Knowledge graph construction -- entity extraction, relation mapping, and pathfinder graph traversal.
Extracts entities (classes, functions, modules, types, concepts) and relations (imports, extends, implements, depends-on, calls) from source code and documentation. Builds a navigable knowledge graph stored in AgentDB with hierarchical nodes and causal edges. Traverses the graph using a pathfinder algorithm that scores paths by edge weight and semantic similarity.
claude --plugin-dir plugins/ruflo-knowledge-graph| Agent | Model | Role |
|---|---|---|
graph-navigator |
sonnet | Entity extraction, relation mapping, knowledge graph construction, pathfinder traversal |
| Skill | Usage | Description |
|---|---|---|
kg-extract |
/kg-extract <path> |
Extract entities and relations from source files to build a knowledge graph |
kg-traverse |
/kg-traverse <entity> [--depth N] |
Pathfinder traversal starting from a seed entity |
kg extract <path> # Extract entities and relations from source files
kg traverse <entity> # Pathfinder traversal from a seed entity
kg relations <entity> # List all direct relations for an entity
kg visualize # ASCII visualization of the knowledge graph
kg search <query> # Semantic search across the graph| Type | Examples |
|---|---|
| class | UserService, AuthController |
| function | calculateDiscount, handleRequest |
| module | auth, payments, api |
| concept | authentication, caching |
| type | User, OrderStatus |
| config | database, redis, jwt |
- CLI: pinned to
@claude-flow/cliv3.6 major+minor. - Verification:
bash plugins/ruflo-knowledge-graph/scripts/smoke.shis the contract.
This plugin owns the kg-graph AgentDB namespace (kebab-case, follows the convention from ruflo-agentdb ADR-0001 §"Namespace convention"). Reserved namespaces (pattern, claude-memories, default) MUST NOT be shadowed.
Entity nodes are stored via agentdb_hierarchical-store; relation edges via agentdb_causal-edge; semantic indexing via embeddings_generate (NOT embeddings_embed — that tool name doesn't exist; ADR-0001 fixes prior references).
- Seed -- start from the target entity node
- Expand -- follow causal edges outward (configurable depth, default 3)
- Score --
relevance = edge_weight * semantic_similarity(query, node) - Prune -- remove paths below threshold (default 0.3)
- Rank -- return top-K paths by cumulative relevance
ADR-095 closed five AgentDB controllers that this plugin's graph traversal can leverage:
gnnService— GNN embeddings + relational scoring over the AgentDB causal graph. Augments the pathfinder'ssemantic_similarity(query, node)term with structurally-aware scoring; nodes that are graph-neighbors of confirmed-relevant nodes get a boost.rvfOptimizer— Quantizes + dedupes vector blocks before persistence. Knowledge-graph indexes commonly have many near-duplicate entity vectors (same class re-exported from multiple modules); rvfOptimizer collapses them transparently.mutationGuard+attestationLog+GuardedVectorBackend— Proof-gated writes to the underlying vector store. Relevant when the graph spans trust boundaries (federated knowledge import) — the attestation chain at.swarm/attestation.dbrecords every mutation for after-the-fact audit.
The yet-pending graphAdapter controller will give this plugin a first-class graph-DB backend (instead of building the graph view on top of AgentDB's flat causal-edge table). Tracked in ADR-095.
Inspect runtime status via the agentdb_controllers or agentdb_health MCP tools.
bash plugins/ruflo-knowledge-graph/scripts/smoke.sh
# Expected: "10 passed, 0 failed"ruflo-agentdb-- The G7 controllers above ship via this plugin's runtime; install both for full graph + traversal coverage; namespace convention ownerruflo-ruvector-- HNSW indexing for fast semantic search across graph nodesruflo-adr-- ADR dependency graphs share the same causal edge model
MIT