Gotcha. Not sure about your setup, but I think a lot of LLM tooling uses graph databases, which typically aren’t great for BI / traditional analytics.

I do a lot of data prep and modeling for that traditional BI use case. You’ll get the best performance out of a columnar database, by far. But you can get a lot of mileage out of Postgres, up to a certain size of data.

Whatever your DB, you’ll want dedicated models/tables to serve those timeseries visualizations or file dumps. Using “materialized views” can be a nice middle ground, to avoid building full-on data pipelines and a lot of change data capture issues.

If you’re doing a lot of data modeling, I’d highly recommend looking into dbt-core (the open source version).