Does run cmds but outputs JSON

[baboulebou]

Describe the bug

This may be something I misunderstood, but I run interpreter locally and it outputs a JSON with what looks like commands it would run, but it's not actually running anything. Example:

`> what's the weather today ?

{"name": "execute", "arguments": {"language": "shell", "code": "curl -X GET https://api.weather.gov/gridpoints/MKX/25,40/forecast"}}`

What am I missing ?

Reproduce

just ask any question

Expected behavior

I expect interpreter to open the browser and find the weather

Screenshots

No response

Open Interpreter version

Version: 0.4.3

Python version

3.11

Operating System name and version

Mac OS

Additional context

No response

[nxtruong]

I have the same issue on MacOS. It prints out the JSON format response like you got, but it didn't execute the code and didn't even ask me to run or not. Seems pretty useless. Have you figured out a solution, @baboulebou ?

[Notnaton]

--no-tool-calling
Or
--no-supports-function-call

[Jinnan-Guo]

What model you are using? From my observation if I use some new models, e.g., qwen3 or llama3.3, the code does not execute. If I rollback to llama3 the code will be execute.