Ollama support

[JCPOCSIN]

Hi team!

Do you plan to add Ollama support in AgentKit?
It would enable me to use your product at my work. As a Swiss State we can only use LLMs locally hosted in our own infrastructure.

Cheers!
JC

[djfarrelly]

Hey @JCPOCSIN - we can add Ollama support given the right config options. The models are currently defined in our main JS repo and imported through Agent Kit. We plan to extract those into a neutral package at some point.

https://github.com/inngest/inngest-js/tree/main/packages/inngest/src/components/ai/models

Is your Ollama model open to the internet w/ a base URL and auth, or is it on the closed network?

[JCPOCSIN]

Hi @djfarrelly ,
For now it's on a lab server in a closed network directly accessible in HTTP on the standard port 11434. I also use a local instance on http://127.0.0.1:11434