API service falling back to Mistral for file uploads for no (apparent) reason

[bwalding]

We have successfully configured one Librechat environment (staging), and are now setting up our main one (production) - however file uploads are refusing to work. File uploads are terminated almost immediately with a 500 error from api.

The error we see in the log is as follows (which is kind of curious as it seems to be a fallback for when RAG API isn't working?)

ERROR 2025-05-27T08:48:30.280929995Z [resource.labels.containerName: api] [getUserPluginAuthValue] No plugin auth OCR_BASEURL found for user xxx
ERROR 2025-05-27T08:48:30.282343906Z [resource.labels.containerName: api] [getUserPluginAuthValue] No plugin auth OCR_API_KEY found for user xxx
ERROR 2025-05-27T08:48:30.282667845Z [resource.labels.containerName: api] Error uploading document to Mistral OCR API An error occurred while setting up the request: No plugin auth OCR_API_KEY found for user xxx
ERROR 2025-05-27T08:48:30.282864726Z [resource.labels.containerName: api] [/files] Error processing file: Error uploading document to Mistral OCR API An error occurred while setting up the request: No plugin auth OCR_API_KEY found for user xxx
INFO 2025-05-27T08:49:24Z Pulling image "ghcr.io/danny-avila/librechat-rag-api-dev-lite:v0.5.0"

I'm kind of out of ideas - connectivity between the api container and rag-api containers appears to work - I can curl to the rag-api based on the environment variables (and have verified it's rag-api and not rag_api for us - due to k8s container naming in our charts).

The API server is claiming - RAG API is running and reachable at http://rag-api:8000. - but I've seen comments online that maybe indicate that is not "completely" true.

I feel like this is some subtle error that I'm missing - I've turned on debug logging and there's nothing I can see.

My staging and production environments should be the same (production has a firewall ingress on api, but that's the only thing different that I know of)

Has anyone got any inspirational thoughts or is there some common case that I've missed that can go wrong?

My bug report I guess is... "it should be more obvious why the api is falling back to mistral rather than using rag-api"!

[bwalding]

So it all miraculously started working - I don't think I changed anything though (after it wasn't working). My working theory is that there's a race condition somewhere - maybe one container booting faster than expected. But for now my problem has gone away.