For a hybrid search app with lots of filters, I’d generally recommend:

Best default choice: Postgres + pgvector

Use this if:

• your dataset is small to medium,
• you already use Postgres,
• you need lots of structured filters,
• you want simple ops and strong consistency.

Why it works well:

• SQL gives you excellent filtering: WHERE, joins, permissions, faceting, etc.
• pgvector handles embeddings in the same database.
• Hybrid search can be done by combining lexical search (Postgres full-text, trigram, or external BM25) with vector similarity.
• Great when filter logic is complex and constantly changing.

Tradeoff:

• Not the fastest for very large-scale ANN search compared to dedicated vector DBs.

---

Best dedicated vector DB for heavy filtering: Weaviate

Good if you want:

• native vector search,
• strong metadata filtering,
• hybrid search,
• a more “search product” feel.

Why:

• It supports hybrid search natively.
• Metadata filters are pretty solid.
• Easier than building your own search stack from separate components.

Tradeoff:

• Less flexible than SQL for deeply relational/filter-heavy applications.

---

Best if you want strong filter performance at scale: Qdrant

Good if:

• you need fast vector search,
• you have lots of metadata filters,
• you want a straightforward API,
• you don’t need full text search to be the main engine.

Why:

• Very good payload filtering.
• Efficient and simple to operate.
• Often a strong fit for “vector + metadata filter” workloads.

Tradeoff:

• For true hybrid search, you may still pair it with a text search engine or implement hybrid ranking in your app.

---

Best for enterprise search at scale: Elasticsearch / OpenSearch

Choose this if:

• hybrid search is core,
• you already want inverted-index text search,
• you have lots of filters and aggregations,
• you need faceting and relevance tuning.

Why:

• Excellent text search + filters + aggregations.
• Vector search is supported now.
• Hybrid search is a natural fit.

Tradeoff:

• More operational complexity.
• Vector capabilities are good, but not always as elegant as a dedicated vector DB.

---

Quick recommendation by scenario

Pick Postgres + pgvector if:

• filters are very important,
• data size is moderate,
• you want simplest architecture.

Pick Qdrant if:

• vector search is primary,
• filters are heavy,
• you want good performance and simplicity.

Pick Weaviate if:

• you want built-in hybrid search with metadata filtering.

Pick Elasticsearch/OpenSearch if:

• search relevance, text search, filters, and faceting matter most.

---

My practical advice

If you said “hybrid search app with lots of filters” and didn’t mention massive scale, I’d start with:

Postgres + pgvector
or
Qdrant + a text search engine

If you want a single system and lots of filters, Weaviate or Elasticsearch/OpenSearch are often better than pure vector DBs.

If you want, I can give you a specific recommendation based on your stack (data size, QPS, cloud/on-prem, need for faceting, etc.).