Start local. Get to your first request fast.
Install Izwi, pull a model, run izwi serve, and make your first /v1 request without detours.
Quickstart
The shortest path through the docs is simple: install Izwi, pull a model, start the server, then make a request against the local OpenAI-compatible interface.
- Install Izwi on your target machine.
- Pull a model for the workflow you want to test.
- Run izwi serve to start the local API surface.
- Use the getting started guide to make your first request.
Docs map
| Section | What it helps you do |
|---|---|
| Quickstart | Install, pull a model, run the server, and make a request |
| Features | Understand speech recognition, text to speech, diarization, and related workflows |
| Models | Choose what to pull and when to use it |
| CLI | Operate the runtime locally, including pull, serve, and status commands |
| Installation | Use platform-specific install steps when the quickstart is not enough |
| Troubleshooting | Resolve hardware, install, or model issues without guessing |
Activation path
The docs are most useful when they keep you moving toward one clear outcome: a local runtime that responds to a real request.
- Use Getting started when you want the fastest first-run path.
- Use Features when you already know the workflow and need deeper detail.
- Use Models when you need to decide what to download and test.
- Use CLI when you are operating the runtime directly.
- Use Troubleshooting when hardware or model behavior blocks the path.
Getting Help
- GitHub Issues — Report bugs or request features
- Discussions — Ask questions and share ideas
License
Izwi is open source software licensed under MIT.