M04 · Knowledge graph · GNN · semantic mapping

Ontology and knowledge graph

Map cryptic legacy schemas onto an enterprise semantic model the agent can reason over.

$ python -m fde_toolkit ontology

Workflow

the order an FDE actually runs it in

  1. 1Profile legacy tables, columns, cardinality and sample values.
  2. 2Propose semantic entities with a confidence score per mapping.
  3. 3Auto-accept high confidence; route the rest to human review.
  4. 4Materialise entities and relationships into the knowledge graph.
  5. 5Expose Cypher and natural-language query paths with provenance.

Function output

deterministic trace

TBL_CUST_99X_REV -> Customer            conf 0.99  AUTO_ACCEPT
EXP_LMT_TB       -> CreditExposure      conf 0.98  AUTO_ACCEPT
TMP_WRK_2019     -> (unmapped)          conf 0.31  HUMAN_REVIEW
MATCH (c:Customer)-[:OWNS]->(:Account)-[:GENERATES]->(t:Transaction) RETURN c, sum(t.amount)