RDF4J / SPARQL Endpoint

Waratah serves an RDF4J-protocol REST facade (waratah_api, axum) over an in-process compute node. In the developer cluster the base URL is http://127.0.0.1:8080/rdf4j/ and the single repository id is main.

There is no authentication in v1. There is also no CORS — proxy /rdf4j and /waratah onto the same origin as your client (the bundled console does exactly that), or use the same-origin /cors_proxy/{compute_id}/… route the cluster exposes.

Endpoints

Path Methods Purpose

Path	Methods	Purpose
`/rdf4j/protocol`	GET	RDF4J protocol version.
`/rdf4j/repositories`	GET	Repository list as SPARQL Results (`uri`, `id`, `title`, `readable`, `writable`) — one row for `main`. Exists for Workbench-compatible clients.
`/rdf4j/repositories/{id}`	GET, POST	SPARQL Protocol query endpoint. The query text arrives in the `query` parameter (query string on GET; form-encoded body on POST; or a raw `application/sparql-query` POST body); repeated `default-graph-uri` / `named-graph-uri` parameters override the query’s `FROM` / `FROM NAMED` dataset.
`/rdf4j/repositories/{id}/size`	GET	Repository quad count.
`/rdf4j/repositories/{id}/explain`	GET, POST	Plan explain — the HTTP twin of `waratah explain`; same formats.
`/rdf4j/repositories/{id}/namespaces`	GET	Namespace prefix listing.

/rdf4j/protocol

GET

RDF4J protocol version.

/rdf4j/repositories

GET

Repository list as SPARQL Results (uri, id, title, readable, writable) — one row for main. Exists for Workbench-compatible clients.

/rdf4j/repositories/{id}

GET, POST

SPARQL Protocol query endpoint. The query text arrives in the query parameter (query string on GET; form-encoded body on POST; or a raw application/sparql-query POST body); repeated default-graph-uri / named-graph-uri parameters override the query’s FROM / FROM NAMED dataset.

/rdf4j/repositories/{id}/size

GET

Repository quad count.

/rdf4j/repositories/{id}/explain

GET, POST

Plan explain — the HTTP twin of waratah explain; same formats.

/rdf4j/repositories/{id}/namespaces

GET

Namespace prefix listing.

Everything else in the RDF4J surface — /statements, /contexts, /rdf-graphs/…, /transactions, per-prefix namespace operations, and PUT/DELETE on the repository resource — answers 501 not implemented. SPARQL Update (INSERT DATA) does not exist in v1; data arrives via the ingest API or waratah import.

Example

curl -G 'http://127.0.0.1:8080/rdf4j/repositories/main' \
  --data-urlencode 'query=SELECT ?s ?p ?o WHERE { ?s ?p ?o } LIMIT 5' \
  -H 'Accept: text/csv'

Content negotiation

Content negotiation (waratah_api/src/serialize.rs) picks the first acceptable Accept entry — q-values are ignored. Wildcards map to a producible default (text/* → Turtle or CSV, application/* → RDF-XML or JSON). An Accept listing only unproducible media types is a 406.

Query form Formats

Query form	Formats
`SELECT` / `ASK` (results)	SPARQL Results JSON, XML, CSV, TSV
`CONSTRUCT` / `DESCRIBE` (graph)	Turtle, N-Triples, TriG, N-Quads, RDF-XML, JSON-LD

SELECT / ASK (results)

SPARQL Results JSON, XML, CSV, TSV

CONSTRUCT / DESCRIBE (graph)

Turtle, N-Triples, TriG, N-Quads, RDF-XML, JSON-LD

Defaults on the results endpoint (SELECT / ASK) for absent / * / */* Accept: CSV (text/ also maps to CSV, application/ to JSON). Graph responses (CONSTRUCT / DESCRIBE) default to Turtle for absent, */*, or text/* Accept, and application/ to RDF-XML — a bare * is *not matched for graphs, so an Accept of only that is a 406. Repository and namespace listings keep the JSON default.

Compression

Responses are compressed by a hand-rolled axum middleware wrapping the whole router: brotli at level 4, gzip at level 5. Accept-Encoding negotiation follows RFC 9110 q-values (x-gzip aliases gzip, * stands in for unlisted codings); equal-weight ties prefer brotli; a header rejecting every coding including identity short-circuits 406 before the handler runs. Compressed responses carry Content-Encoding and Vary: Accept-Encoding; streamed SELECT bodies compress chunk-by-chunk. Bodies already carrying Content-Encoding (CORS-proxy pass-through) are never recompressed, image/* (except SVG) and text/event-stream are skipped, and buffered bodies under 1025 bytes stay identity — which keeps ASK booleans uncompressed in every serialization.

Also served on the same listener

The pre-built grevillea console (static SPA) at / — GET/HEAD fallback after the /rdf4j, /waratah, and /cors_proxy nests. A missing console artifact embeds a stub and waratah start warns.
/waratah/* — admin surface: ingest and cluster snapshot / health.
/cors_proxy/{compute_id}/rdf4j|waratah/… — streams the request and response to that cluster member’s configured HTTP origin, so a browser UI can reach peer computes same-origin. Each compute’s GET /waratah/cluster row carries its cors_proxy base.
/jupyter/* — the optional notebook sidecar reverse proxy (REST
WebSocket kernel channels), when waratah start --with-notebooks (or waratah standalone --with-notebooks) is on.

Every live process also reports a semver+git-describe version string via its Describe RPC, surfaced on GET /waratah/cluster — see the Cluster API page.