feat: add Prompt Studio v9 Generation Console and Evaluation Loop

.github/workflows/pages.yml

@@ -28,189 +28,133 @@ jobs:
     steps:
       - name: Checkout
         uses: actions/checkout@v6
-
       - name: Setup Node
         uses: actions/setup-node@v4
         with:
           node-version: 22
-
       - name: Install dependencies
         run: npm install
-
       - name: Validate Pages production artifact contract
         run: node scripts/validate-pages-contract.mjs
-
       - name: Validate Deep Review Workspace contract
         run: node scripts/validate-deep-review-workspace.mjs
-
       - name: Validate Review Player evidence boundary
         run: node scripts/validate-review-player.mjs
-
       - name: Validate Promotion Workspace contract
         run: node scripts/validate-promotion-workspace.mjs
-
       - name: Validate strategic Coverage Planner contract
         run: node scripts/validate-coverage-planner.mjs
-
       - name: Validate source adapter registry, risk policy and health engine
         run: node scripts/validate-source-adapters.mjs
-
       - name: Validate Operations Command Center
         run: node scripts/validate-operations-command-center.mjs
-
       - name: Validate Operations production wiring
         run: node scripts/validate-operations-production.mjs
-
       - name: Validate global Command Palette
         run: node scripts/validate-command-palette.mjs
-
       - name: Validate Command Palette production wiring
         run: node scripts/validate-command-palette-production.mjs
-
       - name: Validate Workspace Evidence Bundles
         run: node scripts/validate-workspace-bundles.mjs
-
       - name: Validate Workspace Bundle production wiring
         run: node scripts/validate-workspace-bundles-production.mjs
-
       - name: Validate Curated Rotation Planner
         run: node scripts/validate-rotation-planner.mjs
-
       - name: Validate Curated Rotation production wiring
         run: node scripts/validate-rotation-production.mjs
-
       - name: Validate Prompt Studio engine and safety contract
         run: node scripts/validate-prompt-studio.mjs
-
       - name: Validate Prompt Studio local AI contract
         run: node scripts/validate-prompt-studio-ai.mjs
-
       - name: Validate Prompt Studio production wiring
         run: node scripts/validate-prompt-studio-production.mjs
-
       - name: Validate Prompt Studio v2 Rule Packs
         run: node scripts/validate-prompt-studio-v2-rule-packs.mjs
-
       - name: Validate Prompt Studio v2 production wiring
         run: node scripts/validate-prompt-studio-v2-production.mjs
-
       - name: Validate Prompt Studio v3 Variables Ingredients and Timeline
         run: node scripts/validate-prompt-studio-production-tools.mjs
-
       - name: Validate Prompt Studio v3 Shared Ingredient Library
         run: node scripts/validate-prompt-studio-ingredient-library.mjs
-
       - name: Validate Prompt Studio v3 extension persistence
         run: node scripts/validate-prompt-studio-v3-persistence.mjs
-
       - name: Validate Prompt Studio v3 production wiring
         run: node scripts/validate-prompt-studio-v3-production.mjs
-
       - name: Validate Prompt Studio v4 Storyboard Variants and Handoff engines
         run: node scripts/validate-prompt-studio-v4-engines.mjs
-
       - name: Validate Prompt Studio v4 workflow guard behavior
         run: node scripts/validate-prompt-studio-v4-workflow-guard.mjs
-
       - name: Validate Prompt Studio v4 production wiring
         run: node scripts/validate-prompt-studio-v4-production.mjs
-
       - name: Validate Prompt Studio v5 Repair Blueprints and Seedance engines
         run: node scripts/validate-prompt-studio-v5-engines.mjs
-
       - name: Validate Prompt Studio v5 workflow guard behavior
         run: node scripts/validate-prompt-studio-v5-workflow-guard.mjs
-
       - name: Validate Prompt Studio v5 production wiring
         run: node scripts/validate-prompt-studio-v5-production.mjs
-
       - name: Validate Prompt Studio v6 audio reference engine
         run: node scripts/validate-prompt-studio-v6-audio.mjs
-
       - name: Validate Prompt Studio v6 audio reference UI
