QQL runnable examples

The repository examples are executable applications, not duplicated snippets inside the documentation site.

Example	Language / backend	What it demonstrates
`examples/python`	Python (`pyqql`)	Python basics through advanced retrieval: `basic_to_medium.py` (parse → `is_valid` → `explain` → `compile_query` → `inject_filter` → shard key) and `medium_to_expert.py`. Offline — no Qdrant needed.
`examples/nodejs`	Node.js (`@veristamp/nqql`)	Node.js parser and runtime APIs: `basic_to_medium.mjs` and `medium_to_expert.mjs`.
`examples/rust`	Rust (`qql-core`)	`basic_to_medium` (`Parser::parse` + `inject_filter` + `set_shard_key`) and `medium_to_expert` (fail-closed multi-tenant gateway with a per-user `tenant_id` plus a `status` predicate).
`examples/wasm`	WebAssembly (`qql-wasm`)	WASM `parse` / `isValid` / `analyze` / `compile` / `explain`, `Stmt.injectFilter` + `shardKey`, hybrid analysis, `Client` execute (multi-tenant gateway).
`examples/edge-demo`	qql CLI `--edge`	In-process edge execution with the qql CLI. Requires an edge-enabled build (`--features edge`).
`examples/airbnb-demo`	Python (`pyqql`) + Qdrant + Ollama	Berlin geo search: `GEO_RADIUS` / `GEO_BBOX` / `GEO_POLYGON`, hybrid search, turbo binary quantization with rescore.
`examples/medical-retrieval-ops`	qql CLI + Qdrant	Provisions the ChatMED RAGCare-QA benchmark and runs retrieval checks.
`examples/medical-showcase`	qql CLI + Qdrant	12-record end-to-end: hybrid, filters, ACORN, timeout/consistency, grouped retrieval, recommend, context, discover, CTE prefetch DAGs, mutations, `DELETE PAYLOAD`, `SCROLL WITH VECTOR`, `COUNT` exact.
`examples/sec10k-qql`	Python (`pyqql`) + Qdrant + Ollama/LM Studio	Flagship multi-tenant hybrid RAG over SEC 10-K filings: `CREATE SHARD KEY`, `is_tenant` index, `SHARD` + `WHERE tenant_id`, fail-closed `inject_filter`, agent tool-use.

Browse the examples directory on GitHub.

QQLRepresentative geo queryTry in playground

QUERY 'quiet apartment near transit'
FROM listings
USING dense
WHERE location GEO_RADIUS {center: {lat: 52.52, lon: 13.405}, radius: 5000.0}
WITH PAYLOAD INCLUDE (name, neighbourhood, price)
LIMIT 10;

Qdrant 1.19 / QQL 1.5 surface

These statements require a Qdrant 1.19+ server for full semantics. Prefix and slice filters, IDF params, and memory placement parse on every backend; quotas are REST-only.

QQLPrefix filter, slice, and sparse IDFTry in playground

CREATE INDEX ON COLLECTION docs FOR title TYPE keyword WITH (prefix = true);

QUERY 'search'
FROM docs
WHERE title MATCH PREFIX 'Comp' AND SLICE (4, 1)
LIMIT 10;

QUERY 'search'
FROM docs
USING sparse
PARAMS (idf = 'global')
LIMIT 10;

QUERY 'search'
FROM docs
USING sparse
WHERE tenant_id = 'acme'
SHARD 'acme'
PARAMS (idf = WHERE tenant_id = 'acme')
LIMIT 10;

QQLMemory placement and cluster quotas (REST)Try in playground

CREATE COLLECTION docs (
  dense VECTOR(384, COSINE) WITH VECTOR (memory = 'cached', datatype = 'turbo4')
)
WITH HNSW (memory = 'cold')
WITH PARAMS (payload_memory = 'cold');

SHOW QUOTAS;

SET QUOTA (enabled = true, max_resident_memory_percent = 80) WAIT true;

When adding an example, keep its QQL compatible with language/v1, document the required backend and models, and provide a deterministic setup path.