Replies: 3 comments
|
Yes, I am planning to add a small, lightweight LLM to the Kixote service for licensed users. However, I have not had the time to explore this yet, and I am not sure whether such a setup would provide sufficient quality and performance. Regarding local Ollama: the main issue is connectivity. A Typemill website hosted at a regular web host cannot access a service running on localhost on a user’s machine. This would only work for local Typemill installations, which are not the standard use case. In all other cases, the local Ollama instance would need to be exposed to the public internet. That is usually not a good idea, or at least not trivial, due to security, networking, and maintenance concerns. I have not worked with Ollama yet, so I do not know whether it offers built-in features to support such scenarios (for example secure tunneling or access control). If you have experience with this kind of setup, I would be interested to hear about it. |
|
I plan a refactoring of the ai service connector for the next release so you can connect to any ai-service that you want (supporting Openai-cmpatible apis (ollama) and anthrophic api. |
|
Ok, this is live now with Version 2.23.0 |
Uh oh!
There was an error while loading. Please reload this page.
Hello Sebastian,
I would like to suggest Ollama as another AI service. Ollama can be operated locally and does not incur any running costs. It is certainly an interesting solution for small Typemill sites, without having to take out more and more subscriptions.
All reactions