[WARNING] [LLM]: Invalid response format

[phazonix]

Hi, I'm running Llama3.2 locally on Ollama however when I run the main.py I encounter an invalid response format error:

`uv run main.py

Enter a query: can you open a word document
[INFO] [Tree] Tree State capture took 0.22 seconds
[INFO] [Desktop] Desktop State capture took 0.36 seconds
[INFO] [Agent] 🎯 Step: 1/25
[WARNING] [LLM]: Invalid response format, Retrying attempt 1/3...
Response: role='tool' content=None id='tool-call-id' name='desktop_tool' params={'action': 'launch', 'desktop_name': 'Desktop 1'}
[WARNING] [LLM]: Invalid response format, Retrying attempt 2/3...
Response: role='tool' content=None id='tool-call-id' name='desktop_tool' params={'action': 'create', 'desktop_name': 'Desktop 1'}
[WARNING] [LLM]: Invalid response format, Retrying attempt 3/3...
Response: role='tool' content=None id='tool-call-id' name='app_tool' params={'mode': 'launch', 'name': 'notepad.exe'}
Invalid JSON format: Expecting property name enclosed in double quotes: line 1 column 2 (char 1) Content: {'mode':
'launch', 'name': 'notepad.exe'}`

So this model seems to be incompatible - Is there a list of confirmed working models or a fix for this issue?

Thanks

[Jeomon]

Am updating the model wrappers use qwen:vl 4b.
Now onwards we are sticking with tool calling

Make sure the model support tool_call