🎙️ Aniquen Server

Aniquen Server is the backend inference engine powering the Aniquen Voice Assistant.

Instead of running heavy AI models directly on the client device, this server performs all machine learning inference using ONNX Runtime, processes live microphone audio through WebSockets, detects the "Hey Aniquen" wake word in real time, and securely proxies speech recognition and LLM requests without exposing API keys.

The architecture is designed for low-latency voice interaction while keeping client devices lightweight.

---

📑 Table of Contents

• Features
• Architecture
• Screenshots
• Technologies
• Project Structure
• How It Works
• Installation
• API Endpoints
• Future Improvements
• Contributing
• License
• Author

---

✨ Features

✅ Real-Time Wake Word Detection

✅ ONNX Runtime AI Inference

✅ FastAPI Backend

✅ Low-Latency WebSocket Streaming

✅ Voice Activity Detection (VAD)

✅ Automatic Audio Gain Normalization

✅ Mel Spectrogram Generation

✅ Embedding Model Pipeline

✅ Keyword Spotting Model

✅ Rolling Probability Detection

✅ Secure Whisper API Proxy

✅ Secure Groq LLM Proxy

✅ Environment Variable API Keys

✅ Lightweight Client Architecture

✅ Optimized for Always-On Servers

---

🏗 System Architecture

Microphone
      │
      ▼
Browser
      │
WebSocket
      │
      ▼
FastAPI Server
      │
      ├────────► Audio Buffer
      │
      ├────────► Gain Normalization
      │
      ├────────► Voice Activity Detection
      │
      ├────────► Mel Spectrogram Model
      │
      ├────────► Embedding Model
      │
      ├────────► Wake Word Model
      │
      ▼
Wake Probability
      │
      ▼
Detection Result

---

⚙ Technologies Used

• Python 3.12
• FastAPI
• Uvicorn
• ONNX Runtime
• NumPy
• WebSockets
• HTTPX
• Python Multipart

AI Models

• Mel Spectrogram ONNX
• Embedding Model ONNX
• Hey Aniquen Wake Word Model

Cloud APIs

• Groq Whisper
• Groq LLM

---

📂 Project Structure

aniquen-server/
│
├── server.py
├── requirements.txt
├── .python-version
├── embedding_model.onnx
├── melspectrogram.onnx
├── hey_Aniquen.onnx
│
├── firmware/
│
└── README.md

---

⚡ How It Works

Step 1

The browser records microphone audio.

↓

Step 2

Raw Float32 PCM audio is streamed to the FastAPI server using WebSockets.

↓

Step 3

Incoming audio is buffered and normalized.

↓

Step 4

Voice Activity Detection determines whether speech is present.

↓

Step 5

The Mel Spectrogram ONNX model converts audio into spectrogram features.

↓

Step 6

The Embedding Model generates feature embeddings.

↓

Step 7

The Wake Word Model predicts the probability of the "Hey Aniquen" keyword.

↓

Step 8

The server sends detection results back to the client in real time.

↓

Step 9

When activated, speech transcription and AI chat requests are securely proxied through the backend.

---

🌐 API Endpoints

WebSocket

/ws

Streams microphone audio and returns wake-word probabilities.

---

Speech Transcription

POST /api/transcribe

Proxies audio to Groq Whisper while keeping API keys secure.

---

AI Chat

POST /api/chat

Proxies LLM requests to Groq without exposing credentials.

---

🚀 Installation

Clone the repository

git clone https://github.com/dhruvpandit46/aniquen-server.git

Go inside the project

cd aniquen-server

Install dependencies

pip install -r requirements.txt

Run

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

---

🔐 Environment Variables

GROQ_WHISPER_KEY=your_key_here

GROQ_LLM_KEY=your_key_here

Never commit API keys to the repository.

---

🎯 Future Improvements

• GPU acceleration
• Multi-language wake words
• Speaker identification
• Noise suppression
• Echo cancellation
• Multi-device synchronization
• Streaming speech recognition
• Docker deployment
• Kubernetes support
• Authentication
• Rate limiting
• Monitoring dashboard

---

🤝 Contributing

Contributions are welcome.

1. Fork the repository

2. Create your feature branch

3. Commit your changes

4. Push your branch

5. Open a Pull Request

---

📜 License

Licensed under the MIT License.

MIT © 2026 Dhruv Pandit.

See the LICENSE file for full license details.

---

👨‍💻 Author

Dhruv Pandit

GitHub

https://github.com/dhruvpandit46

LinkedIn

https://linkedin.com/in/dhruv-pandit-755786326

Instagram

https://instagram.com/dhruv_pandit2007

---

🌌 Aniquen Ecosystem

Repository	Description
🤖 Aniquen	AI Voice Assistant
🧠 aniquen-server	AI Backend & Wake Word Server
🎤 Anique AI Voice	Text-to-Speech API
💬 Anique AI	Browser-Based AI Chatbot
🎬 Anique	Multimedia Platform

---

⭐ Support

If you found this project useful,

please consider giving it a ⭐ on GitHub.

It helps support future development.