You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm running into issues when my model needs to make successive tool calls for a single user query. Continue provides the tools list on the initial request, allowing the model to make a tool call. However, on the followup request when Continue provides the tool response, it does not provide the original tools list, so the model can no longer see what tools are available and cannot successfully make any subsequent tool calls.
Here's an example of the output:
The model calls ls successfully since tools is provided on the first call, but after the model gets the tool response, there is no tools list provided, so it ends up trying to call a hallucinated function that doesn't exist. Shouldn't the full tools list be provided with each request, not just the initial one?
I'm running gpt-oss-120b on a custom server that emulates Ollama.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
I'm running into issues when my model needs to make successive tool calls for a single user query. Continue provides the
toolslist on the initial request, allowing the model to make a tool call. However, on the followup request when Continue provides the tool response, it does not provide the originaltoolslist, so the model can no longer see what tools are available and cannot successfully make any subsequent tool calls.Here's an example of the output:

The model calls
lssuccessfully sincetoolsis provided on the first call, but after the model gets the tool response, there is notoolslist provided, so it ends up trying to call a hallucinated function that doesn't exist. Shouldn't the fulltoolslist be provided with each request, not just the initial one?I'm running gpt-oss-120b on a custom server that emulates Ollama.
Beta Was this translation helpful? Give feedback.
All reactions