Memory Integration with Custom Endpoint

[cry-po]

Hi,

I'm looking to integrate the memory/conversation history feature for my use case. However, I am using a custom API endpoint instead of the standard OpenAI one.

What modifications are required on my custom endpoint to ensure compatibility with the memory system?

[jasongilman]

I had this same exact question. I did note the documentation says "Note: If you are using a custom endpoint, the endpoint value must match the defined custom endpoint name exactly." But it's not clear what "endpoint value" means. Is that the Agent provider?

[jasongilman]

I was able to get it working. A config like this works:

memory:
  disabled: false
  personalize: true  # Gives users the ability to toggle memory on/off, true by default
  tokenLimit: 2000  # Maximum tokens for memory storage
  messageWindowSize: 5  # Number of recent messages to consider
  agent:
    provider: "My Custom Endpoint"
    model: name-of-my-model


endpoints:

  custom:
  - name: "My Custom Endpoint"
    ...