Graph tools¶

Graph operations¶

Graph algorithms operate on caller-owned topology. Start with scoped IDs and authorized adjacency, then select the smallest operation that answers the question. Use traversal for reachability, ranking for relative importance, and evidence projection to return source-backed context.

For change maintenance, share DependencyKey identities with dependency-aware updates. A graph walk finds potentially affected outputs. The planner uses input fingerprints and completed receipts to decide which outputs need work. Temporal filters and access checks remain explicit inputs to each operation.

Operation	Structure
Resolve entities	Field contributions plus link/review thresholds
Query temporal facts	Valid-time and transaction-time intervals
Expand retrieval	Authorized Personalized PageRank and passage projection
Aggregate a corpus	Connected communities and bounded map/reduce reports
Rebuild derived views	Dependency planning plus ordered event replay
Validate semantic records	Versioned concepts, properties, relations, and violations
Inspect arbitrary topology	Traversal, paths, reachability, components, and cycles
Select evidence	Bounded seed expansion and prize-guided connected subgraphs
Propose links	Common-neighbor, Jaccard, and Adamic–Adar scores
Rank structure	Degree, closeness, betweenness, HITS, and Personalized PageRank
Compare graphs	Explicit node/edge additions, removals, and structural drift
Diagnose quality	Orphans, dangling edges, duplicates, self-loops, and density
Construct candidates	Caller-defined blocking, pair scoring, and clustering
Move between libraries	Loss-visible NetworkX, GraphML, and JSON-LD projections

Example graph flow

Starting object	Graph operation	Returned structure
Query-linked entity	Authorized PageRank	Ranked nodes with propagation scores
Ranked nodes	Passage projection	Evidence passages with source-node trace
Temporal assertion	Bitemporal query	Facts valid and known at requested times
Event log	Projection replay	Derived state and stable build identity