Overview¶

globaltrust-bank is a complete fictional bank generated entirely through apogee CLI commands. No application code was written by hand — every entity, repository, use case, controller and migration came from a generator.

Bounded contexts¶

graph LR
  customers((customers))
  accounts((accounts))
  transactions((transactions))
  fx((fx))
  compliance((compliance))

  customers --> accounts
  accounts --> transactions
  transactions --> fx
  transactions --> compliance
  customers --> compliance

Context	Aggregate root	Storage
`customers`	`CustomerEntity` (full_name, email, KYC status, country, language)	PostgreSQL
`accounts`	`AccountEntity` (IBAN, balance, currency, status)	PostgreSQL
`transactions`	`TransactionEntity` (amount, currency, kind, status)	PostgreSQL + audit in MongoDB
`fx`	`FxRateEntity` (base/quote currency, rate, observed_at)	PostgreSQL
`compliance`	`ComplianceFlagEntity` (subject, reason, severity, state)	PostgreSQL + RAG in Qdrant

Why these specific data stores¶

PostgreSQL — strong typing, FK constraints, money/decimal arithmetic.
MongoDB — schema-flexible audit log; KYC document attachments.
Qdrant — vector similarity for RAG over compliance regulations and customer T&Cs in 4 languages.
pgvector (also added) — alternative vector store, useful for staying within Postgres in some deployments.

How modules wire in¶

The project's requirements.txt includes:

apogee-core — base classes, ApiResponse, Alembic runner
apogee-auth — RBAC + MFA + audit
apogee-ai-providers — LLM unification
apogee-ai-rag — vector store + retrieval
apogee-ai — agents (compliance officer, FX advisor)

Dishka (DI) wires implementations to ports; the generated container.py already registers everything.