Java Code Analyzer with RAG and Ollama

This project implements a Java code analyzer that uses Retrieval-Augmented Generation (RAG) techniques combined with the Ollama language model to provide intelligent responses to queries about Java codebases.

Features

• Parses and indexes Java source code from a specified repository
• Uses Tree-sitter for efficient code parsing
• Implements a RAG (Retrieval-Augmented Generation) system for context-aware code understanding
• Integrates with Ollama for generating human-like responses to code-related queries
• Includes caching mechanisms for improved performance on repeated analyses
• Optimized for handling large codebases with multi-processing capabilities

Prerequisites

Before you begin, ensure you have met the following requirements:

• Python 3.7+
• Ollama installed and running locally with the llama3.1 model
• Git (for cloning the repository)

Installation

1. Clone the repository:

   git clone https://github.com/sydowma/ragCodeKnowledge.git
   cd ragCodeKnowledge
   

2. Install the required Python packages:

   pip install -r requirements.txt
   

3. Ensure you have the Tree-sitter Java grammar:

   git clone https://github.com/tree-sitter/tree-sitter-java.git
   

Usage

1. Update the repo_path in the script to point to your Java repository:

   repo_path = "/path/to/your/java/repository"

2. Run the script:

   python optimized_rag_java_analyzer.py
   

3. When prompted, enter your query about the Java codebase. For example:

   Enter your query (or 'quit' to exit): what is xxxx?
   

4. The system will retrieve relevant code snippets and generate a response using Ollama.

5. To exit, type 'quit' when prompted for a query.

How It Works

1. Code Indexing: The system parses Java files in the specified repository using Tree-sitter and creates an index of code snippets.

2. Caching: Indexed data is cached for faster subsequent runs. The cache is invalidated if the repository content changes.

3. Query Processing: When a query is received, the system retrieves relevant code snippets using semantic similarity search.

4. Response Generation: Relevant snippets are sent to Ollama along with the query to generate a contextualized response.

Configuration

You can modify the following parameters in the script:

• JAVA_LANGUAGE_PATH: Path to the Tree-sitter Java language file
• k in query_code function: Number of relevant snippets to retrieve (default is 5)
• Ollama model in query_ollama function (default is "deepseek-coder-v2")

Future Improvements

• Implement incremental updates for the code index
• Add support for other programming languages
• Improve query understanding with more advanced NLP techniques
• Integrate with IDEs or code editors for seamless usage

Contributing

Contributions to improve the Java Code Analyzer are welcome. Please feel free to submit a Pull Request.

Contact

If you have any questions or feedback, please open an issue in the GitHub repository.