Glossary
The words, defined.
Fourteen terms that come up around code graphs, each in one line, each with the longer version a click away.
Terms
Knowledge graph to blast radius.
Each links to a page of its own with the longer answer.
- Knowledge graph
- Entities and the typed relationships between them, so software can follow how things connect.
- Code graph
- A knowledge graph built from a codebase: functions, config and docs become nodes; calls and imports become edges.
- Node
- A single entity in the graph, a function or class or file, carrying its name, path and kind.
- Edge
- A typed, directed relationship between two nodes, carrying a tag saying where it came from.
- Community
- A group of nodes densely joined to each other and loosely to the rest, usually a real subsystem.
- God node
- The node with far more edges than the rest, the module everything imports, where change is riskiest.
- Confidence tags
- EXTRACTED, INFERRED and AMBIGUOUS, on every edge, so you know what was found and what was guessed.
- MCP server
- A server speaking the Model Context Protocol, the open standard for letting an assistant call an outside tool.
- RAG
- Chunk, embed, and retrieve what looks most like the question. Lookup by resemblance, not structure.
- Vector embedding
- Numbers standing for the meaning of some text. It answers what looks like this, not what calls this.
- Tree-sitter and the AST
- The AST is the parsed structure of source code; tree-sitter is the parser library Graphify runs to get it.
- Token reduction
- What you save by answering from structure instead of pasting whole files into an assistant.
- Traversal
- Following edges from node to node. One hop is one edge, and a path of them answers a question with steps in it.
- Blast radius
- Everything a change reaches: the callers of a function, their callers, and the tests that name them.