Ava WhatsApp Agent Extended

Honorable Mention: This project is based on Ava WhatsApp Agent Course by Neural Maze. We've extended it with additional features like multi-provider support for image generation and analysis.

Ava is a versatile AI companion agent that can interact through chat, generate images, and convert speech to text and back. This project is built on a flexible architecture that supports multiple AI providers including OpenAI, Groq, Ollama, Together AI, and ElevenLabs.

Features

• 💬 Natural Conversation: Have natural, context-aware conversations with your AI companion
• 🖼️ Image Generation: Generate images based on your descriptions, with support for Together AI or OpenAI DALL-E
• 🔍 Image Analysis: Analyze images with detailed descriptions using Groq or OpenAI
• 🗣️ Voice Interaction: Talk to your AI with text-to-speech (TTS) and speech-to-text (STT)
• 🧠 Memory System: Both short-term and long-term memory with vector database support
• 🔄 Multi-Provider Support: Switch between different LLM providers (OpenAI, Groq, Ollama)
• 📱 WhatsApp Integration: Chat with your AI companion directly through WhatsApp
• 💻 Web Interface: Interact through a browser-based Chainlit interface

Architecture Overview

Ava is designed with a modular architecture that allows for flexible configuration and deployment:

System Components:

1. User Interfaces:

   • WhatsApp for mobile messaging
   • Chainlit for web-based interaction

2. Core Processing:

   • LangGraph-powered workflow engine
   • Context injection for personalized responses
   • Memory management system

3. AI Services:

   • Text Generation: OpenAI, Groq, or Ollama
   • Image Generation: Together AI or OpenAI DALL-E
   • Image Analysis: Groq Llama or OpenAI GPT-4 Vision
   • Speech Processing: OpenAI or Groq for STT, OpenAI or ElevenLabs for TTS

4. Storage:

   • Vector Database (Weaviate or Qdrant) for long-term memory
   • Session-based short-term memory

Supported Providers

• LLM: OpenAI, Groq, Ollama
• Speech-to-Text: OpenAI, Groq
• Text-to-Speech: OpenAI, ElevenLabs
• Image-to-Text: OpenAI, Groq
• Text-to-Image: OpenAI (DALL-E), Together AI

Getting Started

See GETTING_STARTED.md for detailed installation and configuration instructions.

Quick Start

1. Clone this repository
2. Copy .env.example to .env and add your API keys
3. Choose your vector database (Qdrant or Weaviate)
4. Run the application with Docker Compose:

# For Qdrant
make ava-run-qdrant

# For Weaviate
make ava-run-weaviate

Deployment Options

• Local: Run on your machine using Docker
• Cloud: Deploy to Google Cloud Run
• Self-hosted Ollama: Deploy Ollama models on vast.ai

License

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

Acknowledgements

This project is an extension of the Ava WhatsApp Agent Course by Neural Maze, a comprehensive course that teaches how to build a WhatsApp agent from scratch. We've built upon their excellent foundation to add multi-provider support and extended capabilities.