Execution
The playground requires JavaScript and WebAssembly.
Plan
Type a statement or pick an example. Query planning runs offline inside WebAssembly.
/collections/…
Loading compiler…
Executable corpus
Every example loads from language/v1/fixtures/valid.
language/v1/fixtures/valid
Live collections
Reading collections from your Qdrant…
41 examples available
ALTER COLLECTION docs WITH VECTOR (on_disk = true);
BATCH {
CLEAR PAYLOAD FROM docs WHERE status = 'archived';
COUNT FROM docs;
CREATE COLLECTION docs (
CREATE COLLECTION docs USING DENSE MODEL 'all-MiniLM-L6-v2';
CREATE INDEX ON COLLECTION docs FOR status TYPE keyword;
CREATE SHARD KEY 'acme' ON COLLECTION docs WITH (shards_number = 2);
DELETE PAYLOAD draft FROM docs WHERE status = 'archived';
DELETE VECTOR colbert FROM docs WHERE id = 42;
DELETE FROM docs WHERE id = 42;
DROP COLLECTION docs;
DROP INDEX ON COLLECTION docs FOR status;
UPSERT INTO docs VALUES {id: 1, text: 'hello', title: 'world'} USING DENSE MODEL 'nomic' ON FIELD title INTO title_vec;
CREATE COLLECTION products (dense VECTOR(512, COSINE));
FACET category FROM documents;
QUERY 'test' FROM docs WHERE field = 'value' LIMIT 10;
QUERY CONTEXT (POSITIVE POINT 1 NEGATIVE POINT 2) FROM docs LIMIT 10;
WITH
QUERY FORMULA $score / views [DEFAULT = 1.0] * 10.0 DEFAULTS (score = 0.0)
QUERY FORMULA $score * 2.0 DEFAULTS (score = 0.0) FROM docs LIMIT 10;
QUERY HYBRID TEXT 'vector database' DENSE dense SPARSE bm25 FUSION RRF
QUERY 'supply chain'
QUERY MMR TEXT 'vector database' DIVERSITY 0.7 CANDIDATES 100
QUERY 'supply chain risks'
QUERY 'vector database' FROM docs LIMIT 10;
QUERY ORDER BY price ASC FROM products LIMIT 10;
QUERY POINTS (1) FROM docs;
QUERY TEXT 'search' MODEL 'e5'
QUERY RECOMMEND POSITIVE (1, 2, 3) FROM products LIMIT 10;
QUERY RELEVANCE FEEDBACK TARGET 'search' FEEDBACK ((VECTOR [0.1, 0.2], 1.0), (VECTOR [0.3, 0.4], -1.0)) STRATEGY NAIVE (a = 1.0, b = 0.5, c = 0.2)
QUERY SAMPLE RANDOM FROM docs LIMIT 10;
QUERY 'hello' FROM docs LIMIT 10;
SCROLL FROM docs LIMIT 50;
SHOW QUOTAS;
SHOW COLLECTIONS;
UPSERT INTO docs VALUES {id: 1, text: 'QUERY $QUERY_TEXT FROM docs USING dense LIMIT $LIMIT;'};
UPDATE docs SET PAYLOAD = {status: 'reviewed'} WHERE id = 42;
UPSERT INTO docs VALUES {id: 1, title: 'Hello', body: 'World'};
Configure connection endpoints and vector inference. Credentials remain strictly local in your browser.
The weights download from the Hugging Face CDN on first use (tens of MB) and stay cached in this browser. The collection's vector size must match the model.
Browser model loads only when execution needs text embeddings.
Host inject
Applied after parse, before plan. The editor source stays as the caller wrote it.
Start from an example
tenant_id = acme
workspace_id = ws_101
deleted = false
region = eu-west-1
Host integration
Use the same validated QQL source from Rust, Python, Node.js, or cURL.
Generate a valid plan before exporting.
No matching action.
Reference