What you shipped for your org. A tool that fetches GitHub commits for an organization and time period (quarter or custom range), for one or more usernames, and shows a report: commit list with diffs plus a written summary (technical initiatives + feature/business work). Built for individuals and teams (including mob programming).
Concept: Dispatch = the report (what was shipped). User(s) = who dispatched. Org = for whom it was dispatched.
- Time period — Quarter + year (custom range planned)
- Single or multiple users — One username or comma-separated list (e.g.
alice, bob) for team/mob - Org + optional repos — Scope by organization; optionally restrict to specific repos (comma-separated)
- Real diffs — Commit list with file-level changes and inline patch viewer
- Noise filtering — Ignores lockfiles and common noise files
- AI or rule-based summary — Optional AI-generated report (Gemini or Ollama) or pattern-based (rule-based) summary
- Export/Import — Export as JSON or Markdown; re-import JSON without refetching
- Regenerate summary — Switch AI analyser without refetching GitHub data
npm install
npm run devOpen http://localhost:5173.
Create a token at github.com/settings/tokens with:
repo— Read commits and diffs (including private repos)read:org— List organization repositoriesread:user— Used by GraphQL for contribution data
The app stores the PAT in your browser’s localStorage; it is only sent to GitHub’s API from your device.
- Enter your GitHub PAT, organization, and at least one GitHub username (or several, comma-separated).
- Optionally enter repositories (comma-separated); leave blank to use all org repos or repos the user(s) contributed to.
- Choose quarter and year.
- Choose AI analyser: Rule-based (default), Ollama (local), or Gemini (API key).
- Click Fetch dispatch.
For detailed usage, AI setup, multi-repo, and multi-user use cases, see docs/USER_GUIDE.md or open Documentation in the app header.
| Option | Description |
|---|---|
| Rule-based | No AI. Summary from commit message patterns and file types. Fast, private, works offline. |
| Ollama | Local models (e.g. llama3.2). No API key; run OLLAMA_ORIGINS="*" ollama serve and pull a model. |
| Gemini | Google’s API. Get a key at aistudio.google.com; free tier is sufficient. |
If you switch only the AI analyser after a fetch, the app re-runs the summary without refetching from GitHub.
- Leave “Repositories” blank — For a single user, the app uses GitHub GraphQL to list only repos they contributed to in the date range. For multiple users, it uses all org repos and fetches commits per user per repo.
- Enter repo names (e.g.
backend, frontend) — Only those repos are scanned.
- Enter comma-separated usernames (e.g.
alice, bob, charlie). The app merges their work: commits are deduplicated by SHA. Use cases: mob programming, team reports, or a single report for a small group in the same org/period.
npm run buildOutput is in dist/. Deploy as a static site to Vercel, Netlify, Cloudflare Pages, or GitHub Pages.
- CONTEXT.md — Project context for developers and LLMs: layout, types, data flow, where to change what, conventions, APIs. Read this first when working on the codebase.
- In-app — Use the Documentation link in the header for Ollama/Gemini setup, multi-repo, multi-user, and export/import.
- Repo — docs/USER_GUIDE.md for the full user guide.