Agent Tool Call
A policy-aware agent loop with planning, approval, tool execution, exception handling, and observability lanes.
Every frame below is a live Archify artifact—not a mockup. Each one is regenerated from versioned JSON IR, checked for structural integrity, and linked to the exact source that produced it.
Every scenario has named reader views with a path-aware Story Trail, shareable navigation, dark/light rendering, self-contained export, and a machine-readable quality receipt.
A policy-aware agent loop with planning, approval, tool execution, exception handling, and observability lanes.
Regions, private networks, workload owners, state, cross-region replication, audit evidence, and named boundary crossings.
A time-ordered request path covering authentication, cache fallback, persistence, return traffic, and async tracing.
A change moves through reproducible build, blocking gates, human approval, canary verification, communication, and rollback.
Detection, incident command, mitigation, stakeholder communication, escalation, rollback, and recovery evidence.
Events move through consent, streaming, PII isolation, warehouse sync, and governed downstream consumers.
A fast acknowledgement leads into durable queueing, background work, retry, final-state storage, webhook, and polling fallback.
Named producers, partitioned topics, consumer groups, idempotent state, dead letters, operator ownership, and controlled replay.
Planning, execution, review, human approval, retry, cancellation, and terminal outcomes in one state model.
The deployment object moves through build, verification, approval, promotion, health pause, rollback, and explicit terminal outcomes.
A classic AWS web stack with edge delivery, authentication, API services, cache, persistence, and background work.
$ node scripts/build-gallery.mjs
render workflow ✓
render sequence ✓
render dataflow ✓
render lifecycle ✓
render architecture ✓
receipt 77 checks / 11 artifacts / deterministic hashes