Skip to content

feat(pilot): design-partner pilot program — framework, kit, ragctl pilot, case study (Step 7.4) - #174

Merged
officialCodeWork merged 1 commit into
mainfrom
build/phase-7/step-7.4-design-partner-pilots
Jun 9, 2026
Merged

feat(pilot): design-partner pilot program — framework, kit, ragctl pilot, case study (Step 7.4)#174
officialCodeWork merged 1 commit into
mainfrom
build/phase-7/step-7.4-design-partner-pilots

Conversation

@officialCodeWork

Copy link
Copy Markdown
Owner

Step 7.4 — Design-partner pilots

Step 7.4 is a GTM/process step. Signing referenceable design partners is an
external, human deliverable (the same way the Step 7.3 external pentest is a
documented process item). What this PR ships in-repo is the machine that makes
every pilot run identically and report objective KPIs pulled from the platform's
own signals
— never self-reported. Delivered as one PR, sliced 7.4a–d.

What shipped

Slice Deliverable
7.4a Operator runbook (guides/design-partner-pilots.md) + five fill-in templates (docs/pilots/templates/) + ADR-0046
7.4b Customer-support/KB pilot kit — sample corpus (synthetic-PII handbook + product FAQ + a planted prompt-injection probe), domain-calibrated success criteria
7.4c ragctl pilotonboard (per-tenant config + checklist) and report (weekly-KPI dashboard + PASS/FAIL verdict from the feedback / drift / cost signal components behind GET /v1/status/*)
7.4d Worked case study — the framework run end-to-end on the kit with real ragctl pilot report output

Key decision (ADR-0046)

Every success criterion maps to a platform signal — quality → eval harness +
GET /v1/status/feedback; latency → /v1/status/metrics; security →
/v1/status/compliance. KPIs are pulled, not self-reported — the same
"checkable, not just prose" posture as the Step 6.10 compliance mapping.

Real tooling output (used in the case study)

$ ragctl pilot report --tenant acme-support --week 3
  volume:        60 signals (explicit=43 implicit=17)
  satisfaction:  mean=+0.733  +52/-8  clicks=17   [ok]
  drift:         0/5 monitors drifting   [ok]
  cost:          status=ok ratio=1.00 tokens=820   [ok]
  verdict:       PASS — all online signals green

Maps to the plan's acceptance

  • Signed success criteria (quality/latency/integration/security) → the
    success-criteria template, each row
    tied to a signal.
  • Weekly status with KPIsragctl pilot report + the
    weekly-kpi template.
  • Feedback incorporated → the closed intake → triage → incorporate → close loop
    (feedback-log template).
  • Reference architecture + customer story → the
    case study.
  • 3 live referenceable tenants → tracked as the external GTM acceptance item
    layered on this framework.

Verification

  • ruff check + ruff format --check
  • mypy --strict (ragctl) ✅
  • Full ragctl test suite (+3 new test_pilot.py) ✅
  • RAG001 logging gate ✅
  • Markdown link-check across all 16 touched docs (441 links) ✅
  • No dist/ / SPI / config / core-or-wire-type change → schema-drift +
    policy-coverage gates unaffected (pilot is a pure ragctl reader)

Documentation

  • Added: docs/guides/design-partner-pilots.md, docs/pilots/ (index + 5
    templates + the customer-support kit: README, 3-doc corpus, case study),
    docs/adr/ADR-0046-design-partner-pilots.md, docs/reference/pilot.md
  • Updated: docs/README.md (guides + reference + a new pilots/ section),
    TRACKER.md (Step 7.4 ✅, 78/84, detailed section, PR-history row)

Deferred / external

Sourcing + signing the real referenceable partners (GTM); the legal +
internal-engineering vertical kits; per-pilot golden-set generation tooling; a
pilot KPI card in the admin console.

🤖 Generated with Claude Code

officialCodeWork pushed a commit that referenced this pull request Jun 9, 2026
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…lot, case study (Step 7.4)

Ship Step 7.4 as a repo-backed pilot machine (signing referenceable
partners stays an external GTM deliverable, like the 7.3 pentest), with
KPIs pulled from the platform's own signals rather than self-reported.

- 7.4a framework: operator runbook (guides/design-partner-pilots.md),
  five fill-in templates (docs/pilots/templates/), ADR-0046.
- 7.4b customer-support/KB kit: sample corpus (synthetic-PII handbook +
  product FAQ + a planted prompt-injection probe), domain-calibrated
  success criteria.
- 7.4c ragctl pilot: `onboard` renders the per-tenant config + checklist;
  `report` assembles a weekly-KPI dashboard + PASS/FAIL verdict from the
  feedback / drift / cost signal components behind GET /v1/status/*.
- 7.4d worked case study: framework run end-to-end on the kit with real
  `ragctl pilot report` output (satisfaction +0.733, 0/5 drift -> PASS).

No dist/SPI/config change; pilot is a pure ragctl reader. Gates green:
ruff, ruff format, mypy --strict, RAG001, full ragctl suite (+3 tests),
markdown link-check.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@officialCodeWork
officialCodeWork force-pushed the build/phase-7/step-7.4-design-partner-pilots branch from a280cc4 to 7a52460 Compare June 9, 2026 20:34
@officialCodeWork officialCodeWork changed the title feat(pilot): design-partner pilot program — framework + customer-support kit + ragctl pilot + case study (Step 7.4) feat(pilot): design-partner pilot program — framework, kit, ragctl pilot, case study (Step 7.4) Jun 9, 2026
@officialCodeWork
officialCodeWork merged commit 292f729 into main Jun 9, 2026
13 of 19 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant