Query Commands

Fresh

Overview

Once a graph exists at graphify-out/graph.json, you can query it without rebuilding. All query commands work as slash commands and direct terminal commands.

Query Types

Semantic Search

/graphify query "what handles authentication"
/graphify query "how is data validated"
/graphify query "database connection setup"

Returns nodes and edges relevant to the question. The graph uses the community structure from Leiden clustering to contextualize results.

DFS Path Tracing

/graphify query "authentication flow" --dfs

Uses depth-first search to trace a specific path through the graph rather than returning a ranked list.

Shortest Path Between Nodes

/graphify path "UserController" "Database"
/graphify path "AuthMiddleware" "SessionStore"
/graphify path "NodeA" "NodeB"

Returns the shortest connection path between two named nodes. Useful for understanding how two seemingly unrelated components interact.

Node Names

Node names are case-sensitive and must match exactly as they appear in GRAPH_REPORT.md or the graph.html visualization.

Node Explanation

/graphify explain "SwinTransformer"
/graphify explain "AuthMiddleware"
/graphify explain "ConceptName"

Returns everything Graphify knows about a specific node: its relationships, community membership, confidence scores, and source files.

Understanding Query Results

flowchart LR
    Q[Query String] --> S[Semantic Match]
    S --> N1[Node: AuthService]
    S --> N2[Node: JWTMiddleware]
    S --> N3[Node: SessionStore]
    N1 --> E1[EXTRACTED edge to Database]
    N2 --> E2[INFERRED edge to AuthService]
    N3 --> E3[AMBIGUOUS edge to RedisCache]

Edge Confidence

Every result includes edge confidence:

Edge Type	Meaning	Trust Level
EXTRACTED	Directly found in source	High
INFERRED	Reasonable inference from context	Medium
AMBIGUOUS	Found but low confidence	Verify before acting

Querying Without an Assistant

All query commands work directly from the terminal:

graphify query "authentication flow"
graphify path "UserController" "Database"
graphify explain "AuthMiddleware"

This lets you explore the graph without opening your AI assistant.

Practical Query Patterns

Onboarding a New Codebase

/graphify explain "main"          # Entry point
/graphify query "data flow"       # How data moves
/graphify query "external deps"   # Third-party integrations

Debugging

/graphify path "ErrorSource" "ErrorHandler"
/graphify query "error handling"
/graphify explain "the failing component"

Architecture Review

/graphify query "database layer"
/graphify query "API boundaries"
/graphify query "authentication"

Token Efficiency

Using /graphify query before asking your AI assistant a complex architecture question can reduce the tokens needed by 71x — the graph gives the assistant a precise subgraph instead of requiring it to read dozens of files.