ChatTensorFlow

ChatTensorFlow is an intelligent assistant that helps users with TensorFlow related queries by utilizing the power of LangGraph based Agents, vector embeddings, and a conversational interface powered by various LLM providers.

Features

• Multi-provider support: Works with multiple LLM providers (Google Gemini, OpenAI, Anthropic, Cohere)
• Content processing: Splits and embeds TensorFlow documentation for semantic search
• Query routing: Intelligently classifies user questions and determines next steps
• Assisted research: Follows a research plan to answer complex TensorFlow queries
• Context-aware responses: Provides answers with citations from official TensorFlow documentation

Architecture

ChatTensorFlow is built with a modular architecture:

1. Document Crawler: Efficiently crawls TensorFlow documentation using sitemap.xml
2. Data Ingestion: Processes web content into chunks for embedding
3. Vector Store: Stores embeddings for semantic search using Chroma
4. Retriever: Retrieves relevant content when answering questions
5. Researcher Graph: Follows a structured plan to research answers
6. Router Graph: Determines how to handle each user query
7. Response Generator: Creates coherent, accurate responses based on retrieved information

Graphs

TF Assistant main Graph

Researcher sub Graph

Getting Started

Prerequisites

• Python 3.9+
• API keys for at least one supported LLM provider

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/chatTensorFlow.git
   cd chatTensorFlow

2. Create a virtual environment:

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

3. Install dependencies:

   pip install -r requirements.txt

4. Set up your environment variables:

   cp .env.example .env
   # Edit .env with your API keys

Data Collection

To crawl and process the Scikit-learn documentation:

1. Run the web crawler:

   python -m processor.crawler

2. Process and ingest the data:

   python -m processor.data_ingestion

This will create a Chroma vector database with the embedded documentation chunks.

Usage

Run the application:

python app/main.py

The assistant can answer questions about:

• TensorFlow API usage
• Deep learning concepts
• Model building and training
• TensorFlow data pipelines
• Common errors and troubleshooting

Example queries:

• "How do I compile and train a Keras model?"
• "How do I use tf.data.Dataset to load and preprocess large datasets?"
• "What are the parameters for a Conv2D layer?"
• "How do I create and use a custom loss function?"

Configuration

The application can be configured by setting the following environment variables:

# LLM Provider (gemini, openai, anthropic, cohere)
LLM_PROVIDER=gemini

# Embedding Provider (gemini, openai, cohere)
EMBEDDING_PROVIDER=gemini

# API Keys
GOOGLE_API_KEY=your_gemini_api_key
OPENAI_API_KEY=your_openai_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key
COHERE_API_KEY=your_cohere_api_key

# Model selections (optional, defaults provided)
GEMINI_LLM_MODEL=gemini-2.0-flash
OPENAI_LLM_MODEL=gpt-4o
ANTHROPIC_LLM_MODEL=claude-3-sonnet-20240229
COHERE_LLM_MODEL=command

Project Structure

chatTensorFlow/
├── app/
│   ├── core/         # Core utilities and model interfaces
│   ├── graphs/       # LangGraph state management and flow
│   ├── retriever/    # Document retrieval components
│   ├── static/       # Static files (CSS, JS, images)
│   ├── templates/    # HTML templates
│   └── main.py       # Application entry point
├── config/           # Application configuration
├── DATA/             # Vector store data
├── logs/             # Application logs
├── processor/        # Web crawling and data ingestion
├── tensorflow_docs/  # Crawled TensorFlow documentation
└── tests/            # Unit and integration tests

How It Works

1. User Query Analysis: The system classifies the query as TensorFlow related, general, or requiring more information
2. Research Planning: For TensorFlow queries, a research plan is created
3. Document Retrieval: Relevant documentation is retrieved using semantic search
4. Response Generation: A detailed, accurate response is generated citing relevant documentation

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built with LangChain and LangGraph
• Uses crawl4ai for crawling
• Uses ChromaDB for vector storage
• Powered by various LLM providers (Gemini, OpenAI, Anthropic, Cohere)
• Documentation content from TensorFlow
• Developed by Lalan Kumar