Can you recommend a vector database for a startup RAG app that needs keyword search, metadata filters, and low ops overhead?

For a startup RAG app that needs vector search + keyword search + metadata filtering + low ops overhead, my short recommendation is:

Best all-around choice: Qdrant Cloud

Why it fits:

• Hybrid retrieval support: vector search plus payload/metadata filtering; good fit for RAG
• Low ops overhead: managed cloud option, easy to run self-hosted later if needed
• Strong filtering: one of Qdrant’s biggest strengths
• Fast enough for startup scale: solid performance without a lot of tuning
• API-friendly: simple to integrate into app stacks

If keyword search is a first-class requirement

Pure vector DBs vary on keyword search quality. If you need true lexical keyword search alongside vectors, consider:

Weaviate Cloud

• Supports hybrid search well
• Strong developer experience
• Managed hosting available
• Good choice if you want a more “search engine + vector DB” feel

Postgres + pgvector + full-text search

• Great if you want lowest operational complexity and already use Postgres
• You get:
  • vector search via pgvector
  • keyword search via Postgres full-text search
  • metadata filters via SQL
• Best when dataset size is moderate and you prefer simplicity over specialized infra

My practical ranking for your use case

1. Qdrant Cloud — best balance of RAG, filtering, and ops simplicity
2. Weaviate Cloud — strong hybrid search, nice managed experience
3. Postgres + pgvector — simplest stack if scale is modest and you want everything in one DB

Quick decision guide

• Choose Qdrant if: you want strong metadata filtering and a clean vector-first RAG setup
• Choose Weaviate if: you want more built-in hybrid/keyword-style retrieval features
• Choose Postgres/pgvector if: you want minimal moving parts and your corpus is not huge

If you want, I can also give you:

• a feature-by-feature comparison table
• a cost/scale recommendation
• or a “best choice based on your expected document count + QPS” recommendation.