Module algo 

Graph algorithms operating on an in-memory CSR snapshot.

These algorithms (PageRank, connected components, shortest path, similarity) are pure-Rust, zero-dependency, and backend-agnostic: they consume a CsrGraph built from the edge set rather than running SQL per step. This keeps the iterative math identical across the SQLite and PostgreSQL backends — the same zero-drift principle as recall.rs’s compute_recency.

Available on the graph feature with no extra feature gate (the algorithms add no dependencies).

Re-exports

pub use community::LABEL_PROPAGATION_ITERS;

pub use community::connected_components;

pub use community::label_propagation;

pub use community::label_propagation_default;

pub use csr::CsrGraph;

pub use pagerank::PAGERANK_DAMPING;

pub use pagerank::PAGERANK_EPS;

pub use pagerank::PAGERANK_ITERS;

pub use pagerank::pagerank;

pub use pagerank::pagerank_default;

pub use pagerank::pagerank_scores;

pub use path::SHORTEST_PATH_W_MIN;

pub use path::dijkstra;

pub use path::shortest_path;

pub use path::shortest_path_ids;

pub use similarity::adamic_adar;

pub use similarity::common_neighbors;

pub use similarity::jaccard_similarity;

pub use similarity::link_prediction;

Modules

community

Community detection: connected components and label propagation.

csr

Compressed Sparse Row (CSR) representation of the knowledge graph.

pagerank

Weighted PageRank centrality.

path

Weighted shortest path via Dijkstra.

similarity

Node similarity and link prediction.