SDK overview

Every SDK implements the same QQL lifecycle: parse input, inspect or rewrite an AST, compile or explain offline, then execute against Qdrant. Choose the host that is already closest to your application; do not translate a query into a different syntax per service.

Host	Install	Live execution	Distinct capability
Rust	qql-core + qql	REST and optional gRPC	Direct AST, planning, and executor APIs
Python	pyqql	REST and optional gRPC	Native statement handles and async execution
Node.js	@veristamp/nqql	REST and optional gRPC	Native N-API bindings and parseJson
Browser/WASM	qql-wasm	REST/fetch	Offline analyze, JavaScript embedder, transferable bytes

QQLOne query, every SDKTry in playground↗

QUERY TEXT 'portable query'
FROM docs
USING dense
LIMIT 5;

Shared workflow

1. Parse or analyze the source before sending traffic.
2. Inject a trusted filter when the caller must be constrained.
3. Use compile or explain to inspect the Qdrant operation.
4. Execute a statement, script, or independent host array.

Capability	Rust	Python	Node.js	WASM
Parse	Parser::parse_all	parse	parse	parse
Validate	parse result	is_valid	isValid	isValid
Inject policy	inject_filter	inject_filter	injectFilter	inject_filter / Stmt.injectFilter
Explain and compile	planner/runtime	explain, compile_query	explain, compileQuery	analyze, compile
Execute	Executor::execute	Client.execute	Client.execute	Client.execute

A shard key is not a security boundary

SHARD chooses routing. Enforce tenant isolation with inject_filter or the equivalent statement method, and use a shard key only as an additional routing choice. The supported injected operators are =, >, >=, <, and <=.

For local in-process execution, the edge packages (pyqql-edge, @veristamp/nqql-edge, and the qql-edge crate) build a self-contained backend with embedded models. They are a separate feature surface from the remote SDKs; see the Edge overview before adopting them.

Choose a host page for complete API and lifecycle guidance.