feat: Change requests against a Baseline (ADR-0006)

A caller restates a request to say how requirements have changed, naming a
completed Execution of the same Customer case as the Baseline.

- app/baseline.py reads a Baseline from its files and refuses one that
  does not exist, did not complete or belongs to another case, before
  anything runs or is recorded. A Baseline is never modified.
- Requirements carry a change (unchanged, added, changed, removed) and a
  reason. Submission is refused when a Baseline requirement disappears, an
  id is reused, or an unchanged requirement's text differs. A removed
  requirement stays on record and asks nothing downstream.
- Each Specialist starts from its delivered version; the graph and the code
  are its working copy, and the description is carried over when the
  design's shape is the same.
- app/workflows/changes.py compares versions by id and traces what each
  changed requirement reaches; anything else that changed is listed as
  unrequested. Approvals show the change for their Phase, and changes.json
  is written with the full Deliverables.
- Intake no longer starts a message read as a change: it names the
  Execution to give as the Baseline.
- A fifth evaluation case runs a request and then a change to it, with
  four checks that the change is recorded, reaches Checks, stays contained
  and costs less than its Baseline.

Tested with scripted agents only; no Change request has run with real
models.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-10-09 11:38:15 +01:00
parent db28d02f56
commit 5ef434d00e
32 changed files with 1824 additions and 40 deletions

View File

@@ -306,5 +306,66 @@
"seconds": 3400,
"tokens": 7500000
}
},
{
"id": "case-05-change-retention",
"name": "A change to a delivered event-driven service: retention goes from 30 days to 7 years",
"workflow_request": "Design an event-driven regional HTTP application on Google Cloud that accepts incoming webhooks, validates their signatures, durably enqueues payloads for asynchronous processing, and retains raw events for 30 days for replay and audit.",
"change": "One thing has changed since this was delivered. Raw events must now be kept for 7 years instead of 30 days, to meet a new audit obligation, and must not be deletable during that time. Everything else stays as it is.",
"required_products": [
"Cloud Run",
"Pub/Sub",
"Cloud Storage"
],
"scripted_answers": [
[
"region|location|geograph",
"A single region, europe-west2 (London). Data must stay in the UK."
],
[
"availability|uptime|sla|rto|rpo|disaster|recover",
"99.9% monthly availability. RTO 1 hour, RPO 5 minutes for accepted events. No multi-region DR in phase one."
],
[
"traffic|volume|throughput|scale|peak|requests per|rps|payload",
"Average 50 webhooks per second, peaks of 500 per second for 10 minutes. Payloads up to 256 KB."
],
[
"latency|response time|performance",
"Acknowledge each webhook within 300 ms at p95."
],
[
"security|compliance|encrypt|privacy|gdpr|pci|access|identity|auth",
"UK GDPR applies. Google-managed encryption keys are acceptable. Senders sign requests with HMAC-SHA256 shared secrets; only processing services may read events."
],
[
"cost|budget|spend",
"Under 1,500 GBP a month at average load; prefer pay-per-use."
],
[
"current|existing|today|on-prem|migrat|legacy",
"This is a new workload; nothing runs today."
],
[
"depend|integrat|downstream|upstream|consum",
"Processed events go to an existing internal REST API that is reachable over HTTPS; it can be slow or unavailable for minutes."
],
[
"operat|monitor|logging|alert|team|support",
"A small platform team of three uses Terraform and wants managed services and Cloud Monitoring alerts."
],
[
"sustainab|carbon",
"Prefer low-carbon regions when there is no other constraint, but the UK residency requirement wins."
],
[
"retention|replay|audit|archive",
"Keep raw events 30 days, then delete. Replay must be possible for any event in that window."
],
[
"approve|confirm|proceed|look right|any changes|happy",
"Approved. Please proceed."
]
]
}
]