Data — What Powers the Suite

Monce Suite has no models of its own. It orchestrates 10 services that each own their models and data.

Downstream Model Inventory

Service	Snake Models	Training Records	Features	Data Source
emailclassifier	10	4,730 (40 classes)	13	Synthetic + Outlook inbox
quoteclassifier	10	5,296	11	Synthetic + CRM
requestclassifier	4	4,000 (12 features)	12	Synthetic tickets
furnitureclassifier	1	800	27	Synthetic CRM visit notes
businessclassifier	1	700	8	Synthetic (hand-tuned distributions)
selfservice	10	~6,000	varies	Synthetic search queries
salesagentclassifier	5	800-1,700/model	21	Synthetic CRM
negociationclassifier	5	500-1,280/model	12-21	Synthetic + real nego data
procurementclassifier	2	700/model	8 + 5	Synthetic invoices
benchmarkclassifier	10	~3,000/model	varies	Synthetic supplier data

What the Suite Stores

Nothing. The suite is stateless. It stores no models, no training data, no user data. Every request fans out to 10 services, collects responses, optionally calls one LLM, returns the synthesis. All intelligence lives in the downstream classifiers.

Real Data Transition

All services currently run on synthetic training data. Each service's /data page documents the planned transition to real production data from ERP, CRM, Outlook, and Salesforce sources. The suite orchestrator is data-agnostic — it works the same whether downstream models are trained on synthetic or real data.