AttributeError: 'NoneType' object has no attribute 'model_endpoint_type' when using Ollama to load local models with MemGPT

[Armanasq]

Describe the bug
I'm trying to use Ollama to load local models with MemGPT, but I'm encountering an AttributeError. The error message indicates that a NoneType object has no attribute model_endpoint_type. This error occurs when I attempt to create a MemGPT agent using the provided configuration.

Please describe your setup

• How did you install memgpt?
  • I installed it using pip install pymemgpt.
• Describe your setup
  • OS: Linux
  • Running memgpt via Jupyter Notebook

Additional context
I've tried different configurations and combinations, but the error persists. Below is the error traceback:

AttributeError: 'NoneType' object has no attribute 'model_endpoint_type'

The error occurs at this line in the memgpt_agent.py file:

if llm_config["model_endpoint_type"] in ["azure", "openai"] or llm_config["model_endpoint_type"] != config.default_llm_config.model_endpoint_type:
    # Additional code

I've followed the default example for using Ollama with MemGPT.

MemGPT Config
Here is the configuration I'm using:

config_list = [
    {
        "model": "NULL",
        "base_url": "http://localhost:11434/v1",
        "api_key": "ollama",
    }
]

config_list_memgpt = [
    {
        "preset": DEFAULT_PRESET,
        "model": "llama3.1",  # only required for Ollama, see: https://memgpt.readme.io/docs/ollama
        "context_window": 8192,  # the context window of your model (for Mistral 7B-based models, it's likely 8192)
        "model_wrapper": "chatml",  # chatml is the default wrapper
        "model_endpoint_type": "ollama",  # can use webui, ollama, llamacpp, etc.
        "model_endpoint": "http://localhost:11434",  # the IP address of your LLM backend
    },
]

Local LLM details

If you are trying to run MemGPT with local LLMs, please provide the following information:

• The exact model you're trying to use: llama3.1
• The local LLM backend you are using: Ollama
• Your hardware for the local LLM backend: Local computer running Linux