git-fork-recon

Summarise key changes in forked repositories.

Synopsis:

git(hub) repository 🠲 pull forks 🠲 ✨LLM✨ 🠲 summary report

Why ?

A popular repository may have many forks.

Most of them are pointless.

A handful have just the bugfix or feature that matters to you.

Through the dark magic of large language models ✨, git-fork-recon helps find these interesting forks.

Features

• Filters and prioritizes forks based on number of commits ahead of parent, starts, recent activity, PRs. Ignores forks with no changes.
• Use locally hosted or remote LLMs with an OpenAI-compatible API.
• Local caching of git repositories and forks (as remotes)
• Detailed Markdown reports with:
  • Repository overview
  • Analysis of significant forks
  • Commit details and statistics
  • Links to GitHub commits and repositories
  • Overall summary of changes highlighting bugfixes, new features and innovations in the most interesting forks
• Local server with a web UI and REST API

Quickstart

The first time you run git-fork-recon, it will start the configuration wizard (see Configuration below). You'll need a Github Access Token and details of an OpenAI-compatible endpoint.

(using uv for convenience)

# Install uv if you haven't already (or use pip and a virtualenv etc if you prefer)
curl -LsSf https://astral.sh/uv/install.sh | sh

# This will start the first-time configuration wizard, or show --help
uvx git-fork-recon

# Once configured, analyse a specific repository
uvx git-fork-recon https://github.com/DunbrackLab/IPSAE

This will generate a Markdown report in the current directory ({username}-{repo}-forks.md).

Tip: you can view the report in the terminal like:

uvx frogmouth DunbrackLab-IPSAE-forks.md

Web interface

To run the simple local web UI:

uvx --from 'git-fork-recon[server]' git-fork-recon-server

Go to http://localhost:8000/ui to see the web UI.

Installation (quick)

uv tool install 'git-fork-recon[server]'

Now you can run: git-fork-recon or git-fork-recon-server like any other command.

Installation (development)

Quick: using uv sync (automatically creates a .venv and installs server and dev dependencies)

# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

uv sync --all-extras

source .venv/bin/activate

Note: The first time you run uv sync, it will create a uv.lock file for reproducible builds. This file should be committed to version control to ensure all developers use the exact same dependency versions.

Or: using uv with manual venv creation:

# Create and activate a new virtual environment
uv venv
source .venv/bin/activate

# Install the package in editable mode, with optional server and dev dependencies
uv pip install -e '.[server,dev]'

Configuration

Configuration is stored in a TOML file located in the platform-specific config directory (e.g., ~/.config/git-fork-recon/config.toml on Linux). On first run, an interactive setup wizard will guide you through configuration.

First-Time Setup

When you run git-fork-recon for the first time, you'll be prompted for some configuration values.

You will need:

1. A Github API token to READ repositories and their forks.

• Go to https://github.com/settings/tokens and make an access token with permissions: public_repo, user:email - provide the key when prompted.

2. Access to an OpenAI-compatible endpoint. You can use:

• OpenRouter (paid but cheap, also some free models: https://openrouter.ai/settings/keys),
• Google AI Studio (https://aistudio.google.com/app/apikey)
• a local server (Ollama, llama.cpp server, LM Studio etc.)
• any remote OpenAI-compatible endpoint (e.g. OpenAI, Cerebras, Groq etc.)

First-time configuration wizard options:

• OpenAI-Compatible Endpoint: Choose from:
  • Environment variable (OPENAI_BASE_URL)
  • Local Ollama server (http://localhost:11434)
  • OpenRouter
  • Google AI Studio
  • Custom URL
• Endpoint API Key: Enter your API key or choose to always use an environment variable
• Model: Select a model based on your chosen endpoint
• GitHub Token: Enter your GitHub token or choose to always use an environment variable
• Cache Directories: Configure repository and report cache locations (defaults to $HOME/.cache/git-fork-recon/repos and $HOME/.cache/git-fork-recon/reports)

Config File Structure

The config file (config.toml) has the following structure:

# Configure an OpenAI-compatible endpoint
[endpoint]
base_url = "https://openrouter.ai/api/v1"
api_key = "sk-..."
model = "deepseek/deepseek-v3.2"
# context_length = 64000  # Optional: Override default context length

[github]
# Get a token https://github.com/settings/tokens with permissions: public_repo, user:email
token = "ghp_..."

[cache]
repo = "$HOME/.cache/git-fork-recon/repos"
report = "$HOME/.cache/git-fork-recon/reports"

[server]
# Server configuration options (commented out by default)

Environment Variable References

You can use environment variable references in the config file by prefixing with $:

[endpoint]
base_url = "$OPENAI_BASE_URL"  # Reads from OPENAI_BASE_URL env var
api_key = "$OPENAI_API_KEY"    # Reads from OPENAI_API_KEY env var

[github]
token = "$GITHUB_TOKEN"        # Reads from GITHUB_TOKEN env var

Custom Config File

You can specify a custom config file location using the --config option:

git-fork-recon --config /path/to/config.toml https://github.com/user/repo

Commandline options

$ git-fork-recon --help

 Usage: git-fork-recon [OPTIONS] [REPO_URL]

 Analyze a GitHub repository's fork network and generate a summary report.


