Source

How a remote graph is queried lazily

Remote datasets (e.g. wikidata-1GB, ohm-full) are never downloaded whole — each query fetches only the dictionary chunks and index tiles it touches.

The self-describing card

Each .rete embeds a small card: curated metadata (title, license, source) plus auto-derived stats — datatype/language histograms, the effective schema (class→predicate→class), top hubs, geo/temporal extents, and a schema pyramid with per-semantic-zoom-level type rollups. A remote reader fetches just the header + card in two HTTP ranges (~10 KB) to learn a multi-GB graph's shape and legend without ever touching the triple index.

Query helpers: label decode & entity finder

Two aids sit by the editor and never change what a query does — they only read and write the query text:

🏷 Labels floats a human-readable name over each IRI: wd:Q100001 reads as Bemelen, wdt:P625 as coordinate location. Names come from a built-in vocabulary, predefined per-dataset hints, and a live rdfs:label / skos:prefLabel lookup.
🔎 Find an entity turns a name you type into its IRI through rete's bounded label index (prefix_search — a binary search, never a full scan); click a hit to drop its <IRI> in at the caret.

Both are lazy-mode native: on a remote .rete they fetch only the few label tiles they need over HTTP range — a small spins while those bytes arrive.

Selected Example

Progressive Metadata

Provenance

Schema Summary

SPARQL Console

Default path for exact queries, summary-safe progressive queries, and graph rendering.

Sources

Output Strategy 🏷 Labels

Explore

Browse the dataset as entity tables, the community pyramid as nested clusters, and the file's byte layout.

One table per rdf:type class: entities as rows, their properties as columns.

Write your own SQL over this dataset's companion tables — quote the names, e.g. SELECT * FROM "Class_" (same table names across all three engines). Pick an engine; cached files (see Settings) are used automatically, otherwise it reads lazily over HTTP range.

Engine

The pyramid clusters the graph by Louvain modularity over several coarsening rounds (a dendrogram); the build ships one round whose tiles fit a ~64 KiB budget. That gives a semantic zoom from coarse communities down to finer ones — and lets the progressive query strategy answer aggregate questions exactly from this summary in a few small range reads, never touching the triple index.

File byte map

Cell size

SHACL

Validate the loaded graph against Turtle shapes without leaving the static page.

Reach

Run transitive closure over one predicate for impact analysis and lineage checks.

Predicate Seeds Reverse

Schema

Ontology-aware class and relation summary from the loaded graph.

Classes

Relations

Provenance

Explain triple-pattern matches with selected index permutation and byte ranges.

Subject Predicate Object

Coherence

Check the loaded graph for OWL/RDFS incoherent points — schema-level (Tier-0, index-free) and instance-level (full reasoner) — entirely in your browser.

Create a dataset

Assemble a brand-new .rete from your own RDF — add an ontology, write the dataset card, attach example SPARQL & SHACL — and keep it in this browser (IndexedDB) right beside the bundled ones. Nothing leaves your machine until you export it.

RDF data

Format

Paste your graph, or upload a file. Compatible formats: N-Triples (.nt), N-Quads (.nq), Turtle (.ttl), JSON-LD (.jsonld / .json) and RDF/XML (.rdf / .owl). Uploading sets the Format from the file's extension; JSON-LD and RDF/XML are converted in your browser before building.

Upload file .nt · .nq · .ttl · .jsonld · .rdf · .owl

Ontology / schema

optional

Class/property definitions (RDFS / OWL / SHACL vocabulary) merged into the same graph, so the Schema, Reasoning and SHACL tabs light up. Use the same format you picked above — JSON-LD and RDF/XML (incl. .owl) are converted in the browser before building.

Upload ontology file

Dataset card

in sync

Title Key url-safe id Icon License Source URL Tags comma-separated Description Provenance how it was built (optional)

Card JSON edit either side — they sync both ways Import JSON

Example queries

optional

Saved with the dataset and shown in its Query Library / SHACL tab — the same shape the bundled datasets use.

…

SPARQL Console

Explore

File byte map

SHACL

Reach

Schema

Classes

Relations

Provenance

Coherence

Create a dataset