Add OpenRouter as Ai Assistent Connection

[tobiashochguertel]

Alternatively, we could consider enabling a generic LLM API URL that adheres to the OpenAI schema APIs. This approach would provide flexibility and ease of integration. A straightforward solution would be to allow users to specify the OpenAI URL themselves. By doing so, we would be able to utilize OpenRouter effectively, streamlining the process and enhancing user experience.

[antonreshetov]

OpenRouter makes sense as an additional provider. I’d start with a dedicated integration so we can test model selection, streaming, and tool calling properly. A custom OpenAI-compatible URL is also worth considering, but it needs more than exposing the URL field: our OpenAI integration uses the Responses API, and compatibility can vary across services. I’d keep that as a separate follow-up.

[antonreshetov]

OpenRouter is now supported as a dedicated provider for the AI assistant (#971). It will be available in the next release.

To use it, open Preferences → AI assistant, choose OpenRouter, enter your OpenRouter API key and click Save connection. The model list includes only models that support tool calling, so the assistant can search and perform actions.

A custom OpenAI-compatible URL is still a possible follow-up, but it is not part of this change.