Sandy's Chat Reactor

A multiservice application that reacts to chat messages via text generation and audio speak via TSS using local-only AI models.

Currently, supports Twitch integration (when viewers type !react <message>).

Architecture

[Twitch IRC] --> [.NET 10 Console App] --> [Ollama REST API]
                       |                          |
                       |  <-- reaction text ------+
                       |
                       +--------------> [Python TTS Service]
                       |                          |
                       |  <-- audio bytes --------+
                       |
                       +--> [NAudio Playback Queue]

Quick Start

Prerequisites

• .NET 10 SDK
• Docker with NVIDIA GPU support (for Ollama + TTS)
• A Twitch account with OAuth token

1. Start backend services

docker compose up -d
docker exec ollama ollama pull qwen3:4b

This will run up ollama (with loaded qwen3) for text generation and tts-service for TTS generation.

2. Configure the .NET app

Edit src/Sandy.ChatReactor.Console/appsettings.json.

3. Run the app

dotnet run --project src/Sandy.ChatReactor.Console

4. Test it

Type !react hello world in channel chat (Twitch).

Contributing

Contributions are welcome! Feel free to open a pull request with your changes.

Documentation

• Architecture & Design
• Twitch OAuth Setup
• Ollama Setup
• Qwen3-TTS Setup
• Building
• Running