Help Desk Voice Bot

A highly modular, decoupled Voice Bot leveraging Retrieval-Augmented Generation (RAG). The system features a modern FastAPI backend and a glassmorphic Streamlit frontend, allowing voice-to-voice interaction using state-of-the-art open models.

Key Features

• Decoupled Architecture: Independent frontend and backend for easy scalability.
• Provider Pattern: Easy to swap out different LLMs, Vector Stores, and Voice models.
• RAG Enabled: Powered by LangChain and Pinecone.
• Voice Capabilities: Built-in support for Speech-To-Text (Whisper / OpenRouter) and Text-To-Speech (Edge-TTS).

Project Structure

Voice-bot/
├── backend/                  # FastAPI Application
│   ├── api/                  # Routes and endpoints
│   ├── core/                 # Config and base interfaces
│   ├── providers/            # Provider implementations and routing files
│   │   ├── vector_store.py   # Factory routing for Vector DBs (e.g. Pinecone)
│   │   ├── llm_provider.py   # Factory routing for LLMs (e.g. OpenRouter)
│   │   └── ... 
│   ├── rag/                  # RAG logic (Document loaders & LCEL Chains)
│   ├── voice/                # Audio processing orchestration
│   └── main.py               # Entrypoint
└── frontend/                 # Streamlit UI
    └── app.py                # UI Code

Getting Started

1. Backend

Navigate to the backend folder:

cd backend
pip install -r requirements.txt
cp .env.example .env

Fill in the .env file with your OpenRouter, Pinecone, and (optional) HuggingFace keys. Then start the server:

uvicorn main:app --reload

2. Frontend

In a new terminal, navigate to the frontend folder:

cd frontend
pip install -r requirements.txt
cp .env.example .env

Start the Streamlit application:

streamlit run app.py

Modularity & Providers

The project uses routing files like backend/providers/vector_store.py and backend/providers/llm_provider.py to dynamically load the implementations specified in your .env. The API uses dependency injection (backend/api/dependencies.py) so providers are created once and reused.

If you wish to change from openrouter STT to local_whisper, simply update the STT_PROVIDER=local_whisper setting in your .env without modifying any code!

Testing

From the Voice-bot directory (project root), with backend dependencies installed:

cd Voice-bot
pip install -r backend/requirements.txt -r backend/requirements-dev.txt
PYTHONPATH=. python -m pytest tests/ -v

Some tests are skipped when optional packages (e.g. edge_tts, langchain_huggingface, whisper) are not installed. For full API tests, ensure the backend .env is configured and all dependencies are installed.