Support for GPT-5 Model Family

[ChronicLynx]

Does the library work with the OpenAI Responses API or only Chat Completion API for now? The library seemed to have trouble capturing my interactions at all over the Responses API.

Also:

When setting the model to a GPT-5 model it appears the background processing tries to call it with unsupported parameters.

ERROR: Fallback memory processing failed for x: Error code: 400 - {'error': {'message': "Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.", 'type': 'invalid_request_error', 'param': 'max_tokens', 'code': 'unsupported_parameter'}}

[harshalmore31]

Thanks for the detailed report! You've identified two real issues:

1. Responses API Support: Confirmed - the library currently only supports the Chat Completions API. The Responses API uses a different architecture that our interception system doesn't capture yet.

2. GPT-5 max_tokens Error: Valid bug. The library uses the deprecated max_tokens parameter, which GPT-5 models reject in favor of max_completion_tokens. This affects background memory processing in three locations.

Both issues are now tracked. For GPT-5, a quick workaround would be falling back to GPT-4 models for memory operations until we patch the parameter usage. we will push an update for this !