Dialogos: Real-time Russian ASR System

Dialogos is a real-time Automatic Speech Recognition (ASR) system for Russian language with low latency and minimal VRAM usage. It features a Docker-based backend running the T-one ASR engine (streaming WebSocket + endpointing) and a Windows client that captures audio via WASAPI and streams it over WebSocket.

Features

• Low Latency: Real-time streaming ASR with minimal delay
• Lightweight: Optimized for low VRAM consumption
• Multi-client: Supports multiple simultaneous connections
• Endpointing: Automatic phrase detection and segmentation
• GPU Accelerated: Leverages NVIDIA GPU for fast inference

Documentation

For detailed documentation, please see:

• Project Overview
• Architecture Documentation
• Client Documentation
• Server Documentation

Quick Start

Prerequisites

• Docker installed on your system
• Go 1.19+ for building the client
• Audio input device (microphone)

Building and Running

1. Build the client:

   make build-client

2. Run the server:

   docker-compose up -d

3. Run the client:

   make run-client

Project Structure

• asr-client/ - Windows client application for audio capture and streaming
• asr-server/ - WebSocket server implementation for speech recognition
• docs/ - Project documentation
• Dockerfile - Server Docker image definition
• docker-compose.yml - Multi-container deployment configuration

License

This project is licensed under the Apache License, Version 2.0. See LICENSE for details.