GitHub Agentic Search

A CLI tool that answers questions about a GitHub repository on demand, without pre-ingestion. It fetches the repo tree, README, searches code, and inspects files, PRs, and issues via the GitHub API.

Prerequisites

• Python 3.10+ installed and available as python3
• A GitHub Personal Access Token (classic or fine-grained) with repo read permissions
• An OpenAI API key to run the agent model
• macOS/Linux shell (examples use zsh/bash)

Quick Start

1) Clone the repository

git clone https://github.com/<your-username>/github-agentic-search.git
cd github-agentic-search

2) Create and activate a virtual environment (Python 3)

python3 -m venv .venv
source .venv/bin/activate

3) Install dependencies

pip install --upgrade pip
pip install -r requirements.txt

4) Configure environment variables

Create a .env file in the project root with the following variables:

GITHUB_TOKEN=ghp_your_personal_access_token
OPENAI_API_KEY=sk-your_openai_api_key

Notes:

• Check how to get a GitHub Acess Token Here
• OPENAI_MODEL defaults to gpt-5-mini if omitted.

5) Run the CLI

The CLI entrypoint is index.py. Use:

• --repo as owner/name
• --question as your query
• --ref for a branch or commit SHA (Optional).

python index.py --repo owner/name --question "What is the deployment process?" --ref main

Examples:

# Ask about project structure
python index.py --repo pytorch/pytorch --question "Where is the build configuration defined?"

# Ask about recent PRs/issues (agent may check PRs/issues if relevant)
python index.py --repo owner/private-repo --question "What PR added feature X?" --ref 123abc4

Output

• The tool prints a concise, sourced answer to stdout.
• Responses may cite paths like path/to/file.py and include short snippets from fetched files.
• If information is uncertain, it may state additional places it would check.