Skip to main content
Deep Dive Exhibit · ~8 min

Generic Entity Scoring Engine, Native to the CRM

The hard problem is not adding an agent. It is deciding which agent - and whether any agent - should act.

01 · Orientation

An exhibit, not an article

Seven layers, each with something to touch. Start anywhere - the path remembers where you are.

You are here: Orientation

Read the exhibit contract, then pick a layer.

You are not just reading about this architecture. You are going to inspect how it works.

This exhibit walks through a generic entity scoring engine: one deterministic algorithm that grades any record type, a router that picks exactly one specialist, and schema gates around every mutation. Six short chapters, each with something to touch.

Nothing here is a production system. The fixtures are synthetic, the runs are described rather than executed, and every number comes from the author's own test harness. What is real is the reasoning: why each boundary exists and which failure it absorbs.

How to use this exhibit:
  • Inspect any component to see its purpose, boundary, and the failure it prevents
  • Step one request through the system hop by hop
  • Answer two design questions before reading how the architect reasoned
  • Try to break the system, then see where each failure lands
RECORDany entityscore firstdeclinedroutedone onlydeclinedSCHEMA GATEnothing half-writesAUDITreconstructable
Triage before tokens
1 / 7
“A well-architected system makes the right thing easier, and the wrong thing harder.”
— Kuldeep Singh