Framework
07 — GraphRAG Retrieval
PartialCombines vector search, graph traversal, structured data, and GNN signals.
Framework07
Why this exists
Purpose
Combines vector search, graph traversal, structured data, and GNN signals.
Input / output
Input
reference query; seed rows
Output
evidence packs; lexical similarity hits
Granularity
material, attribute
Memory / agents
Memory
reference
Agents
material, cost, lifecycle, regulation
Components
graphrag.py, GET /reference/search
Implementation status
Status
Partialpartial
Retrieval is MD5 lexical hashing over synthetic seeds, not pgvector. GNN signals are not present.