╭─ Arguments ──────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│   repo_url      [REPO_URL]  URL of the GitHub repository to analyze          │
│                             [default: None]                                  │
╰──────────────────────────────────────────────────────────────────────────────╯
╭─ Options ────────────────────────────────────────────────────────────────────╮
│ --output              -o      PATH     Output file path (defaults to         │
│                                        {repo_name}-forks.md)                 │
│                                        [default: None]                       │
│ --active-within               TEXT     Only consider forks with activity     │
│                                        within this time period (e.g. '1      │
│                                        hour', '2 days', '6 months', '1       │
│                                        year')                                │
│                                        [default: None]                       │
│ --config                      PATH     Path to config.toml file [default:     │
│                                        None]                                  │
│ --model                       TEXT     OpenRouter model to use (overrides    │
│                                        MODEL env var)                        │
│                                        [default: None]                       │
│ --context-length              INTEGER  Override model context length         │
│                                        (overrides CONTEXT_LENGTH env var)    │
│                                        [default: None]                       │
│ --api-base-url                TEXT     OpenAI-compatible API base URL        │
│                                        [default: None]                       │
│ --api-key-env-var             TEXT     Environment variable containing the   │
│                                        API key                               │
│                                        [default: None]                       │
│ --parallel            -p      INTEGER  Number of parallel requests           │
│                                        [default: 5]                          │
│ --verbose             -v               Enable verbose logging                │
│ --clear-cache                          Clear cached repository data before   │
│                                        analysis                              │
│ --force-fetch                           Force fetch updates from cached     │
│                                        repositories and remotes              │
│ --force                                Force overwrite existing output file  │
│ --max-forks                   INTEGER  Maximum number of forks to analyze    │
│                                        (default: no limit)                   │
│                                        [default: None]                       │
│ --output-formats              TEXT     Comma-separated list of additional    │
│                                        formats to generate (html,pdf)        │
│                                        [default: None]                       │
│ --install-completion                   Install completion for the current    │
│                                        shell.                                │
│ --show-completion                      Show completion for the current       │
│                                        shell, to copy it or customize the    │
│                                        installation.                         │
│ --help                                 Show this message and exit.           │
╰──────────────────────────────────────────────────────────────────────────────╯

Server-mode

Server configuration

For the REST API server, in addition config.toml, some configuration can be overriden using environment variabless:

• ALLOWED_MODELS: Comma-separated list of allowed LLM models (default: unrestricted)
• SERVER_HOST: Host to bind the server to (default: 127.0.0.1)
• SERVER_PORT: Port to bind the server to (default: 8000)
• REPORT_CACHE_DIR: Directory for server report cache (defaults to ~/.cache/git-fork-recon/reports using platformdirs)
• DISABLE_AUTH: Set to 1 to disable authentication (default: enabled)
• AUTH_BEARER_TOKEN: Bearer token for API authentication
• PARALLEL_TASKS: Maximum concurrent analysis tasks (default: 2)
• DISABLE_UI: Set to 1 to disable the web UI at /ui endpoint (default: enabled)

Run the server

Start the server:

# Using installed package
git-fork-recon-server --host 127.0.0.1 --port 8000

# Using uvx
uvx --from 'git-fork-recon[server]' git-fork-recon-server --host 127.0.0.1 --port 8000

Go to http://localhost:8000/ui to see the web UI.

REST API Endpoints

• POST /analyze - Start repository analysis
• GET /report/{owner}/{repo}/{timestamp}/report.{format} - Get cached report
• GET /report/{owner}/{repo}/latest/report.{format} - Get latest cached report
• GET /report/{owner}/{repo}/{timestamp}/status - Get status for specific report version
• GET /report/{owner}/{repo}/latest/status - Get status for latest report
• GET /metadata/{owner}/{repo}/{timestamp} - Get metadata for specific report version
• GET /metadata/{owner}/{repo}/latest - Get metadata for latest report
• GET /health - Health check endpoint
• GET /health/ready - Readiness check endpoint
• GET /ui - Web UI for repository analysis (unless disabled with DISABLE_UI=1)

Example Request

curl -X POST "http://localhost:8000/analyze" \
  -H "Authorization: Bearer your-token" \
  -H "Content-Type: application/json" \
  -d '{
    "repo_url": "https://github.com/martinpacesa/BindCraft",
    "model": "deepseek/deepseek-chat-v3-0324:free",
    "format": "markdown"
  }'

Example Response

{
  "status": "generating",
  "retry-after": "2025-10-04T12:35:00Z"
}

When analysis is complete:

{
  "status": "available",
  "link": "/report/martinpacesa/BindCraft/latest/report.md",
  "last-updated": "2025-10-04T12:34:56Z"
}

Retrieving the Generated Report

Once the analysis is complete, you can retrieve the report using the provided link:

# Get the latest report
curl -X GET "http://localhost:8000/report/martinpacesa/BindCraft/latest/report.md" \
  -H "Authorization: Bearer your-token" \
  -o martinpacesa-BindCraft-forks.md

# Or get a specific version by timestamp
curl -X GET "http://localhost:8000/report/martinpacesa/BindCraft/2025-10-04T12-34-56Z/report.md" \
  -H "Authorization: Bearer your-token" \
  -o martinpacesa-BindCraft-forks-v2025-10-04.md

# Get report in different formats (markdown, json, html, pdf)
curl -X GET "http://localhost:8000/report/martinpacesa/BindCraft/latest/report.json" \
  -H "Authorization: Bearer your-token" \
  -o martinpacesa-BindCraft-forks.json

Checking Analysis Status

If you request a report while it's still being generated, you'll receive a 202 Accepted response with a Retry-After header:

curl -X GET "http://localhost:8000/report/martinpacesa/BindCraft/latest/report.md" \
  -H "Authorization: Bearer your-token"

Response (while generating):

{
  "status": "generating",
  "retry-after": "Wed, 05 Oct 2025 12:35:00 GMT"
}

Output is generated as {username}-{repo}-forks.md by default (use -o to specify a different file name, -o - to print to stdout).

See also

• Useful forks
• frogmouth - a quick viewer for the generated Markdown