Code Navigator CLI

Code Navigator is a command-line tool designed to help developers understand complex codebases. It works by indexing a Git repository into a Neo4j graph database and allowing you to ask natural language questions about the code. The tool uses Large Language Models (LLMs) to parse your questions, generate database queries, and provide answers about the codebase's structure, including files, classes, functions, and their relationships.

Features

• Codebase Indexing: Clones any public Git repository and maps its structure into a graph database.
• AST-based Parsing: Analyzes Python source files to identify key entities like classes, functions, imports, and calls.
• Natural Language Queries: Ask questions like "What does the run_query_agent function do?" or "Which classes inherit from BaseMessage?".
• LLM-Powered Query Generation: Automatically converts your questions into Cypher queries to be executed against the Neo4j database.
• Interactive CLI: A simple command-line interface for indexing repositories and asking questions.

Prerequisites

Before you begin, ensure you have the following installed:

• Python 3.11 or higher
• Docker and Docker Compose

Installation and Setup

Follow these steps to get the Code Navigator CLI running on your local machine.

1. Clone the Repository First, clone this repository to your local machine:

git clone https://github.com/Player256/code-rag-graph
cd code-rag-graph

2. Set Up a Python Environment It is recommended to use a virtual environment to manage dependencies.

python -m venv venv
source venv/bin/activate
# On Windows, use: venv\Scripts\activate

3. Install Dependencies(will be updated later) Install the required Python packages. You will need to create a requirements.txt file with the necessary libraries (e.g., langchain, langchain-aws, neo4j, prompt-toolkit, python-dotenv, gitpython).

pip install -r requirements.txt

4. Start the Neo4j Database The application uses a Neo4j database running in a Docker container. Start the service using Docker Compose:

docker-compose up -d

This command starts the Neo4j container in the background. You can access the Neo4j browser interface at http://localhost:7474.

5. Configure Environment Variables The application requires credentials to connect to the Neo4j database. Create a file named .env in the root of the project directory.

touch .env

Add the following lines to the .env file. These credentials must match the ones specified in the docker-compose.yml file.

NEO4J_URI=neo4j://localhost:7687
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=<password>

How to Use

Once the setup is complete, you can run the application from the root directory of the project.

1. Start the CLI

python main.py

You will be greeted with a welcome message and a prompt.

2. View Available Commands Type help to see a list of all available commands.

> help

3. Index a Repository Before you can ask questions, you must index a codebase. Use the index command followed by the URL of a public Git repository.

> index https://github.com/vllm-project/vllm

The tool will clone the repository, parse its Python files, and load the structure into the database.

4. Ask a Question Once a repository is indexed, you can ask questions about it using the ask command.

> ask How is the model class designed in the vllm repository?

The tool will use an LLM to generate and execute a Cypher query to find the answer and return a human-readable response.

5. Exit the Application To close the CLI, type exit.

> exit

Project Structure

The project is organized into several directories to separate concerns:

└── src/
    ├── agents/        # Contains the specialist agents (ingestion, query) and the orchestrator
    ├── cli/           # Handles the command-line interface logic
    ├── graph_db/      # Manages the Neo4j connection, data ingestion, and Cypher queries
    ├── model.py       # Configures and provides the Bedrock LLM
    └── utils/         # Contains utility functions and tool definitions for the agents