CommitIt

An AI-powered codebase understanding platform that analyzes GitHub repositories and helps developers quickly understand project architecture, dependencies, and implementation details through intelligent search and natural language explanations.

---

Features

• Clone and analyze any public GitHub repository
• Build a structured knowledge model of the codebase
• Parse source files to extract classes, functions, imports, and modules
• Generate dependency graphs
• Search code symbols and files
• Semantic repository queries
• Context generation for developer questions
• Deterministic explanation engine
• Optional Gemini AI integration with automatic fallback
• REST API built with FastAPI

---

Tech Stack

Backend

• Python
• FastAPI
• Pydantic
• GitPython

AI

• Google Gemini API (optional)
• Deterministic explanation engine (offline fallback)

Testing

• Pytest

---

Project Structure

backend/
│
├── app/
│   ├── api/
│   ├── core/
│   ├── models/
│   ├── services/
│   │   └── llm/
│   └── utils/
│
├── tests/
├── requirements.txt
├── run.py
└── README.md

---

API Overview

Repository

• Clone Repository
• Repository Health

Analysis

• Scan Repository
• Parse Repository
• Dependency Graph
• Knowledge Model

Query Engine

• List Classes
• List Functions
• List Imports
• List Files
• Symbol Search
• Relationship Search

AI Features

• Context Builder
• Explanation Engine
• AI Explain Endpoint

---

Installation

Clone the repository

git clone https://github.com/pragya-shree/commitit.git
cd commitit/backend

Create a virtual environment

python -m venv .venv

Activate it

Windows

.venv\Scripts\activate

Install dependencies

pip install -r requirements.txt

Create a .env file

GEMINI_API_KEY=
GEMINI_MODEL=gemini-2.5-flash

The Gemini API key is optional. If it is not provided, CommitIt automatically falls back to its deterministic explanation engine.

---

Run

python run.py

The API will be available at

http://localhost:8000

Interactive documentation

http://localhost:8000/docs

---

Architecture

GitHub Repository
        │
        ▼
 Repository Clone
        │
        ▼
 Repository Scanner
        │
        ▼
 Source Parser
        │
        ▼
 Dependency Graph
        │
        ▼
 Knowledge Model
        │
        ▼
 Query Engine
        │
        ▼
 Context Builder
        │
        ▼
 Explanation Engine
        │
        ▼
 Optional Gemini AI

---

Design Highlights

• Single-build repository analysis
• Cached in-memory knowledge model
• Read-only query layer
• Deterministic explanation pipeline
• Optional AI enhancement
• Automatic fallback when AI is unavailable
• No databases
• No Redis
• No Docker required

---

Testing

Run all tests

pytest

---

Future Work

• Support additional LLM providers
• Repository visualization
• Interactive frontend
• Multi-repository analysis
• Conversation memory
• Code editing assistance

---

License

This project is licensed under the MIT License.