Main inputs
- graph paths from Neo4j
- edge provenance and confidence
- relationship freshness
- email and meeting evidence
- alumni, community, and investor context
- intro outcome history
- connector and target reputation
Output requirements
Every warm path response should help the user understand:- why the route exists
- why that connector ranked where they did
- what evidence supports the route
- what is weak or stale
- what to do next
Current implementation map
| Responsibility | File |
|---|---|
| Edge provenance and quality | server/services/graph/pathIntelligence.ts |
| User-facing path labels and explanations | server/services/networkSearchScoring.ts |
| Fallback pathfinding | server/services/pathFinder.ts |
| Connector and target outcome scoring | server/services/graph/introMarketplaceService.ts |
| Dashboard graph bridge | server/routes/dashboardGraphIntelligenceBridge.ts |
Ranking principles
The engine should prefer:- higher-confidence edges
- fresher relationships
- stronger observed interaction
- connectors with better outcome history
- routes with fewer weak links
- clearer explainability
- inferred-only edges
- stale or decaying relationships
- paths with weak connector-target links
- longer routes that do not add trust
What the user should see
- ranked connector
- intro success score
- graph confidence
- evidence chips
- why-ranked reasons
- weak-link diagnostics
- next recommended action