I'm building a semantic search app and need a vector database recommendation

Here’s a practical way to choose a vector database for a semantic search app.

Best default recommendation

Start with PostgreSQL + pgvector if:

• your dataset is small to medium
• you already use Postgres
• you want the simplest architecture
• you need metadata filtering, transactions, and SQL joins

It’s usually the fastest path to a working semantic search app.

When to choose a dedicated vector DB

Use a dedicated vector database if you expect:

• millions to billions of embeddings
• high query throughput
• low-latency similarity search
• heavy hybrid search needs
• distributed scaling and replication without managing it yourself

Good options:

• Pinecone — very easy managed service, strong production choice
• Weaviate — feature-rich, good for hybrid search and schema support
• Qdrant — strong open-source option, good filtering, efficient, production-ready
• Milvus — strong for large-scale workloads, more infrastructure-heavy

Quick recommendation by scenario

1) Fastest MVP

pgvector

• easiest if you already have Postgres
• great for prototype-to-production for many apps

2) Managed, minimal ops

Pinecone

• easiest dedicated vector DB to run in production
• good if you want to avoid infrastructure work

3) Open-source + strong filtering

Qdrant

• excellent metadata filtering
• lightweight and practical for many teams

4) Larger-scale / advanced search features

Weaviate or Milvus

• better if you need more advanced search patterns or scale

My opinionated pick

If you didn’t give any extra constraints, I’d recommend:

• pgvector if you want simplicity and your scale is not huge
• Qdrant if you want a dedicated vector DB with a good balance of ease, filtering, and performance
• Pinecone if you want the least operational overhead

Questions that would let me recommend one precisely

1. How many embeddings do you expect?
2. Do you need metadata filters?
3. Are you already using Postgres?
4. Do you want managed or self-hosted?
5. What latency/QPS target do you have?

If you want, I can give you a direct recommendation between pgvector, Pinecone, Qdrant, Weaviate, and Milvus based on your exact use case.