Skip to main content
The graph APIs are the core of Introd. They compute TrustRank, rank paths, expose graph health, surface sharing controls, and feed intro execution back into future ranking.

​
Route groups

Route groupWhat it doesAuth
/api/graph/healthNeo4j health and availabilityPublic
/api/graph/recompute, /sync, /org-syncRefresh TrustRank and graph materializationSession
/api/graph/person/:id, /top, /trustrank, /stats, /clustersDirect graph intelligence and TrustRank viewsSession
/api/graph/paths, /searchRaw graph pathfinding and graph-aware searchSession
/api/graph/alumni, /communities, /investor-networks, /events, /relationship-mapCommunity and network overlaysSession
/api/graph/sharing*Graph sharing settings and unlock logicSession
/api/graph/intro-outcomes*, /api/graph/intro-marketplace*Outcome dashboards and connector-route rankingSession
/api/dashboard/people, /companies, /graph/pathsDashboard-specific graph summariesDashboard bridge

​
First call

curl https://api.getintrod.ai/api/graph/health

​
Dashboard path response

The highest-value dashboard graph endpoint today is GET /api/dashboard/graph/paths. It resolves a target by contact ID, LinkedIn identity, or name and returns ranked connector routes.

​
Useful query parameters

ParameterMeaning
emailIntrod user whose graph should be queried
targetContactIdDirect graph person ID if already known
targetPersonIdentityIdStable person-identity key
targetLinkedInUrlLinkedIn URL for target resolution
targetNameFallback name lookup
targetCompanyExtra disambiguation for name lookup
limitResult cap, default 5, max 25

​
Response shape

{
  "userId": "usr_123",
  "email": "[email protected]",
  "target": {
    "id": "per_target",
    "name": "Sarah Guo",
    "title": "Founder",
    "company": "Conviction",
    "linkedinUrl": "https://www.linkedin.com/in/example/"
  },
  "paths": [
    {
      "id": "per_target:per_connector:0",
      "pathLength": 2,
      "source": "neo4j",
      "reason": "Recent email and shared investor context",
      "confidenceScore": 84,
      "introSuccessScore": 88,
      "relationshipStrength": "strong",
      "relationshipStrengthLabel": "Strong",
      "mutualCount": 3,
      "evidenceChips": [
        "6 emails exchanged",
        "1 meetings",
        "Shared investor network"
      ],
      "graphConfidence": 0.84,
      "scoreBreakdown": {
        "connectorTargetStrength": 0.83,
        "userConnectorStrength": 0.91,
        "emailEvidenceScore": 0.78,
        "meetingEvidenceScore": 0.64
      },
      "connector": {
        "id": "per_connector",
        "name": "Owen",
        "title": "Founder",
        "company": "Introd",
        "linkedinUrl": "https://www.linkedin.com/in/example/",
        "strength": "strong",
        "mutualCount": 3
      }
    }
  ],
  "pathCount": 1,
  "source": "neo4j"
}

​
What Introd scores

Graph ranking is more than shortest path:
  • edge strength
  • provenance confidence
  • freshness of recent interaction
  • email and meeting evidence
  • weak-link penalties
  • connector and target trust
  • intro outcome history
  • community, alumni, and investor context
Those rules live primarily in server/services/graph/pathIntelligence.ts, server/services/networkSearchScoring.ts, and server/services/graph/introMarketplaceService.ts.

​
Fallback behavior

If Neo4j is unavailable, some routes fall back to server/services/pathFinder.ts. The fallback exists to keep the product responsive, but it should still preserve user scoping and low-confidence labeling.

​
Production rules

  • Never fabricate a warm path for an unknown target.
  • Keep graph reads tenant-scoped and workspace-aware.
  • Return weak-link diagnostics when confidence is limited.
  • Prefer explainable routes over shorter but weaker ones.