Weaviate Advanced Client Setup (v4 Python) Collection Creation with Modules Native Hybrid Search (BM25 + Vector) Weaviate computes BM25 and vector search server-side and fuses with a weighted convex combination or Relative Score Fusion. Reranking (module-based) Pairing: broad hybrid (limit=50) → reranker trims to the top-K that actually answers the query. Generative Search (RAG in one call) Multi-Tenancy Weaviate implements first-class tenants: each tenant is a physical shard and can be hot / cold / frozen. Compression Tradeoffs | Option | Memory reduction | Speed | Recall impact | |---|---|-…