Graph/query module

Agentic query layer proposed on top of a real, loaded supply chain graph. The built foundation carries parts, vendors, weekly positions, ontology, provenance, and caveats; the chat layer turns that into source-cited answers.

Integrated pipeline: Plant and Procurement Intelligence Pipeline
Role in stack: Graph foundation -> agentic query layer
Current state: Built graph, chat layer in active development
Best demo use: Anchor for the agentic-AI story - graph is built and loaded, the query layer is what is being built next
Honest boundary: The query layer does not yet answer real questions; development has started but nothing is verified or running here yet

Eight stages of answering one question: intake and intent classification, entity resolution, hybrid query planning, graph traversal, provenance and caveat assembly, answer composition, an analyst sign-off gate, and the answer record.

Completing the run opens a question explorer. Each answer expands into three parts — the paths actually walked, the source evidence behind them (collection, field, fact key, load timestamp), and the caveats that qualify the answer, graded info through high.

One question is deliberately unanswerable. Rather than inventing a number, the assistant explains that no relationship links the two entities involved, because that absence is recorded in the ontology as a first-class fact.

Pipeline position

1. Hybrid Supply Chain Chatbot is the graph and query layer: it holds the connected part, supplier, shortage, PO, provenance, and caveat facts that the other two surfaces need.

Feeds next

Shortage Control Tower uses those weekly positions to classify exceptions, then ERP PO Intelligence explains or proposes PO action beside the ERP.

View integrated pipeline Open module demo ↗ Business case

The demo is a static mock. It runs entirely in your browser on synthetic data — nothing is gated, nothing is collected.

See the shared reasoning core — sim mode ↗ — the same classify/validate/traverse/evidence pattern the query layer above is being built on, running live on a small invented dataset.