What this cockpit is actually plugged into. The architecture diagram lists thirteen execution and integration targets; this page states plainly which are wired, which are partial and which are not connected — because a demo that implies otherwise fails at the first question.
Live sources are read at request time, and every page that depends on one shows its status.
Read-only session against datagov_meta, live on every request.
Row-level policies, claims, customers, KYC and branches. Synthetic demo rows, not Generali's book.
Providers, payers, encounters, diagnoses and health claims.
Chat and streaming deployments in use across every AI surface.
Bill and estimate extraction from scans and photographs.
The audited figures verify.mjs asserts. Local, not an integration.
Used by the market ticker and the CFO scenario planner. Unauthenticated and rate-limited — it returns nothing under load.
These need the real systems. Nothing in this cockpit reads from or writes to any of them, and no page implies it does.
No tenant configured. Sign-in is a persona picker, not authentication.
Not provisioned.
Documents are pasted or uploaded by hand.
Payment rows exist as records; no gateway is connected.
No feed. Fraud precedents come from the seeded SIU case book.
Obligations are on the register; no filing integration.
Agent findings write to the local decision log only.
Not connected.
No semantic model published to either.
Why this page exists: the SSO, data-lake, TPA, GARV/IIB, regulator and BI connections in the architecture cannot be built without those systems and credentials. Listing them as available would be the easiest claim in this build to disprove. Each becomes real when the counterpart is provided.