to have ollama provider on agents(previed)
to have mi ollama running as agent with my tools
they can use the same providers if they have installed ollama
Sorry but i do not know how to build something in node.js
I believe this is already supported
1 „Gefällt mir“
Not in my case and i am running in docker with the updated version of the n8n
Do you have an Ollama container running?
Yes in the same docker network, and i can use it in the docker agent inside a workflow, and chat but not in the agent (previed)
I think you can do it this way
i tried but i can not put my ollama inside the chat gtp credential type, maybe i am missing something but the wierd thing is that in your image, you have ollama with your credential i think but at least the provieder and i am not, there is something that i am missing like a configuraction or something in the new agents (previed) section?
This should help
Hi @sneakylenny Welcome!
Docker Model Runner’s Ollama-compatible endpoints are a thin shim and tool calls don’t survive it. Your capture shows exactly that: finish_reason comes back as tool_calls but the message carries no tool_calls array, so the agent gets an empty assistant turn and returns { "output": "" }. Docker documents function calling only on its OpenAI-compatible surface, which is why all five models behaved identically.
Swap the Ollama Chat Model node for the OpenAI Chat Model node…
You need to create an OpenAI credential using Ollama base url
1 „Gefällt mir“
