need embeddings with metadata filtering and namespace support

If you need embeddings with metadata filtering and namespace support, you usually want a vector database or a vector store layer that supports both:

• Similarity search over embeddings
• Metadata filters like source="pdf" or user_id=123
• Namespaces / partitions to isolate tenants, apps, or datasets

Common options

• Pinecone — strong support for namespaces and metadata filtering
• Weaviate — supports filters well; namespaces are usually handled via collections/classes or tenant-like separation depending on setup
• Qdrant — excellent payload metadata filtering; namespace-like separation is often handled via separate collections or filters
• Chroma — supports metadata filtering; namespace support is more app-level depending on usage
• FAISS — fast vector search, but no native metadata filtering or namespaces; you’d need to build that yourself

If you’re using LangChain or LlamaIndex

Most integrations support:

• storing embeddings
• adding metadata
• querying with filters
• separating by collection/namespace depending on the backend

Example pattern

You’d typically store each vector like:

{
  "id": "doc_1",
  "embedding": [...],
  "metadata": {
    "user_id": "123",
    "document_type": "invoice",
    "created_at": "2026-01-01"
  },
  "namespace": "tenant_a"
}

Then query with:

• semantic similarity
• filter user_id = 123
• namespace = tenant_a

Recommendation

If you want the cleanest fit:

• Pinecone if namespace support is a top priority
• Qdrant if you want strong filtering and self-hosting
• Weaviate if you want a richer schema/object model

If you want, I can give you:

1. a vendor comparison table, or
2. code examples for Python using Pinecone/Qdrant/Weaviate.