Memory usage over time

[proschowsky]

We are running librechat-rag-api-dev-lite image and are experiencing increased memory usage over time. We are using version: v0.6.0 (but we have experienced similar issues with prior versions).

We have few daily requests:

• /query: <1000 per day
• /embed: <300 per day (<1000 calls to embedding endpoint, ~80 retries due to rate limits on endpoint)

Last 24 hours:

7 days, until restart:

The spikes seems to be related to requests being made to the /embed endpoint.
There is nothing unexpected in the logs.

I have not done any memory inspections or further analysis. Are other people experiencing similar behavior?