Weaviate

Use it when

• Your semantic search corpus has outgrown an in-memory or embedded vector store and needs replication, horizontal scaling, and RBAC
• You want hybrid BM25 plus vector search with filtering and optional automatic vectorization at import time behind one query interface

What it solves

Not the fit when

• Running fully in-process without a server; requires a running Weaviate instance
• Relational or transactional workloads
• Integrated vectorizer modules may call external embedding APIs unless a local model container is configured
• using the local quickstart configuration unchanged for internet-facing production; configure authentication/RBAC, tenancy, persistence, backups, and network controls for the deployment
• document parsing and chunking
• llm model serving
• embedded in-process vector store with no server

Install

docker-compose.yml with image cr.weaviate.io/semitechnologies/weaviate:1.36.0 (ports 8080 and 50051), docker compose up -d; then pip install -U weaviate-client

Invoke

client = weaviate.connect_to_local(); client.collections.create(name=..., vector_config=Configure.Vectors.text2vec_model2vec()); collection.query.near_text(query='...', limit=1)

Alternatives

No reviewed alternatives recorded yet.