         run: node scripts/validate-prompt-studio-v6-audio-ui.mjs
-
       - name: Validate Prompt Studio v6 production wiring
         run: node scripts/validate-prompt-studio-v6-production.mjs
-
       - name: Validate Prompt Studio v7 Generation Results engine
         run: node scripts/validate-prompt-studio-v7-generation-results.mjs
-
       - name: Validate Prompt Studio v7 staged UI
         run: node scripts/validate-prompt-studio-v7-ui.mjs
-
       - name: Validate Prompt Studio v7 production wiring
         run: node scripts/validate-prompt-studio-v7-production.mjs
-
       - name: Validate Prompt Studio v8 variant batch plan
         run: node scripts/validate-prompt-studio-v8-batch.mjs
-
       - name: Validate Prompt Studio v8 batch UI
         run: node scripts/validate-prompt-studio-v8-ui.mjs
-
       - name: Validate Prompt Studio v8 production wiring
         run: node scripts/validate-prompt-studio-v8-production.mjs
-
+      - name: Validate Prompt Studio v9 evaluation and decision readiness
+        run: node scripts/validate-prompt-studio-v9-evaluation.mjs
+      - name: Validate Prompt Studio v9 decision integrity audit
+        run: node scripts/validate-prompt-studio-v9-decision-audit.mjs
+      - name: Validate Prompt Studio v9 Generation Console and guards
+        run: node scripts/validate-prompt-studio-v9-console-ui.mjs
+      - name: Validate Prompt Studio v9 production wiring
+        run: node scripts/validate-prompt-studio-v9-production.mjs
       - name: Validate protected exact-100 browser render
         run: npm run validate:render
-
       - name: Build Research Corpus + strategic operations snapshots for Pages
         continue-on-error: true
         run: npm run case:refresh
-
       - name: Prepare static Pages site
         shell: bash
         run: |
           set -euo pipefail
           rm -rf _site
           mkdir -p _site
-
           cp studio/library.html _site/index.html
-
-          # Every browser module and stylesheet in Studio is published automatically.
-          # `validate-pages-contract.mjs` verifies the import graph and critical assets
-          # before this step, so new workspaces cannot silently 404 in production.
           find studio -maxdepth 1 -type f \( -name '*.js' -o -name '*.css' \) -exec cp {} _site/ \;
-
-          # Research/operations snapshots are best-effort. The curated app stays live
-          # if an upstream source is temporarily unavailable.
           for snapshot in case-candidates.json case-review-queue.json coverage-plan.json source-health.json; do
-            if [[ -f "studio/$snapshot" ]]; then
-              cp "studio/$snapshot" "_site/$snapshot"
-            fi
+            if [[ -f "studio/$snapshot" ]]; then cp "studio/$snapshot" "_site/$snapshot"; fi
           done
-
           touch _site/.nojekyll
-
-          # Project Pages is hosted under /Seedance-Porter/, so local root-relative
-          # entry assets become relative in the published index.
           sed -i 's|href="/library.css"|href="./library.css"|g' _site/index.html
           sed -i 's|href="/digest.css"|href="./digest.css"|g' _site/index.html
           sed -i 's|href="/case-intelligence.css"|href="./case-intelligence.css"|g' _site/index.html
           sed -i 's|href="/case-corpus.css"|href="./case-corpus.css"|g' _site/index.html
           sed -i 's|src="/library.js"|src="./library.js"|g' _site/index.html
           sed -i 's|src="/sidebar.js"|src="./sidebar.js"|g' _site/index.html
           sed -i 's|src="/case-ui.js"|src="./case-ui.js"|g' _site/index.html
-
-          test -f _site/sidebar.js
-          test -f _site/prompt-studio-ui.js
-          test -f _site/prompt-studio.css
-          test -f _site/prompt-studio-rule-packs-bootstrap.js
-          test -f _site/prompt-studio-production-tools-bootstrap.js
-          test -f _site/prompt-studio-v4-bootstrap.js
-          test -f _site/prompt-studio-v4.css
-          test -f _site/prompt-studio-v5-bootstrap.js
-          test -f _site/prompt-studio-v5.css
-          test -f _site/prompt-studio-v6-bootstrap.js
-          test -f _site/prompt-studio-v6.css
