Skip to content

Features

One statement covers every Qdrant retrieval mode — dense, hybrid, recommendation, context, discover, sort, and random sample. No separate APIs per mode.

QUERY 'emergency care' FROM docs LIMIT 5 USING HYBRID

Build prefetch DAGs declaratively with WITH clauses. Each stage can have its own filter, limit, score threshold, and named vector — composed into a single fused result.

WITH dense AS (QUERY 'care' USING dense LIMIT 200),
sparse AS (QUERY 'care' USING sparse LIMIT 300)
QUERY 'care' FROM docs LIMIT 10 PREFETCH (dense, sparse) FUSION RRF

Full expression algebra executed directly inside Qdrant — arithmetic, math functions, geo-distance, decay functions, datetime expressions, and CASE WHEN conditionals. No client-side post-processing.

Store and search with token-level vector representations for late interaction models. Define multiple named vectors with multivector comparators in a single CREATE COLLECTION statement.

Start a Connect RPC server (gRPC + gRPC-Web + HTTP/1.1) with JWT authentication, YAML policy rules, AST-level tenant filter injection, rate limiting, query templates, and structured audit logging.

Embed QQL execution directly in Go applications without the CLI.

result, _ := qql.Exec(ctx, client, "QUERY 'search' FROM docs LIMIT 5")

Connect RPC clients for every language. Call the gateway from Python, TypeScript, Go, or curl.

Migrate from Qdrant REST API or Python SDK to QQL in seconds with the built-in converter and request interceptor.

Full DDL surface: CREATE, ALTER, DROP, SHOW, and index management with HNSW and quantization configuration.

Structured JSON output mode for pipelines and agents. Script execution, collection dumps, and explain plans for deterministic, repeatable operations.