Decision support, not a diagnosis. Correlate every finding with a licensed clinician and the full clinical picture.
MACH
Athletic Insight is a modular monolith with service contracts. Diagrams below are the 2026-09-25 snapshot shipped with the app. The repo is private, so this page does not fetch GitHub raw files at runtime.
Review this slice posts the current PNG through /api/v1/imaging/interpret. Vision extracts observations and an optional focus circle. The graph returns a ranked estimate — decision support, not a diagnosis.
M
Bounded contexts: identity, knowledge, imaging, dicom, cases. One deploy unit today; contracts are split so each can move to its own runtime.
A
Every capability is reachable as JSON over /api/v1. Production DICOM catalog is also published as ORDS /ords/aiv/v1. The SPA is a client, not the source of truth.
C
Stateless handlers, env-based config, health endpoint, Postgres via Neon or embedded PGLite, Oracle ATP aivdb via ORDS when ORDS_BASE_URL is set.
H
UI never talks to SQL. Knowledge graph, DICOM catalog, cases, and vision interpretation are consumed over HTTP.
Better Auth sessions (Google / X). Cookie or preview bearer.
/api/auth
Sports-injury graph catalog and match (imaging + clinical tests → grade → RTP).
/api/v1/kg
Vision interpretation of a Cases slice or uploaded JPEG/PNG, optional focus {cx,cy,r}, then graph match. Decision support, not a diagnosis.
/api/v1/imaging
Patient / Study / Series / Instance catalog. Metadata on aivdb via ORDS. Pixels in private Object Storage aiv-dicom and the VM slice cache.
/api/v1/studies
Per-user saved matches. Isolated by session user id.
/api/v1/cases