Adding Models¶
In this tutorial, you'll explore model catalogs, add models from both Ollama and OpenRouter, and use them in an agent.
Time: ~10 minutes
Prerequisites: Getting Started complete. Verify the daemon is running:
If this errors with "Cannot reach registry", the daemon isn't running.
Two Model Sources¶
VlinderCLI supports two model catalogs:
| Catalog | Runs | Cost | Best for |
|---|---|---|---|
| Ollama | Locally on your machine | Free | Development, privacy, offline work |
| OpenRouter | Cloud API | Pay per token | Larger models, production quality |
The installer already set up Ollama and registered phi3. In this tutorial you'll add more models from both sources.
Step 1: See What You Have¶
You should see phi3 registered from the install step.
Step 2: Browse Available Models¶
Ollama¶
See what's available locally through Ollama:
Filter by name:
OpenRouter¶
Browse cloud models via OpenRouter:
Note
OpenRouter requires an API key. Set it before browsing:
Or add it to ~/.vlinder/config.toml:
Step 3: Add an Ollama Model¶
Pull a model through Ollama first, then register it with Vlinder:
The --catalog ollama flag is optional — Ollama is the default catalog.
Verify it's registered:
Step 4: Add an OpenRouter Model¶
With your API key configured, add a cloud model:
This registers the model without downloading anything — inference happens via the OpenRouter API.
Step 5: Use a Model in Your Agent¶
Open your agent's agent.toml and add the new model to the requirements:
[requirements.models]
phi3 = "ollama://localhost:11434/phi3:latest"
llama3 = "ollama://localhost:11434/llama3:latest"
Your agent code can now call infer() with either model. The model alias (left side) is what your agent uses; the URI (right side) tells Vlinder where to route the request.
For OpenRouter models, the URI uses the openrouter:// scheme:
Step 6: Remove a Model¶
This deregisters the model from Vlinder. It doesn't delete the Ollama model from disk — you can re-add it anytime.
What You Learned¶
- VlinderCLI has two model catalogs: Ollama (local) and OpenRouter (cloud)
- How to browse available models with
vlinder model available - How to add and remove models with
vlinder model addandvlinder model remove - How to reference models in
agent.tomlusingollama://andopenrouter://URIs
Next Steps¶
- Your First Agent — scaffold and run your own agent
- Your First Fleet — compose multiple agents with delegation
- Manage Models — embedding models and advanced configuration
- model.toml reference — full manifest schema