Skip to content

Calling get_next_assistant_message() with function_call_option="none" and then parsing the response as if the function call was made #3

@ve1nard

Description

@ve1nard

In act.py lines 203-220:
assistant_messages.append(get_next_assistant_message(system_message, user_messages, assistant_messages, model=agent_config.actor_model, function_call_option="none"))
since the argument is function_call_option="none", it is expected that the response will not contain any function selection by the model. However, later the response is parsed as
error_message = { 'tool_call_id': response['id'], 'name': response['function']['name'], 'return_value': json.dumps({ 'error_message': f'You did not provide the text content in the required template. Please provide the content to be inputted in the textfield in the following format: "Text: <your text content>". For example, if you want to input "Hello World" into the textfield, you should answer with "Text: Hello World".' }) }
as if a function selection was expected, which leads to the TyperError:
'name': response['name'], TypeError: string indices must be integers

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions