Local models

Point Tyrn at a model server on your Mac or your network. Open Settings > Inference > Local. The page says “Use models running on your machine or network, like Ollama or LM Studio.” Most local servers need no key.

Add a server

1. Start your model server.
2. Open Settings > Inference > Local and click Add server.
3. Choose a Server type. The Base URL fills in with the default address.
4. Add an API key (optional) if your server needs one.
5. Leave Active on. Inactive servers are skipped by Tyrn router.
6. Save. Tyrn tests the server and reports the models it found, for example “Connected. 2 models: a, b.”

Presets

Server type	Default address
Ollama	http://localhost:11434/v1
LM Studio	http://localhost:1234/v1
vLLM	http://localhost:8000/v1
llama.cpp	http://127.0.0.1:8080/v1
Llamafile	http://127.0.0.1:8080/v1
Lemonade Server	http://localhost:13305/api/v1
NVIDIA Triton	http://localhost:8000/v1
Docker Model Runner	http://localhost:12434/v1
XInference	http://localhost:9997/v1
oobabooga	http://localhost:5000/v1
MLX Gemma 26B	http://localhost:11435/v1
MLX Qwen 3.8 27B	http://localhost:11436/v1

Custom endpoint

For any other server or gateway, choose Custom OpenAI-compatible endpoint or Custom Anthropic-compatible endpoint. Give it a name and a base URL such as https://host:port/v1. It can run on this machine or your network.

Use a local model

Pick it in a session whose Inference is Tyrn router. See Start a session. Model ids appear exactly as the server reports them.

If a server is not running, adding it fails with a message from the test. Start the server and try again.

Next steps

• Usage-based keys
• Autocomplete