Providers & models
Bring your own keys, pick a model from the provider's live list, and set how hard it thinks.
Actovia does not resell model access. You add a key for the providers you use, and every request goes from your computer to that provider with your key.
Supported providers
| Provider | What you enter |
|---|---|
| OpenAI | API key, optional base URL |
| Azure OpenAI | API key, endpoint, API version, deployment name |
| Anthropic | API key, optional base URL |
| Google Gemini | API key, optional base URL |
| xAI | API key, optional base URL |
| Ollama | The address of your local Ollama server (no key) |
Open Settings → AI Providers, choose a Provider, and paste its key under Provider keys. Then Retrieve model list: Actovia asks the provider which models your key can use right now and shows them in the Model picker, searchable. A model the provider is retiring is marked retiring with its date, and one that is no longer offered asks you to pick a current one. Getting API keys says where to get each provider's key.
For Ollama, Status says Ollama is running once the server answers, and the list shows the models installed on your machine.
Saved keys
A saved key is never shown again — only its last four characters, with Saved beside it. Typing into the field replaces the whole key; deleting it clears the key. Changing a provider's Base URL clears its saved key (Actovia never sends a key to an address it was not saved for), so paste it again after the change. If a key ever cannot be read — for example after moving to a new computer — Settings → General lists the keys to enter again.
Switching models
The model chip in a chat's header shows the provider and model in use. Pick another from it at any time: the next message uses the new model, and earlier messages stay as they are.
Reasoning
Models that can think before answering get a Reasoning choice in the composer: Off, Low, Medium, High, Extra high or Max — only the levels the current model supports are offered. The chosen level shows as a pill (Reasoning: High); tap it to turn it off. A higher level takes longer and uses more tokens, and the thinking shows in the timeline.