-          test -f _site/prompt-studio-v7-bootstrap.js
-          test -f _site/prompt-studio-v7.css
-          test -f _site/prompt-studio-v8-bootstrap.js
-          test -f _site/prompt-studio-v8.css
-
+          for asset in \
+            sidebar.js prompt-studio-ui.js prompt-studio.css \
+            prompt-studio-rule-packs-bootstrap.js prompt-studio-production-tools-bootstrap.js \
+            prompt-studio-v4-bootstrap.js prompt-studio-v4.css \
+            prompt-studio-v5-bootstrap.js prompt-studio-v5.css \
+            prompt-studio-v6-bootstrap.js prompt-studio-v6.css \
+            prompt-studio-v7-bootstrap.js prompt-studio-v7-workflow-guard.js prompt-studio-v7-results-ui.js prompt-studio-v7.css \
+            prompt-studio-v8-bootstrap.js prompt-studio-v8-batch-ui.js prompt-studio-generation-batch.js prompt-studio-v8.css \
+            prompt-studio-v9-bootstrap.js prompt-studio-v9-workflow-guard.js prompt-studio-v9-console-ui.js prompt-studio-generation-evaluation.js prompt-studio-generation-evaluation-audit.js prompt-studio-v9.css; do
+            test -f "_site/$asset"
+          done
       - name: Configure Pages
         uses: actions/configure-pages@v6
-
       - name: Upload Pages artifact
         uses: actions/upload-pages-artifact@v4
         with:

.github/workflows/prompt-studio-v9-ci.yml

@@ -0,0 +1,99 @@
+name: Prompt Studio v9 Generation Console CI
+
+on:
+  pull_request:
+    paths:
+      - 'studio/prompt-studio-*.js'
+      - 'studio/prompt-studio*.css'
+      - 'studio/sidebar.js'
+      - 'schemas/prompt-studio-generation*.json'
+      - 'scripts/validate-prompt-studio-v7-*.mjs'
+      - 'scripts/validate-prompt-studio-v8-*.mjs'
+      - 'scripts/validate-prompt-studio-v9-*.mjs'
+      - 'scripts/validate-pages-contract.mjs'
+      - 'docs/PROJECT-MEMORY*'
+      - 'docs/PROMPT-STUDIO-V9-*.md'
+      - 'docs/PROMPT-STUDIO-V9-STATE.json'
+      - 'docs/V9-*.md'
+      - 'docs/ADR-017-*.md'
+      - 'docs/ADR-018-*.md'
+      - '.github/workflows/pages.yml'
+      - '.github/workflows/prompt-studio-v9-ci.yml'
+  push:
+    branches: [main]
+    paths:
+      - 'studio/prompt-studio-*.js'
+      - 'studio/prompt-studio*.css'
+      - 'studio/sidebar.js'
+      - 'schemas/prompt-studio-generation*.json'
+      - 'scripts/validate-prompt-studio-v7-*.mjs'
+      - 'scripts/validate-prompt-studio-v8-*.mjs'
+      - 'scripts/validate-prompt-studio-v9-*.mjs'
+      - 'scripts/validate-pages-contract.mjs'
+      - 'docs/PROJECT-MEMORY*'
+      - 'docs/PROMPT-STUDIO-V9-*.md'
+      - 'docs/PROMPT-STUDIO-V9-STATE.json'
+      - 'docs/V9-*.md'
+      - 'docs/ADR-017-*.md'
+      - 'docs/ADR-018-*.md'
+      - '.github/workflows/pages.yml'
+      - '.github/workflows/prompt-studio-v9-ci.yml'
+  workflow_dispatch:
+
+jobs:
+  v9:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        node: [20,22,24]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: ${{ matrix.node }}
+      - run: npm install
+      - run: npm run check
+      - run: npm test
+      - name: Validate V4 Variant/Handoff baseline
+        run: node scripts/validate-prompt-studio-v4-engines.mjs
+      - name: Validate V7 Generation Results baseline
+        run: node scripts/validate-prompt-studio-v7-generation-results.mjs
+      - name: Validate V7 staged UI baseline
+        run: node scripts/validate-prompt-studio-v7-ui.mjs
+      - name: Validate V7 production baseline
+        run: node scripts/validate-prompt-studio-v7-production.mjs
+      - name: Validate V8 Variant Batch baseline
+        run: node scripts/validate-prompt-studio-v8-batch.mjs
+      - name: Validate V8 return-path UI baseline
+        run: node scripts/validate-prompt-studio-v8-ui.mjs
+      - name: Validate V8 production baseline
+        run: node scripts/validate-prompt-studio-v8-production.mjs
+      - name: Validate V9 evaluation/readiness/winner/retake engine
+        run: node scripts/validate-prompt-studio-v9-evaluation.mjs
+      - name: Validate V9 decision integrity audit
+        run: node scripts/validate-prompt-studio-v9-decision-audit.mjs
+      - name: Validate V9 Generation Console UI and workflow guards
+        run: node scripts/validate-prompt-studio-v9-console-ui.mjs
