Docker Guide

ai_term uses Docker to run the STT (Speech-to-Text) and TTS (Text-to-Speech) backend services.

Pre-built Images

Official Docker images are available on GitHub Container Registry:

Service	Image
STT	ghcr.io/vsaravind01/ai-term-stt:latest
TTS	ghcr.io/vsaravind01/ai-term-tts:latest

Quick Start

The easiest way to start the services is using the built-in CLI command:

ai-term start

This pulls the pre-built images and starts both services in the background.

Detached Mode

The command runs in detached mode by default (-d), freeing up your terminal immediately.

First Run

The first api call to the services will download the required models (Whisper/Coqui) and may take a while. Please be patient.

To check the status of running services:

ai-term status

Port Configuration

Service	Container Port	Host Port
STT	8001	8901
TTS	8002	8902

Port Conflicts

Ensure ports 8901 and 8902 are available on your host machine. If they are in use, the services will fail to start.

Manual Docker Compose

If you prefer to run Docker Compose directly:

docker compose up -d

To force rebuild of images:

ai-term start --build

Volumes and Caching

To speed up startup, the services mount local cache directories:

• Whisper Models: ~/.cache/whisper -> /root/.cache/whisper
• TTS Models: ~/.local/share/tts -> /root/.local/share/tts

GPU Support

If you have an NVIDIA GPU, you can enable hardware acceleration by modifying the Dockerfile and docker-compose.yml to use CUDA-enabled base images.

CPU Default

The default setup uses CPU for maximum compatibility across different hardware (Mac M-series, generic Linux). GPU support is not tested yet.