Forward-deployed / AI Knowledge Graph
Flowable

Topology — the decision path

  1. How many teams need processes, now and in 18 months?
    • One, JVM → embedded (atomic commits are yours; capstone-shaped systems live here).
    • Several, or any non-JVM → 2.
  2. Do the teams’ flows share data/transactions with their own domains heavily?
    • Yes → engine-per-team (embedded each): separate schemas, separate upgrade cadences.
    • No / mixed / platform-team appetite exists → standalone behind your API layer.
  3. Is the constraint actually UIs and modeling velocity, not runtime?
    • Yes → price Flowable Work (lesson 05’s worksheet) on top of the standalone shape.

Non-negotiables (any topology)

  • [ ] Never share one embedded engine’s schema across services.
  • [ ] Engine APIs are admin surfaces: behind your services, never end-client facing (Phase 3.03).
  • [ ] Whoever owns the topology owns Phase 9: the probe, the retention decision, the restore drill — write the pager rota down as part of this decision.
  • [ ] Model ownership and review path decided with the topology (repo per team vs central registry), not after the fiftieth model.

Mixed-topology boundary (write yours)

Core atomic products: __________ (embedded). Departmental/long-tail flows: __________ (standalone). UI layer: build (Phase 3 clients) / buy (Work) — decided by: __________.