Documentation Generator with LangChain and Google Generative AI

This project is a Documentation Generator for local repositories that leverages LangChain, Google Generative AI (Gemini-1.5-flash), and LangGraph to analyze code and generate detailed, high-quality API documentation. The tool processes JavaScript and TypeScript files, splits them into manageable chunks, and generates documentation with suggestions for improvement.

---

Table of Contents

1. Features
2. Installation
3. Usage
4. Configuration
5. Technologies Used
6. File Structure
7. Future Improvements
8. License: MIT

---

Features

• Automatically loads and analyzes files from a specified directory.
• Uses Google Generative AI to generate detailed documentation for:
  • API Endpoints
  • Functions and Methods
  • Data Models
  • Design Patterns
  • Security Considerations
• Includes a reflection step for documentation critique and improvement.
• Implements LangGraph for iterative workflows.
• Provides customizable chunk size and overlap for text splitting.
• Streams results to handle large outputs efficiently.

---

Usage

Installation

Prerequisites

Ensure the following are installed on your machine:

• Node.js (v18+ recommended)
• npm or yarn
• Google API Key for Generative AI
• Unstructured API Key (optional, for advanced text loading)

Steps

1. Install the package

   npm install docs-gen-js-ts

2. RUn the script

   docs-gen

   • The project will scan the specified directory (process.cwd() by default), split the documents, and generate API documentation based on the loaded code files.

Technologies Used

• LangChain: Framework for building AI-driven workflows.
• Google Generative AI (Gemini): Generates detailed and insightful documentation.
• LangGraph: Manages iterative workflows with state graphs.
• TypeScript: Ensures type safety and maintainability.
• Node.js: Executes the backend logic.

Future Improvements

• Add support for more programming languages.
• Integrate with cloud storage for generated documentation.
• Improve error handling for edge cases (e.g., large repositories).
• Enhance the reflection process with additional critique points.

Developed by Shalwin. If you have questions or feedback, feel free to reach out!

• Email : [shalwinsanju.25cs@licet.ac.in]
• Github : shalwin04