manager, and this plugin. Configure a Git review flow, repository access, an enabled AI employee, and an available LLM service/model. Open Git Review Workflows in plugin settings. Existing flows are provisioned automatically when their bindings can be resolved; unresolved flows remain disabled and display their setup blockers. A flow with manual publication needs one or more valid assigned users. When available and authorized for AI workflow tasks, the original flow creator is used as the initial approver; no administrator is substituted automatically. Set explicit approver IDs on the setup page if needed.

Choose Quick for a correctness pass and synthesis, or Full for separate correctness, security, change-impact, and synthesis passes. Create a new revision to change the starter or approval bindings. Provisioning writes a disabled graph and its binding transactionally, then activates the revision only after its dependencies are rechecked. It does not overwrite existing or edited workflow graphs. Open the workflow from the setup page to inspect it. Create a disabled revision before editing a graph; the integration validates and freezes that revision when it is enabled. Changes to a legacy flow's model, publication policy, or approval bindings require a new revision.

Both templates use the native collection CREATE trigger on the persisted gitReviewRequests record. Git Manager first reserves the canonical project; the first node claims the request for that execution and obtains an immutable Git context. The request pins a workflow revision and Git SHAs. A periodic check recovers a missed collection event and reports terminal execution states. A first-node claim prevents duplicated events from owning the same request.

An LLM job that has already persisted its terminal result is also the durable evidence for a missed workflow continuation. Reconciliation may resume that existing job after checking its request owner, project reservation, deadline and downstream progress; it does not call the model again. An unresolved or uncertain model request is not restarted automatically. Access to managed execution and job results, including expanded workflow associations, also requires access to the review's Git repository.

The review passes and synthesis use individual Git review LLM nodes. They use the native AI provider and model configuration, message roles, and structured-output contract. The node persists its pending workflow job before invoking the provider, so even an immediate provider response can be recorded. These review nodes accept text messages; image and file inputs are not supported. The existing employee's static role text and model selection are retained. Dynamic variables in the employee role text block setup until an administrator supplies a static equivalent; user-specific employee prompts and employee tools are not run by these templates. Flow instructions and per-run instructions remain part of the review context. The result node requires structured output from each pass, checks finding paths and lines against the supplied target-side diff, and saves the resulting Markdown with a SHA-256 fingerprint. Collector limitations are included in the result. These checks do not prove that the model's conclusions are correct or that tests ran. A manual-publication flow pauses at a native manual task. Assigned users with repository access can inspect the exact saved review and submit an approval or rejection from Git review approvals in the modern Workflow task center. Publication requires the same execution's approved task and matching fingerprint. Git Manager rechecks the current merge-request head before publishing.

Managed review graphs use four integration node types: Git review context, Git review LLM, Validate Git review, and Publish Git review. Native manual approval remains a separate node. Graphs using the general native llm instruction are not accepted for managed reviews, because its provider response can arrive before the pending job is persisted. Create a new revision from setup when upgrading a graph that uses the earlier node. Existing edited or executed graphs are not overwritten.

Cutover and recovery

A fresh installation can start without an operator cutover. If legacy reviews exist, a root administrator must stop every old review worker and its AI calls before confirming cutover on the setup page. Confirmation alone does not release a running or uncertain legacy review. The page lists pending, quarantined and orphaned reviews; an administrator can record recovery only after verifying the underlying AI work has stopped. A timeout, cancellation, process restart, provider error, or elapsed time is not proof that a model request stopped, so these outcomes do not automatically release the reservation. A different workflow failure can release it only when persisted jobs establish successful completion of every started review LLM call. A verified manual rejection can release it after the same settlement proof. Review recovery may require operational investigation of workers and provider requests.

Legacy publication holds are listed separately. An administrator may release a hold only after the prior outbound request has stopped and settled. Releasing the hold does not assert whether that request posted a comment; use the review history to check the provider before any retry.

The project reservation is shared by the legacy migration and new requests. Waiting for human approval retains it. A queued inbox item binds to a validated current revision when dispatched; once its request is created, that revision is immutable for the request. Direct execution, job resume and rerun routes are blocked for managed reviews. The Git Manager service owns admission, repository authorization, source collection, publishing, cutover state, and project reservation; this plugin uses that service rather than running the previous review worker.

Validation boundary

A successful plugin build or workflow setup does not establish provider availability, browser behavior, successful approval, external publication, or production readiness. Exercise the configured model and provider, manual approval, repository permissions, stale-head handling, and recovery in the target environment before relying on the integration.