01 / DECIDE
Architecture decision
Use it when one application contains fast paths, deep analysis, realtime, or background workloads with different constraints.
Describe each workload by modality, tool support, latency, quality, and budget, then resolve that profile to a deployable model in configuration. Map input, output, state, and side effects as one observable path before deciding which layer owns route by capability, not by a fashionable model name.
02 / BUILD
Three-step implementation
- 01
Frame the contract
Write down the caller, data classification, success condition, timeout, cancellation, and ownership. Use it when one application contains fast paths, deep analysis, realtime, or background workloads with different constraints.
- 02
Build one narrow path
Implement one end-to-end path with request correlation, typed state, and reversible failure handling. Describe each workload by modality, tool support, latency, quality, and budget, then resolve that profile to a deployable model in configuration.
- 03
Prove the outcome
Turn acceptance into a repeatable fixture, contract test, or browser test. Evaluation gates compare quality, tool success, schema success, latency, and cost against the current route.
03 / BOUND
Production boundary
Model catalogs and availability change; keep identifiers out of domain logic and retain a tested rollback.
04 / PROVE
Acceptance evidence
Evaluation gates compare quality, tool success, schema success, latency, and cost against the current route.
SOURCE / HTTP
Reproducible source probe
curl -fsSI 'https://developers.openai.com/api/docs/models' | sed -n '1,5p'