StandupBot

A GitHub Action that writes your PR descriptions for you. It runs the real diff and commits through an LLM you choose, and fills in a structured Summary / Changes / Testing description plus up to three labels — so you stop hand-writing PR bodies and standup updates.

Bring your own model: it talks to any OpenAI-compatible endpoint (OpenAI, OpenRouter, Ollama, LM Studio, …). No provider, URL, or model is hardcoded.

---

Sample output

Generated end-to-end by StandupBot for a real PR (nodejs/node#64573) from its actual diff, using a real LLM endpoint. This is the description it produced verbatim:

Summary

Add lchownSync to the VFS implementation so that symbolic link ownership can be changed without following the link, matching the behavior of fs.lchownSync.

Changes

• doc/api/vfs.md: added lchownSync(path, uid, gid) to the list of VFS API signatures.
• lib/internal/vfs/file_system.js: added synchronous lchownSync method to VirtualFileSystem and updated the async wrapper to call provider.lchownSync.
• lib/internal/vfs/provider.js: added default lchownSync method to VirtualProvider (delegates to chownSync) with JSDoc comment.
• lib/internal/vfs/providers/memory.js: implemented lchownSync that updates the uid/gid of the link entry itself.
• lib/internal/vfs/setup.js: changed the lchownSync handler to invoke vfs.lchownSync instead of vfs.chownSync.
• test/parallel/test-vfs-lchown-symlink.js: new test verifying sync, callback, and promise variants of lchownSync on VFS-mounted symlinks.

Testing

The new test exercises fs.lchownSync, fs.lchown (callback), and fsp.lchown (promise) on symlinks inside a VFS mount, asserting correct uid/gid changes.

Alongside the body it also returns a normalized title and labels (clamped to bug | feature | chore | docs | refactor).

---

Use it in your repo

Add .github/workflows/standupbot.yml:

name: StandupBot
on:
  pull_request:
    types: [opened, synchronize]

permissions:
  pull-requests: write
  contents: read

jobs:
  describe:
    runs-on: ubuntu-latest
    steps:
      - uses: XenoCyber0/StandUpBot@main
        with:
          llm-base-url: ${{ secrets.LLM_BASE_URL }}
          llm-api-key: ${{ secrets.LLM_API_KEY }}
          model: ${{ vars.LLM_MODEL }}

Set LLM_API_KEY (and LLM_BASE_URL) as repo secrets; LLM_MODEL can be a repo variable.

It only writes when the PR description is empty (or already carries the StandupBot marker) — it never clobbers a hand-written description.

Inputs

All optional in YAML; each overrides its env var, which in turn overrides .standupbot.yml. A run needs a base URL, a key, and a model from somewhere or it fails fast.

Input	Env var	Notes
llm-base-url	LLM_BASE_URL	e.g. https://api.openai.com/v1, OpenRouter, …
llm-api-key	LLM_API_KEY	secret — env/input only, never in config
model	LLM_MODEL	no default — must be provided
github-token	GITHUB_TOKEN	defaults to the workflow's ${{ github.token }}

Input names map to the UPPER_SNAKE INPUT_* / LLM_* env vars (llm-base-url → INPUT_LLM_BASE_URL, model → INPUT_MODEL). The Action then falls back to LLM_BASE_URL / LLM_MODEL / .standupbot.yml.

---

Configuration (.standupbot.yml, optional)

Place at the repo root. Env vars override it.

tone: concise        # free-form tone steer
maxDiffTokens: 6000  # diff budget before per-file summarisation kicks in
exclude:             # gitignore-style globs kept out of the prompt
  - package-lock.json
  - '**/*.lock'
  - '**/dist/**'
llm:                 # non-secret connection details (apiKey is rejected here)
  baseUrl: https://openrouter.ai/api/v1
  model: anthropic/claude-3.5-sonnet

See .standupbot.example.yml.

---

Known limitations

• "Testing" claims can be inferred, not evidenced. The model sometimes writes a Testing section that reads as if code was built/run when it wasn't (e.g. a doc-only change claiming it "builds and renders correctly"). Verification statements in summaries are not guaranteed — treat them as a draft. Known issue — prompt tightening planned.
• Diffs larger than maxDiffTokens (default 6000 tokens ≈ 24KB) fall back to a map-reduce summarisation path (per-file summaries, capped at 8 LLM calls) instead of embedding the full diff. Works, but not yet load-tested.
• The fetched diff itself is hard-capped at maxDiffTokens × 4 characters (≈24KB by default) before summarising, so the tail of an enormous PR may not influence the output.
• GitHub PR reads are unauthenticated (~60 req/hr) unless you pass a github-token.

---

License

MIT