commit-insights

Generate a local git contribution dashboard (static HTML) from your repo's commit history. Optional AI-written narrative summaries via BYOK/BYOM providers.

npm install -g commit-insights
commit-insights generate .
open dashboard.html

Features

• Self-contained dashboard — single dashboard.html, dark theme, Chart.js charts, works offline
• Zero network by default — no data ever leaves your machine unless you explicitly opt in
• Optional AI narratives — aggregated stats sent to Anthropic, OpenAI, or local Ollama (--narrative)
• Incremental — SQLite cache reuses historical data, only extracts new commits
• Git-native — respects .gitignore, lives inside .git/ for cache, uses git log -z under the hood

Privacy

commit-insights makes zero external network calls unless you pass --narrative. The core dashboard (statistics, charts, tables) is generated entirely from your local git history. No data is collected, no telemetry, no analytics.

When AI narratives are enabled (--narrative), only aggregated statistics (commit counts, type breakdowns, top tickets) are sent to the configured provider — never raw commit messages, diffs, or code.

Usage

commit-insights generate [path]                 generate contribution dashboard
commit-insights cache status|clear [path]       inspect or clear cache
commit-insights config [--json] [--explain]     show effective config

Options:
  --author <pattern>          filter commits by author
  --all                       bypass cache, full reconciliation
  --out <file>                output path (default: dashboard.html)
  --narrative                 include AI-written narrative summary
  --narrative-audience <type> manager | retro | resume | self
  --narrative-length <len>    short | normal
  --strict                    exit non-zero if AI narrative fails
  --no-cache                  skip cache, re-extract everything
  --cdn-charts                load Chart.js from CDN (requires internet)

Configuration (precedence)

Lowest → Highest:

1. Built-in defaults
2. Repo config (.commit-insights.json — team-shared: AI provider/baseUrl/model, areas, ticket patterns)
3. User config (~/.config/commit-insights/config.json — personal: AI provider/model)
4. Environment variables (ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY, AI_MODEL, AI_BASE_URL, OLLAMA_HOST)
5. CLI flags

User config beats repo config — your AI provider preference shouldn't be forced by a team config file.

API keys are read from a .env file in the current working directory (gitignored via .gitignore). Never commit API keys to your repo config.

{
  "ai": {
    "provider": "ollama",
    "model": "llama3.2"
  },
  "areas": {
    "src/components/": "UI",
    "src/api/": "Backend"
  },
  "ticketPattern": "PROJ-\\d+"
}

AI providers

| Provider | Default model | Key required | |----------|--------------|--------------| | Anthropic | claude-sonnet-4-20250514 | ANTHROPIC_API_KEY | | OpenAI | gpt-4o | OPENAI_API_KEY | | Gemini | gemini-2.0-flash | GOOGLE_API_KEY | | Ollama | llama3.2 | No key (local) |

Default models can be overridden via config (model) or AI_MODEL environment variable.

Development

git clone https://github.com/hinedy/commit-insights
cd commit-insights
npm install

# Run from source
npm run dev -- generate .

# Build
npm run build

# Test
npm test

Why not git-insights?

The tool is named for commits, not git — it works on any repo's commit history regardless of VCS, and the name avoids implying it covers all git operations (blame, reflog, etc.). The scope is contribution insight, not git insight.

MIT © 2026 hinedy