The Programmatic Stream

A Python FastAPI server deployed alongside the Next.js frontend as a Vercel Service. One domain. Two runtimes. Pull any thread.

Stack
RuntimePython 3.12, FastAPI
DeploymentVercel Services — same domain as Next.js frontend
DataEmbedded — every response is verified, structured, queryable
PatternOne domain, two runtimes, zero CORS
Endpoints
/evidenceVerified platform metrics — four categories
/gapsFifteen gaps, named honestly
/rosterTwenty-two agents with ecosystem summary
/documentsEight governing documents — the full prompt layer architecture
/documents/{layer}/{name}Read any locked document by layer and name
/transfer-evidenceSemper Fi engagement — twelve artifacts, ten confirmed decisions
/timelineBuild timeline across twenty-four sessions
/decisionsTwenty architectural and strategic decisions
/correctionsEleven corrections — what was wrong, how it was caught
/methodology-in-actionTwelve moments where methodology governed its own build
/design-principlesAgent ecosystem design principles
/architectureFive-layer composable architecture overview

The server was built in Python during the application process — the first production Python artifact. Every endpoint returns verified data. The eight governing documents that shape agent behavior are queryable through /documents. A reviewer who pulls the Soul Layer and then talks to the Intel Agent is seeing architecture become behavior.