← Back to Blog

Traverse 0.7: A Graph Data Science Library, and Traverse in the Browser

Traverse 0.7 is out, and two big things land with it: a built-in graph data science library - more than 36 algorithms you can call straight from Cypher - and a browser build that runs the entire engine in a tab, with no server at all. You can try the browser version right now at traverse.truespar.com.

Neither one needs a plugin or an extra service, and there are no breaking changes: existing queries, drivers, and SDK calls keep working exactly as before.

Graph data science, from Cypher

Traverse now ships a graph data science library with more than 36 algorithms, each callable from Cypher like any other procedure - for example CALL traverse.pageRank.stream(...). Every algorithm supports stream, stats, and write modes where they apply, and write mode saves the result straight back onto your nodes as a property, so you can query it like any other data.

  • Centrality: PageRank, Article Rank, Eigenvector, Closeness, Betweenness, HITS, Degree, CELF
  • Community detection: Louvain, Leiden, Label Propagation, Weakly and Strongly Connected Components, K-Core, Triangle Count, Modularity
  • Paths and traversal: Dijkstra, A*, Bellman-Ford, Yen's K-Shortest, All-Pairs Shortest Paths, Delta-Stepping, BFS, DFS, Random Walk, Spanning Tree, Steiner Tree
  • Similarity: Node Similarity (Jaccard), Cosine Similarity, and kNN (structural and feature-vector)
  • Link prediction: Common Neighbors, Adamic-Adar, Resource Allocation, Preferential Attachment, and more
  • Embeddings: FastRP, Node2Vec, HashGNN, GraphSAGE

Studio has a GDS drawer that builds the config for you with a form, previews results live on the graph (node size, colour gradients, community colours), and shows result tables for path, node-pair, and embedding outputs. Full details are in the graph algorithms reference.

Traverse in the browser

The full engine now compiles to WebAssembly and runs entirely in a browser tab - no server, no install. It's the same engine and the same APIs as the server build: Cypher, the graph data science library, the schema browser, and the query editor are all identical.

  • Local persistence - databases are stored in the browser's Origin Private File System and survive tab reloads.
  • Multi-threaded - it uses the browser threads available to it.
  • Sample databases - Movies, OurAirports, OpenFlights, and Pokec download straight in, each with starter queries to get going.
  • Shareable queries - a share link carries the full query and the active database, with no server round-trip.

Try it at traverse.truespar.com, or embed the engine in your own app with the @truespar/traverse-wasm npm package. The embedding guide covers installation, hosting, and TypeScript types.

Studio

Studio got a broad pass this release:

  • A mobile-responsive layout - a bottom tab bar, slide-over side panels, and a full-screen AI chat - plus Add to Home Screen support, so on iOS it launches fullscreen like a native app.
  • First visit follows your system light or dark preference; your explicit choice sticks from then on.
  • The AI chat now knows the graph algorithm catalog, so it suggests correct CALL syntax with real algorithm names and modes instead of guessing at them.
  • Click the query-time chip in the result panel to see a parse, plan, execute, and round-trip breakdown.

Performance and fixes

A round of algorithm and query speedups came with this release:

  • Weakly Connected Components is about 44% faster.
  • Random Walk is roughly 3.7x faster.
  • kNN-Property is about 5.3x faster, with its nearest-neighbour descent parallelized across cores.
  • Label Propagation converges in fewer iterations.

On top of that: planner fixes for OPTIONAL MATCH and multi-MATCH patterns with pre-bound endpoints, a timezone fix in datetime.truncate for non-current dates, and several browser-mode rendering fixes for node, edge, and path values in result tables.

Upgrade

0.7 is a drop-in upgrade with no breaking changes - existing Cypher queries, Bolt connections, embedded SDK calls, and HTTP requests all keep working. Download it from the Traverse product page or follow the getting started guide. And if you just want to try it, the browser build at traverse.truespar.com needs nothing at all.