Has anyone successfully configured Posit Assistant in Positron 2026.06.0 to use a local Ollama model via Custom Provider? I have Ollama serving qwen2.5-coder:7b on localhost:11434, the Custom Provider signs in successfully, Anthropic is signed out, OpenAI is disabled, and Posit AI is disabled in settings.json. However, Assistant still reports 'Posit credits depleted,' and ollama ps never shows the model being invoked.
Positron Version: 2026.06.0 (build 211)
Ollama: Installed and working
Model:qwen2.5-coder:7b
Ollama API: Verified at http://localhost:11434
Custom Provider: Successfully signs in
Anthropic: Signed out
OpenAI: Disabled
Posit AI: Disabled in JSON
Problem: Assistant reports "Posit credits depleted" and ollama ps never shows the model running during Assistant requests
I'm using qwen3.6:35b
I replaced the api url from Openai and replaced it with the ollama local api url http://localhost:11434/v1
and did put 'test' as the api key
Then you need to start ollama and check that it is running
Going to http://localhost:11434/ you should see the message 'Ollama is running'
Then in the "chat" section in the left, select openai as the llm provider and select the model you want to use; note that you need to execute in your terminal :
ollama run qwen:latest
For ollama to download the model and for it to appear in the list of models available from openai
I did get the qwen2.5-coder:7b model to be utilized by Posit Assistant, but alas, it isn't going to work out. The model is too small and my computer not powerful enough. I will stick to my Claude connection to Posit Assistant. My workflow is pretty minimal, so I can't justify getting a set-up like you must have to run the 35b model.
Thanks again, I am glad to have learned how to set it up and the lessons learned were very valuable.