This is a project by h4ck1ng.science Support it → This is a project by h4ck1ng.science Support it → This is a project by h4ck1ng.science Support it →
rete
Docs

Open source · RDF · SPARQL

Your RDF graph in one file. Host it anywhere. Query it everywhere.

Query it from a browser, Python, JavaScript, R, Java, the command line or Claude

rete packs your triples and the indexes that find them into a single .rete file. Put it on object storage, GitHub Pages, Zenodo or any web server. Each SPARQL query reads only the parts of the file it needs, so there is no database to run.

Try it

Query a published graph

Choose a graph and a SPARQL query. Your browser reads the file straight from the bucket and answers it, with the same engine as the command line.

Graph
Question
byte 0

Each mark on the bar is a part of the file this page fetched.

How it works

Build once, host it, query from anywhere

1

Build

Turn Turtle, N-Triples, N-Quads, TriG, RDF/XML, JSON-LD or Jelly into one .rete file. It holds the triples, the indexes that find them and a Dataset Card that says what is inside.

The parts are laid out so a reader can find its way from the front of the file alone. The figure is a real file from the demo, drawn to scale.

rete build data.ttl -o data.rete --card
2

Host

Upload the file anywhere that serves byte ranges: object storage, GitHub Pages, Zenodo, your own web server. There is no database to run or keep up.

The engine runs where you query, in the browser tab, the notebook or the command line. The host only hands out the bytes it is asked for.

curl -r 0-1023 https://…/data.rete   # 206 Partial Content
3

Query

A client opens the file with one read of its front, then fetches only the dictionary chunks and index tiles the query touches. Pick a graph and a question: the marks are the ranges this page really fetched for it, set against a plain download on the same link.

Selective questions stay small. A question that scans the whole graph still has to read the whole graph.

rete sparql-url https://…/data.rete "SELECT …"

Any size

The first read is the same whether the file is a few megabytes or a hundred gigabytes, and a lookup grows with what it finds, not with the file.

The biggest file in the catalog, the OpenCitations Index, holds 15.8 billion statements in 111 GB. Asking who cites one paper reads about 17 MB of it.

Use it

The same file, from wherever you work

Every client below asks the jazz graph from the demo the same question.

# the CLI ships as a Docker image; see Install for the alias
rete sparql-url https://data.graphplaza.com/linked-jazz/linked-jazz-2026-06-13.rete \
  "PREFIX foaf: <http://xmlns.com/foaf/0.1/>
   SELECT ?name WHERE { ?p foaf:name ?name } LIMIT 5"

Built on it

Apps that read .rete files straight from a bucket

Support

Help keep rete free and open

rete is a project by h4ck1ng.science. The engine, the file format and every client are open source under the Apache 2.0 license, and the public graphs on this page are free for anyone to query. Support keeps it that way.

Where support goes

Hosting the graphs The catalog at data.graphplaza.com serves dozens of graphs, some over 100 GB, to anyone who asks for a byte range.
Building new graphs Turning a large open dataset into a .rete file can take days of compute and a lot of memory.
Time Work on the engine, the file format, the clients and these docs.

Ways to help

Chip in on Ko-fi A one-off or a monthly contribution, in a couple of clicks. ☕ Support on Ko-fi ↗
Sponsor as an organization If rete is useful to your lab, library or company and you would like to sponsor it, get in touch through h4ck1ng.science. h4ck1ng.science ↗
Help without money Star the repository, cite rete in your papers, report a bug, or publish a graph of your own as a .rete file. Star on GitHub → Cite (Zenodo DOI) → Report an issue →