Vector Store

Enterprise-grade vector database for document embeddings.

Technology

GateFlow uses PostgreSQL with pgvector extension:

• HNSW indexing for fast similarity search
• Row-level security for multi-tenancy
• ACID transactions
• Point-in-time recovery

Architecture

Collections

Organize vectors into collections:

curl -X POST https://api.gateflow.ai/v1/data/collections \
  -H "Authorization: Bearer gw_prod_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "legal-documents",
    "embedding_model": "text-embedding-3-small",
    "dimensions": 1536
  }'

Indexing

HNSW Configuration

{
  "index": {
    "type": "hnsw",
    "m": 16,
    "ef_construction": 64
  }
}

Search Performance

Vectors	Search Time (p99)
100K	<10ms
1M	<50ms
10M	<100ms

Multi-Tenancy

Automatic isolation via row-level security:

-- Automatically applied
SELECT * FROM vectors
WHERE organization_id = current_org_id()

Next Steps

• Semantic Search - Query vectors
• Data Isolation - Security model