AI-Powered CLI for Exploring Git Repositories

GitLLM Explorer is an AI-powered command-line tool that leverages LLM-based agents to help developers and teams explore and understand Git repositories. By asking natural language questions, users can:

• Summarize a repository's purpose and structure.
• Analyze branch activity and workflows.
• Generate detailed repository reports.
• Search for GitHub repositories based on topics.

• Natural Language Interface
  • Ask questions in plain English; the AI translates them into Git commands or queries.
  • Maintains conversational context for follow-up queries.
• Modular LLM Agents
  • Setup Agent: Clone and initialize repositories.
  • Commands Agent: Execute safe Git commands and fetch data.
  • Report Agent: Generate human-readable reports on repository history and structure.
  • Assistant Agent: Orchestrates agent interactions for smooth workflows.
• Repository Insights and Analytics
  • Summaries based on README and code.
  • Branch activity and merge history.
  • Workflow strategy detection (e.g., Git Flow, GitHub Flow).
• Repository Search
  • Discover GitHub repositories by topic, stars, or activity.
• Asynchronous and Extensible
  • Built with asyncio for responsive CLI interactions.
  • Easily extendable with new agents and tools.
• Configurable LLM Models
  • Customize each agent's model via environment variables.

• Python 3.8 or higher
• uv

1. Install uv:

   curl -LsSf https://astral.sh/uv/install.sh | sh
   # On macOS, follow instructions at https://docs.astral.sh/uv/getting-started/installation/

2. Clone the repository:

   git clone https://github.com/DSmmartin/msft-hackaton-gitllmexplorer.git
   cd msft-hackaton-gitllmexplorer

3. Create and activate the virtual environment:

   uv sync && uv venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate

Set the following environment variables:

• GITHUB_TOKEN: Your GitHub token (optional; for the search functionallity).
• Agent-specific model variables (optional; defaults to gpt-4o-mini):
  • GIT_ASSISTANT_MODEL
  • GIT_COMMANDS_MODEL
  • GIT_REPORT_MODEL
  • GIT_SETUP_MODEL
  • OPENAI_API_KEY: Your OpenAI API key, in case you want to use the OpenAI provider.
  • AZURE_OPENAI_API_KEY, AZURE_OPENAI_API_VERSION, AZURE_OPENAI_ENDPOINT: The credentials for the Azure OpenAI provider.

Run the CLI:

uv run ./src/main.py

Once launched, type your questions about Git repositories. Type exit to quit.

• Summarize a repository:

  💬 Could you provide a summary about this repo based on the README? The URL is https://github.com/microsoft/promptflow.git

• List active branches in a local repo:

  💬 What are the current active branches in /path/to/repo?

• Generate a repository report:

  💬 Generate a report of the repository: https://github.com/microsoft/promptflow.git

• Search for Python repos:

  💬 Find me the best repositories about transformers

• Developer Onboarding: Quickly understand a new codebase.
• Branch Strategy Audits: Analyze branch workflows (Git Flow, GitHub Flow, etc.).
• Project Audits: Generate automated reports for stakeholders.
• Repository Discovery: Find relevant open-source projects.

1. Agent Orchestration: Uses a Runner to sequentially call LLM agents based on responses, this is the Git Assistant and the Git Reporter when the handoff operations is executed.
2. Tools Integration: There are other Agents used as tools that group high level tasks, these agents facilitate the execution of complex workflows.
3. Context Management: Pydantic models (GitRepository, ListOfRepositories) store agent context.
4. Asynchronous CLI: Built with asyncio and Rich for interactive consoles.

• Improving the Git Exploratory analisis tool which more detailed information such as:
  1. The tags
  2. Core components/folders of the repository
  3. The main issues and pull requests
• Packaging the tool in a CLI library
• Building a TUI for the tool (the Textual library)
• Integrating some Git commands with the Git MCP
• Adding unit tests for core functionalities
• Building a RAG with the book "PRO GIT" by Scott Chacon and Ben Straub, in order to extend the capabilities of this tool to tutor/learning.
• Add traceability to the tool (promptflow), in order to track the redirections, the use of the tools and the agents, and the use of the LLMs.
• Build a Evaluation system to evaluate the performance of the tool/agents, and the LLMs used in the tool.