TittoosAI

Tittoo AI Agent

Tittoo is a powerful AI agent that can understand voice and text input in multiple Indian languages, execute code, and respond with voice or text. It uses local models by default with automatic fallback to cloud services when needed.

Features

• Multilingual Support: Understands and processes Indian languages (Hindi, Tamil, Marathi, Bengali, etc.)
• Voice & Text Input: Interact using voice or text commands
• Local-First Processing: Uses Whisper.cpp for STT and Ollama for LLM inference
• Automatic Cloud Fallback: Falls back to OpenAI services when local processing fails
• Code Execution: Safely runs Python code in a controlled environment
• Voice Response: Speaks responses using offline TTS
• VS Code Integration: Seamless integration as a VS Code extension
• Hugging Face Deployment: Ready for deployment on Hugging Face Spaces

Project Structure

Tittoo/
├── utils/
│   ├── __init__.py
│   ├── stt.py           # Speech-to-Text service (Whisper.cpp & OpenAI)
│   ├── llm_local.py     # Local LLM service (Ollama)
│   ├── llm_fallback.py  # Cloud LLM service (OpenAI)
│   ├── runner.py        # Safe Python code execution
│   └── tts.py           # Text-to-Speech service
├── vscode-extension/
│   ├── package.json     # Extension metadata and dependencies
│   ├── tsconfig.json    # TypeScript configuration
│   └── src/
│       └── extension.ts # Extension implementation with webview
├── app.py               # FastAPI backend application
├── huggingface_app.py   # Hugging Face deployment configuration
├── requirements.txt     # Python dependencies
├── setup_tittoo.ps1     # Windows setup script
└── README.md            # This file

Prerequisites

• Python 3.8+ installed
• Node.js and npm installed (for VS Code extension)
• Whisper.cpp compiled and available
• Ollama installed with llama3 model pulled

Setup Instructions (Windows)

1. Clone the repository:

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

2. Run the setup script:

   .\setup_tittoo.ps1
   

3. Configure environment variables:

   • Edit the .env file created by the setup script
   • Set the path to your Whisper.cpp executable
   • Optionally add your OpenAI API key for cloud fallback

4. Pull the Llama3 model in Ollama:

   ollama pull llama3
   

Running the Application

Backend Server

Start the FastAPI backend server:

uvicorn app:app --host 0.0.0.0 --port 8000

VS Code Extension

1. Open the vscode-extension folder in VS Code
2. Press F5 to start debugging
3. The extension will be loaded in a new VS Code window
4. Use the command palette (Ctrl+Shift+P) and type "Tittoo: Start" to open the UI

Hugging Face Deployment

1. Create a new Hugging Face Space with a Gradio SDK
2. Upload all files to the Space
3. Install dependencies:

   pip install -r requirements.txt
   pip install gradio
   

4. Set the Space to use the huggingface_app.py as the entry point

Local vs Cloud Fallback

Tittoo automatically switches between local and cloud services based on:

1. Availability: If local services (Whisper.cpp or Ollama) are not available or fail
2. Performance: If local processing takes too long
3. Manual Override: When the "Use Cloud" option is selected in the UI

The fallback mechanism is implemented in each service class and provides seamless transitions between local and cloud processing.

Usage Examples

Voice Mode

1. Click the "Record" button in the UI
2. Speak your command (e.g., "Tittoo, ek Python program likh jisme Fibonacci print hota ho")
3. Click "Stop" when finished
4. Tittoo will transcribe your speech, process the request, and respond

Text Mode

1. Type your command in the text input field (e.g., "Write a Python function to check if a number is prime")
2. Click "Send"
3. Tittoo will process your request and respond with the appropriate code or explanation

Troubleshooting

• STT Issues: Ensure Whisper.cpp is properly compiled and the path is correctly set in the .env file
• LLM Issues: Check if Ollama is running and the llama3 model is downloaded
• VS Code Extension: If the extension doesn't load, check the VS Code Developer Tools console for errors

License

MIT

---

✅ One-line PowerShell command to run backend:

uvicorn app:app --host 0.0.0.0 --port 8000

✅ One-line VS Code command to launch extension:

Tittoo: Start

✅ Example prompt for testing:

Tittoo, ek Python program likh jisme Fibonacci print hota ho.