GraphForge v0.4.0 completes the three-surface API: db.execute, db.gds, and db.search.

That is the release where GraphForge becomes less like a query engine alone and more like an embedded graph workbench for Python.

What Shipped

• db.gds graph algorithms backed by igraph or NetworkX
• PageRank, centrality, Louvain, connected components, clustering coefficient, and triangle count
• db.search with FTS5 text search, vector cosine similarity, and reciprocal-rank-fusion hybrid retrieval
• Bring-your-own vectors for text embeddings, geo coordinates, temporal features, or numeric arrays
• graphforge.recipes, including neighbourhood context helpers
• Bulk property write-back and improved export parameters
• Benchmark suite and integration tests for three-surface interoperability

Why We're Excited

This release expresses the larger GraphForge idea: graph work is not only storage and query. Analysts need query, algorithms, retrieval, recipes, and export to fit together in one local workflow.

Source

• GitHub release: GraphForge v0.4.0