How Lumo Projects and Drive actually work | Clearing up some confusion
I've seen a few posts recently from people frustrated that Lumo "isn't seeing" their linked Drive files. I want to clear up how this feature actually works, because the behavior people are reporting isn't a bug, it's by design.
Connecting a Drive folder or adding files to Project Knowledge does not mean every file gets loaded into every response. That would be wasteful and would actively hurt response quality.
Here's what actually happens when you link Drive to a Project:
- Sync: Lumo pulls in and indexes the files it can access
- Retrieve: when you send a message, Lumo searches that index for what's relevant (typically fewer than 5 files, depending on relevance scores)
- Ground: only the matching files are included in the request to the model
- Respond: the model answers using that targeted context
The retrieval step is the whole point. A Project with 500 documents doesn't send 500 documents to the model, it sends the two or three that matter for your question.
If Lumo didn't use a file you expected it to, it most likely means retrieval didn't rank it as relevant not that the file is missing or the integration is broken. That's not to say that bugs can't exist, but current reports or messages here are 99% likely down to this behaviour.
Link everything you need. Trust the retrieval 🙂