Ra.One - Human-like AI WhatsApp Assistant Development by Keshav GargRa.One - Human-like AI WhatsApp Assistant Development by Keshav Garg

Ra.One - Human-like AI WhatsApp Assistant Development

Keshav Garg

Completed work

AI Engineer

ElevenLabs

PostgreSQL

Artificial Intelligence

🤖 Ra.One - Human-like AI WhatsApp Assistant

Ra.One is a sophisticated AI-powered WhatsApp assistant that delivers remarkably human-like conversations. Built on the Langgraph framework, Ra.One goes beyond basic chatbot functionality to provide natural, contextual, and emotionally intelligent interactions right in your WhatsApp conversations.

✨ Human-like Capabilities

🧠 Contextual Understanding: Maintains conversations across multiple messages

🗣️ Natural Language: Communicates with nuanced, human-like responses

🎭 Emotional Intelligence: Recognizes and responds appropriately to user emotions

🧩 Personality: Maintains a consistent, engaging personality throughout interactions

🤔 Memory: Remembers past conversations to provide personalized experiences

🚀 Key Features

✅ WhatsApp Integration: Seamless connection via WhatsApp Cloud API

💬 Real-time AI Conversations: Instant intelligent responses

🔊 Voice Capabilities: Converts voice messages to text and responds with natural voice

🔗 Multi-turn Dialogues: Maintains conversation context over time

📊 Dual Memory System: Postgres for short-term and Pinecone for long-term memory

🧰 Extensible Architecture: Easily add custom tools and capabilities

🔐 Privacy-Focused: Secure design with local-first approach

📱 Experience Ra.One

Chat with Ra.One like you would with a friend! It understands context, responds naturally, and gets smarter with every interaction.

🛠️ Setup Instructions

Prerequisites

Python 3.8+

PostgreSQL database

WhatsApp Business Account

API keys (Groq, ElevenLabs, Pinecone, etc.)

1. Clone the Repository

git clone https://github.com/KeshavG69/RaOne.git
cd RaOne

2. Install Dependencies

python3 -m venv venv
source venv/bin/activate  # On Windows use `venv\Scripts\activate`
pip install -r requirements.txt

3. Configure Environment Variables

cp .env.example .env

Edit the .env file with your API keys:

GROQ_API_KEY='your_groq_api_key'
ELEVENLABS_API_KEY='your_elevenlabs_key'
ELEVENLABS_VOICE_ID='your_voice_id'
PINECONE_API_KEY='your_pinecone_key'
PINECONE_ENVIRONMENT='your_pinecone_environment'
PINECONE_INDEX='your_pinecone_index'
TOGETHER_504='your_together_key'
WHATSAPP_TOKEN='your_whatsapp_token'
WHATSAPP_PHONE_NUMBER_ID='your_phone_number_id'
WHATSAPP_VERIFY_TOKEN='your_verify_token'
COHERE_API_KEY='your_cohere_key'
SHORT_TERM_MEMORY_DB_PATH='postgres-path'

4. Database Setup

Ra.One requires PostgreSQL for short-term memory storage. Ensure you have PostgreSQL installed and create a database for Ra.One to use. Configure the connection details in your .env file as shown above.

5. Launch the Webhook Server

uvicorn webhook_endpoint:app --host 0.0.0.0 --port 5000

6. Expose Your Webhook

Using Ngrok:

ngrok http 5000

7. Configure Meta Developer Portal

Go to Meta for Developers

Navigate to your app > WhatsApp > Configuration

Set your callback URL: https://your-ngrok-url.ngrok.io/webhook

Enter your verify token (same as WHATSAPP_VERIFY_TOKEN)

8. Start Chatting!

Your Ra.One assistant is now online and ready to engage in human-like conversations on WhatsApp.

🧠 Memory Architecture

Ra.One uses a sophisticated dual-memory system:

Short-term Memory (PostgreSQL): Ra.One utilizes PostgreSQL as its short-term memory database to store recent conversations, user preferences, and contextual information. This relational database provides quick access to recent interactions, allowing Ra.One to maintain context during active conversations. The structured nature of PostgreSQL makes it perfect for storing conversation threads, user details, and immediate contextual data that Ra.One needs for responsive, coherent exchanges.

Long-term Memory (Pinecone): For deeper semantic understanding and long-term knowledge retention, Ra.One leverages Pinecone's vector database capabilities. Conversation histories are embedded into semantic vectors and stored in Pinecone, enabling Ra.One to retrieve relevant past interactions based on meaning rather than just keywords. This vector-based approach allows the assistant to maintain a more human-like memory of past conversations, recognizing patterns and recalling relevant information even months later.

Memory Integration: The system seamlessly transitions information between short and long-term memory, with important conversational details gradually moving from PostgreSQL to Pinecone as they age. This dual-layer approach ensures Ra.One maintains both immediate responsiveness and long-term personalization, creating a more natural conversational experience.

🔧 Customization

Ra.One is designed to be highly customizable:

Personality Tuning: Adjust conversation style in prompt.py

Memory Settings: Configure memory retention periods and importance thresholds

Voice Characteristics: Customize speech patterns via ElevenLabs

Custom Skills: Add specialized capabilities through the modular architecture

Conversation Flow: Design custom interaction patterns in graph.py

📁 Project Files

keshavg69-raone/
├── README.md               # Project documentation
├── app.py                  # Main application entry point
├── edges.py                # Connection definitions for graph
├── graph.py                # Conversation flow architecture
├── nodes.py                # Processing nodes for the graph
├── prompt.py               # Personality and instruction templates
├── requirements.txt        # Project dependencies
├── schedule.py             # Scheduled message functionality
├── schedule_manager.py     # Management of timed interactions
├── settings.py             # Configuration settings
├── speech_to_text.py       # Voice message processing
├── state.py                # Conversation state management
├── text_to_speech.py       # Voice response generation
├── utils.py                # Utility functions
├── webhook_endpoint.py     # WhatsApp webhook handler
├── whatsapp_response.py    # Response formatting for WhatsApp

└── .env.example            # Environment variable template

👨‍💻 Contributing

Contributions are welcome to make Ra.One even more human-like and capable! Please feel free to:

Submit pull requests with enhancements

Report bugs or suggest features

Share your custom personality configurations

Add new integration modules

🚀 License

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

Created with ❤️ by [Keshav Garg]gargkeshav504@gmail.com)

"Ra.One: Your AI companion that feels like chatting with a friend"

Like this project