Core concepts

Semantic intelligence

Map repository vocabulary, feature ownership, and developer intent to stable graph evidence — without embeddings or ML.

Semantic Index

The Semantic Index resolves repository-local language into candidate graph nodes. It maps concepts, symbols, features, configuration keys, environment variables, routes, and entrypoints to stable node IDs using structural analysis — no vector embeddings needed.

TERMINAL

ontoly search "authentication"
ontoly search "payment" --category feature
ontoly search "DATABASE_URL" --category environment

Intent resolution

When a developer or agent asks a question, Ontoly resolves the natural language intent into a structured query plan. The pipeline moves through intent parsing, candidate expansion, stable-node selection, scoped traversal, ranking, and bounded evidence construction.

TERMINAL

ontoly find "user authentication"
ontoly locate "payment processing"

Semantics artifact

The semantics enhancer derives feature ownership, aliases, vocabulary, and neighborhoods from graph structure. Build and validate it separately when you need to inspect or debug the semantic layer.

TERMINAL

ontoly semantics build
ontoly semantics inspect
ontoly semantics validate

Evidence packs

Evidence packs are compact, bounded subgraphs designed for AI agent consumption. They rank and deduplicate graph facts instead of dumping the full graph. Maximum 500 nodes per pack. Missing evidence produces next-step commands instead of hallucinated answers.

TERMINAL

ontoly evidence "authentication flow" --limit 12