These are the flows the automated suite exercises, shown as request sequences you can run against a local engine. Each example states the version and environment it was last verified in.
Opens the documented local setup. Running needs an engine on 127.0.0.1:4310 and the bootstrap token printed by bash scripts/dev.sh — no account, no cloud cost.
When a parameter is renamed, the affected reference entries, examples and agent metadata change together; the old release keeps serving the old semantics.
Last verified: engine 1.0.0 · local PostgreSQL 17 · 2026-09-24
git commit -m "rename user to username"
POST /v1/software/$ID/jobs {"kind": "analyze"} → impact: 3 pages
POST /v1/software/$ID/publish → rel_... served atomically
GET /v1/published/$OLD/pages/$PAGE → still the previous semantics
A change touching company/* or legal/* pauses at awaiting_approval; rejecting it blocks the job instead of publishing.
Last verified: engine 1.0.0 · local PostgreSQL 17 · 2026-09-24
POST /v1/software/$ID/jobs → status: awaiting_approval
GET /v1/approvals → pending item with the protected paths
POST /v1/approvals/$ID/decision {"decision": "approved"|"rejected"}