Jelly
The high-performance streaming protocol for knowledge graph systems
Jelly is an open-source, high-performance binary serialization format and streaming protocol for RDF knowledge graphs. Jelly accelerates knowledge graph applications where high throughput or low latencies are needed.
Blazing-fast by design
Text-based formats like Turtle and JSON-LD struggle with high-throughput streams and large-scale knowledge graphs due to heavy string parsing and repeated IRIs.
Jelly replaces this text overhead with a compact, Protobuf-based binary encoding designed specifically to be as efficient as possible.
✔ Cut transmission times
In real-world deployments (such as the award-nominated Nanopublication Network case study), Jelly reduced the transmission time of customer data from hours to under 4 seconds.
✔ Ultra-fast deserialization
Jelly outperforms traditional text formats and binary alternatives, processing millions of triples per second with minimal CPU and memory footprints.
✔ On-the-fly stream compression
Uses lookup-table and delta compression without requiring prior knowledge of the dataset, scaling effortlessly from edge IoT devices up to distributed server clusters.
Compatible with your tech stack
Jelly integrates seamlessly into your existing data infrastructure and language ecosystem. Whether you are running microservices, feeding stream pipelines, or populating graph databases, Jelly supports the tools you already rely on:
✔ Language support
✔ Knowledge graph frameworks
Built-in adapters for Apache Jena, RDF4J, RDFLib, Titanium, and Neo4j.
✔ Streaming infrastructure
Out-of-the-box support for streaming over gRPC, Apache Kafka, MQTT, and reactive streams
✔ Powerful command-line tool
jelly-cli will help you convert, inspect, and validate RDF streams directly in CI/CD pipelines or batch scripts with a single command.
Quick start
To install jelly-cli, add Jelly to your project dependencies, or view integration guides for Java, Scala, and Python, view the full documentation on GitHub:
Our commitment to open source
Jelly is 100% open-source, licensed under Apache 2.0.
At NeverBlink, we build reliable, neurosymbolic AI solutions that convert fragmented enterprise data into trusted, actionable knowledge graphs. Streaming knowledge graph systems require fast, standardized communication layers to operate in real time.
Jelly was developed to resolve inter-service communication and on-disk persistence bottlenecks of knowledge graph applications. We maintain Jelly as a 100% open-source specification and suite of libraries to empower the knowledge graph, AI, and data engineering communities.
Read more
Read our research papers detailing the architecture, benchmarks, and the motivation behind this project: