ai-agents-configurable-llm-you-need-to-support-using-different-local-or-third-d533b48e

THEN Modify the function that calls the AI model to read a model identifier from a configuration file (e.g., config.ini). Use a conditional structure to dispatch to the appropriate API based on the configured model, falling back to a default like GPT-4. This allows flexible model selection without hardcoding.

import configparser

def call_ai_function(function, args, description, config_path='config.ini'):
    config = configparser.ConfigParser()
    config.read(config_path)
    model = config.get('AI', 'Chosen_Model', fallback='gpt-4')
    # ... build messages ...
    if model == 'gpt-4':
        response = openai.ChatCompletion.create(model=model, messages=messages, temperature=0)
    elif model == 'some_other_api':
        response = some_other_api_call(parameters)
    else:
        raise ValueError(f'Unsupported model: {model}')
    return response.choices[0].message['content']