The language pages explain individual constructs. These guides combine them into production retrieval patterns.
Hybrid retrieval
Build dense and sparse candidate stages, then fuse them with RRF or DBSF.
Embeddings and reranking
Resolve vector roles and choose late-interaction or cross-encoder reranking.
Multitenancy
Keep isolation, shard routing, index layout, and partition lifecycle distinct.
Backends
Choose between Qdrant REST, gRPC, and in-process edge execution.
Filter injection
Enforce tenant and authorization predicates at the AST level with inject_filter.