JGF 2.0

JGF 2.0 (JSON Graph Format, version 2) is a lightweight, open specification for encoding graph structure in plain JSON. It is maintained by the "jsongraph" organization on GitHub, published under json-graph-specification with a companion site at jsongraphformat.info, and validated against JSON Schema. The format registers the media type application/vnd.jgf+json. Its stated goal is to capture basic graph structure in a convenient, human-readable form while leaving room for arbitrary application data through metadata.

A JGF document is a single top-level JSON object that contains either one graph object or a graphs array of them, so several graphs can travel in one file. A graph object carries a directed flag (defaulting to true), optional type, label, and metadata, plus a nodes container and an edges array. The defining change from version 1 to version 2 is that nodes are no longer an array but a map (JSON object) keyed by node id; each node value holds an optional label and a free-form metadata object. Using the id as the object key guarantees node-id uniqueness structurally. Each edge is an object with source and target (referencing node keys) plus optional relation, directed, label, id, and metadata. A later revision (early 2021) added hyperedges for undirected node sets and directed many-to-many relations. The metadata escape hatch is where layout, styling, weights, and algorithm results are conventionally stored.

Within the format-conversion landscape, JGF occupies the same niche as other JSON graph encodings such as format:jgf, format:jgf-1.0, format:graphjson, format:gjgf-0.1, format:networkx-node-link, and format:cytoscape-json: a generic, tool-agnostic interchange for node/edge data, contrasting with heavier XML formats like format:graphml or format:gexf-1.3. Its schema-driven design has spawned domain child schemas, notably the BEL JSON Graph Format for biological expression data.

Strengths: minimal, readable, JSON-native, schema-validated, multi-graph and metadata friendly, easy to parse in any language. Tooling includes the Node.js library tool:jay-gee-eff and the json-graph-specification npm package; the format is also consumed by visualization libraries such as tool:gravis. Limitations: it standardizes structure but not semantics, so metadata conventions vary between producers; the keyed-node map complicates ordering and streaming of very large graphs; and adoption remains narrower than GraphML or GEXF, with tooling comparatively sparse.