| name | maddie-windows-runtime |
| description | Use when: debugging or changing Windows runtime support for MADDIE, PowerShell startup, llama.cpp endpoint configuration, docker compose, OmniVoice service URL, or app_settings.json defaults. Keywords: Windows, PowerShell, 0.0.0.0:8080, run.ps1, Docker, OmniVoice. |
MADDIE Windows Runtime
Use this skill when the task is about starting, configuring, or debugging the app on Windows.
Primary files
run.ps1
docker-compose.yml
docker/omnivoice/Dockerfile
data/app_settings.json
app/llm_client.py
README.md
Runtime facts
- The configured llama.cpp URL may be
http://0.0.0.0:8080, but the Python client normalizes it to loopback for actual requests.
- The recommended Windows path is: llama.cpp on the host, OmniVoice in Docker, FastAPI app in the host venv.
- The default OmniVoice HTTP service runs on
http://127.0.0.1:8010.
Safe workflow
- Check
data/app_settings.json before changing runtime code.
- If TTS is failing, verify
docker compose up omnivoice and GET /health on port 8010.
- If LLM calls fail, verify
/v1/models on the configured llama.cpp host.
- Prefer fixing URL normalization or startup scripts over hardcoding machine-specific paths.