metadata

title: chat-with-avd-doc
app_file: app.py
sdk: gradio
sdk_version: 5.30.0

Network Fabric Documentation Chat Assistant

A powerful AI-assisted tool for querying network fabric documentation using semantic search and large language models. This application allows network engineers and administrators to easily access and retrieve information about their network fabric through natural language queries.

🚀 Key Features

Natural Language Queries: Ask questions about your network in plain English
Semantic Search: Uses embeddings to find relevant information beyond simple keyword matching
Device-Aware Context: Distinguishes between global and device-specific information
Interactive UI: User-friendly Gradio web interface
Vector Database: Persistent storage of document embeddings using ChromaDB
LLM Integration: Powered by Google's Gemini Flash model
Customizable Prompts: Configure system and user prompts through YAML

🛠️ Technical Stack

Vector Search: ChromaDB for efficient semantic search
Embeddings: Nomic AI's text embeddings (nomic-embed-text-v1)
LLM Orchestration: SmolagentsCodeAgent for tool use and reasoning
UI Framework: Gradio for the web interface
Model: Google's Gemini 2.0 Flash via LiteLLM

📋 Requirements

Python 3.8+
Dependencies listed in requirements.txt

🔧 Installation

Clone the repository:

git clone https://github.com/yourusername/chat-with-avd-doc.git
cd chat-with-avd-doc

Create a virtual environment (recommended):

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

Install dependencies:
```
pip install -r requirements.txt
```
Configure environment variables (if needed):
- For LiteLLM to access Gemini, you may need to set API keys
- Create a .env file or set them directly in your environment

🔍 Working with the Vector Database

The application uses ChromaDB to store document embeddings. The database is created in the vector_db directory.

Documents are organized by "fabric" collection
Metadata includes source information that can identify device-specific documentation
The system distinguishes between global network information and device-specific details

📝 Configuration

Prompt Templates

Edit prompts.yaml to customize:

System instructions for the AI agent
Formatting of responses
Behavior and capabilities

Model Configuration

The application uses Gemini Flash by default. To modify:

# In app.py
model = LiteLLMModel("gemini/gemini-2.0-flash")  # Change model here

🚀 Usage

Start the application:
```
python app.py
```
Access the UI: Open your browser to http://127.0.0.1:7860
Query your network documentation:
- Ask natural language questions about your network
- Example: "What is the loopback Pool address used by the fabric?"
- Example: "How many IP addresses are in use in the management network?"

📂 Project Structure

chat-with-avd-doc/
├── app.py              # Main application with Retriever Tool and Gradio UI
├── prompts.yaml        # Configuration for AI prompts and behavior
├── requirements.txt    # Python dependencies
└── vector_db/          # ChromaDB database files
    ├── chroma.sqlite3  # SQLite database for ChromaDB
    └── [UUID folders]  # Vector data storage

🧪 How It Works

User Query: User enters a natural language question through the Gradio UI
Semantic Encoding: The query is converted to a vector embedding
Vector Search: ChromaDB searches for similar document vectors
Context Assembly: Top matches are assembled with source metadata
LLM Processing: The Gemini model processes the query with retrieved context
Response Generation: The system returns a natural language answer

📄 License

MIT License

🤝 Contributing

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

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