Can't seem to call tools with gpt-oss 20b

[fetchezar]

Hi! I configured gpt-oss 20b on my ollama instance like so, this is my entire agent configuration file:

name: agentconf
version: 1.0.0
schema: v1
models:
- name: gpt-oss 20b
  provider: ollama
  model: gpt-oss:20b
  roles:
    - apply
    - chat
    - edit
  capabilities:
    - tool_use
- name: qwen2.5-coder 3b
  provider: ollama
  model: qwen2.5-coder:3b
  roles:
    - autocomplete
  defaultCompletionOptions:
    contextLength: 16384
- name: nomic-embed-text latest
  provider: ollama
  model: nomic-embed-text:latest
  roles:
    - embed

However when I ask in plan or agent mode if gpt-oss can read a specific file, I get the following invalid tool call error:

assistant failed because the arguments were invalid, with the following message: Tool assistant not found
Please try something else or request further instructions.

What does this mean? I can't find anything related to a "tool assistant" in the configuration.

[duaneking]

I'm not sure continue can call any tools at all on any model; Are you able to get it working on any other model that is self hosted?

[fetchezar]

GPT OSS is one of the suggested models for the agent/plan mode, that's why I tried. I imagine it ought to be soported if it's here https://docs.continue.dev/customize/models

That being said, I gave a try on the other similarly sized model, devstral small 2 which is supposed to be designed for agent usage, and it fared much better. It's not perfect but it can run tools like file reads and so on.

[arthurcohen]

This issue is related to this one #8098. Someone noticed that the extension only includes the available tools on the first call to Ollama (as it is initiated by a user message). Subsequent messages do not contain the tools array. He noticed that if the tools array is included when the message role is set to "tool" the agent stops messing up tool calls.

The fix seems too simple, I can't understand how this is still a problem after all this time.