Mem0 `self._store_messages` blocks the conversation flow

[lucasrothman]

pipecat version

0.0.65

Service Name

Mem0

Service or model version

No response

Issue Description

When using the Mem0 memory service the self.store_messages(context_messages) (line 194) seems to be blocking the conversation. This should be refactored to to be called async (or made non-blocking some other way).

Also, the position input param seems to be unused.

Reproduction Steps

Add Mem0MemoryService. Observe that there is a noticeable long pause after logs for storing messages in Mem0 and the LLM service generating chat.

Expected Behavior

The storing of messages should not be blocking, only the retrieving and enhancing of context.

Actual Behavior

Storing the messages seems to be blocking, seems like it's waiting for responses from all the API calls.

Error Logs

[lucasrothman]

@Dev-Khant

[lucasrothman]

Also could be useful to do a quick check to make sure duplicate information isn't being added to the context.

[erikrubinov]

Same problem, generating an answer takes up to 10 seconds

[harshjain2408]

@markbackman @aconchillo Is there any way we can solve this issue? Is there a way we can make the data-storing step of Mem0 run in the background/separately without hampering the conversational flow of the pipeline, as we only require data retrieval during the conversation flow?

[markbackman]

Sorry for the delay. I've asked the mem0 team to take a look. They maintain this integration.

[markbackman]

Fix is here:
#4156