Canonical graph schema, provenance records, queries, and durable snapshots for the repository model.

Derives a tree of cards from the graph's own containment - repository, module, capability or package, type, member - so retrieval can descend structure rather than rank a flat list. Deterministic, fingerprinted against the graph it came from, ...

Deterministic Java source analysis with Eclipse JDT, producing a provenance-bearing code graph.

BM25 and optional dense retrieval over the symbol vocabulary, index-tree descent and the hybrids of those, query planning, token budgets, and citation verification for answers.

Modules, centrality, communities, workflows, capabilities, and explainable change-risk scoring over the code graph.

Interprets Java facts as Spring operational structure: bean wiring, endpoints, guards, queries, and clients.

Grounded question format, scoring for recall, precision, ranking quality, evidence and groundedness, plus a text-search baseline and an oracle chooser that measures the ceiling a tree descent could reach.

Renders the code graph as one self-contained HTML file, and exports GraphML, DOT, and Cytoscape JSON.