AI assistant: Ollama doesn't seem to work, and across several providers, the displayed model is different than the configured model
AI assistant: Ollama doesn't seem to work, and across several providers, the displayed model is different than the configured model
\1. I tried using Ollama, hosted locally (same computer as RDM) and it doesn't work:
I can see the query from RDM against the Ollama server - Ollama is able to query the models
When I send a chat message through RDM, I see this error in Ollama's server.log:
[GIN] 2026/04/08 - 12:35:24 | 404 | 1.0388ms | 127.0.0.1 | POST "/api/chat"
The error may be in the display-name (the name of the model shown on the RDM chat window - instead of saying "phi4:latest" it simply says "phi4") I also notice the model selector changes to models which don't exist in the local Ollama server, and though I can change that to the model I do have loaded, it returns a '404 (not found)' error that RDM session.
\2. I tried Gemini as well - it too allows me to select a model in the entry for AI assistant, but when I load that entry to chat with it, the available models are reduced to three (three older models) and I can't tell which model is being used when I send a message
Last, not so much a bug as much a question: Is there a way to track "AI assistant" interactions in RDM? For example, which API key, user, destination, model selected, message sent and response received?
All Comments (1)
Hello @oschaudt
1) Sorry about the problems you are having connecting to Ollama. You are correct there definitely is a problem loading the correctly installed models in the drop down, I have logged a ticket and I will fix the issue shortly. If you select "phi4" and have "phi4" installed though it should work, maybe try changing the model and hard code it to "phi4", remove the ":latest" manually. Here is my RDM with a local Ollama and phi4:
2) Correct it's a similar issue we have with Gemini models, the drop down while in a chat does not show the correct models.
3) This is a great feature request, I will log that one as well, we could put when changes happen in the connection log? Or in a report?
Thanks for the feedback this is very valuable for us to improve the AI Assistant inside RDM!
Paul Dumais
32f8b687-c093-41da-b39c-3f1b8cb548fb.png