Highlight abstraction layer over vector databases

[dandv]

I did a bit of advocacy for Mastra here, showcasing how it makes it far easier to evaluate multiple vector DBs before committing to one, by providing a consistent interface to upsetting and querying.

My comment links to https://mastra.ai/docs/rag/vector-databases. Could that page be even more effective at highlighting this consistent interface? For example, by

• mentioning this specifically
• using the same collection name, my-collection, unless the name matters for embeddings and my-index
• only using one example, with the vector DB initialization factored out
• using tabs for the vector DB (askin to npm/pnpm/yarn/...), so it's immediately visible how the actual usage is unchanged

[calcsam]

Thanks! Working on this in #1676

[calcsam]

Okay completed, closing

[dandv]

Nice! Perhaps the heading and description of the DB should be included in the tab? Otherwise it's Postgres for all: