Is there any sense to pass many tool schemas and use tool_choice not set to "auto"?

API
1

The documentation suggests passing a list of tool schemas - and then using the tool_choice parameter to force the LLM to choose one of them. Isn’t that kind of wasteful in terms of tokens? If you want to use just one tool - then maybe pass just one schema?

For my usage I wrote a wrapper around ‘completions.create’:

def openai_query(self, tool_schemas, force_auto_tool_choice=False):

        if len(tool_schemas) == 1 and not force_auto_tool_choice:
            tool_choice = tool_choice={'type': 'function', 'function': {'name': tool_schemas[0]['function']['name']}}
        else:
            tool_choice = "auto"

        completion = self.client.chat.completions.create(
            model=self.model,
            messages=self.prompt.to_messages(),
            tools=tool_schemas,
            tool_choice=tool_choice
        )
        return completion

I infer tool_choice from the number of schemas. To cover all the possible cases I added that force_auto_tool_choice optional parameter that would make it possible for the LLM to choose no function call even when passed a schema with just one tool. For now I don’t need that.

Anything against that wrapper?