using local ollama LLM example is broken https://github.com/evalstate/fast-agent/tree/main/examples/mcp_basic_ollama_agent

[suvarchal]

error:

i guess the core issue is probably trying to use 2 kinds of OpenAI-API models in the example https://github.com/evalstate/fast-agent/tree/main/examples/mcp_basic_ollama_agent

and default augumented openai llm is gpt-4o and the config uses one url
https://github.com/evalstate/fast-agent/blob/8891d00e5d069eaddebadf8bbf1d775426e3ef1c/examples/mcp_basic_ollama_agent/mcp_agent.config.yaml#L23-24
and the example needs 2 openai model api's and keys.

fast-agent is wonderful, I think having atleast one purely-local llm with mcp tools is worth the effort.

[yizhangliu]

You can use "request_params=RequestParams(model=your_ollama_model)" in every "llm.generate_str()".

[evalstate]

Hi @suvarchal - great you are enjoying fast-agent. I'm wondering if it's worth adding an ollama provider similar to the DeepSeek one: #4 ?

[evalstate]

Hi @suvarchal -- please see this update : #12 (comment)