+      - name: Validate V9 production and project-memory wiring
+        run: node scripts/validate-prompt-studio-v9-production.mjs
+      - name: Validate generalized Pages browser module graph
+        run: node scripts/validate-pages-contract.mjs
+      - name: Validate protected exact-100 renderer
+        run: npm run validate:render
+      - name: Parse V9 schema/state/memory JSON
+        run: |
+          node -e "for (const p of ['schemas/prompt-studio-generation-evaluation.schema.json','docs/PROJECT-MEMORY.json','docs/PROMPT-STUDIO-V9-STATE.json']) JSON.parse(require('fs').readFileSync(p,'utf8')); console.log('json-ok')"
+      - name: Syntax-check V9 runtime and contracts
+        run: |
+          node --check studio/prompt-studio-generation-evaluation.js
+          node --check studio/prompt-studio-generation-evaluation-audit.js
+          node --check studio/prompt-studio-v9-workflow-guard.js
+          node --check studio/prompt-studio-v9-console-ui.js
+          node --check studio/prompt-studio-v9-bootstrap.js
+          node --check studio/sidebar.js
+          node --check scripts/validate-prompt-studio-v9-evaluation.mjs
+          node --check scripts/validate-prompt-studio-v9-decision-audit.mjs
+          node --check scripts/validate-prompt-studio-v9-console-ui.mjs
+          node --check scripts/validate-prompt-studio-v9-production.mjs
+          node --check scripts/validate-pages-contract.mjs

docs/ADR-017-V9-HUMAN-WINNER.md

@@ -0,0 +1,88 @@
+# ADR-017 — Winner is a human decision after full comparison review
+
+Status: Accepted  
+Date: 2026-08-08  
+Scope: Prompt Studio v9 Generation Console
+
+## Context
+
+V9 has a 13-dimension production Evaluation and transparent average score. The score helps inspection but cannot safely answer the final production question by itself.
+
+Two failure modes must be prevented:
+
+1. highest score silently becomes Winner;
+2. preferred candidate is reviewed deeply while alternatives are barely reviewed, then Winner is selected anyway.
+
+A third problem appears later: if Evaluations are edited after the decision, future learning must still know which evidence existed **at decision time**.
+
+## Decision
+
+Evaluation, Decision Readiness and Winner are separate states.
+
+### Evaluation
+
+- one canonical Evaluation per generated task;
+- verdict is `candidate | retake | reject`;
+- `winner` is not an Evaluation verdict;
+- overall score is only the average of rated dimensions.
+
+### Per-candidate readiness
+
+An Evaluation is decision-ready only when:
+
+- at least 5 dimensions are rated;
+- at least 3 dimensions contain note/evidence;
+- production-readiness is rated.
+
+### Comparison-wide readiness
+
+A Comparison contains 2–8 unique succeeded visual results.
+
+**Every candidate must be decision-ready before any Winner action is enabled or accepted by the engine.**
+
+### Human Winner
+
+Winner requires:
+
+- comparison is decision-ready;
+- selected task belongs to comparison;
+- task/export lineage is intact;
+- non-empty human rationale;
+- explicit user action.
+
+No numeric score, threshold or ranking automatically chooses the Winner.
+
+### Historical evidence snapshot
+
+At Winner selection time, V9 snapshots the Evaluation version for **all comparison candidates**:
+
+- task ID;
+- Evaluation ID;
+- Evaluation updatedAt;
+- overall score;
+- verdict;
+- rated/evidence counts;
+- decision-ready flag.
+
+Later Evaluation edits do not rewrite the Winner.
+
+Decision Audit reports post-decision review changes as `winner-evidence-drift` warning. A malformed or incomplete snapshot set is a hard integrity error.
+
+## Consequences
+
+Positive:
+
+- alternatives receive symmetric minimum review;
+- score remains evidence, not authority;
+- rationale remains human and durable;
+- historical decision evidence is reconstructable;
+- V10 can learn from the evidence state that actually produced the decision, not from later edits.
+
+Tradeoffs:
+
+- Winner requires more review work;
+- changing an Evaluation after Winner may produce an audit warning until the team deliberately re-evaluates the decision.
+
+## Invariant
+
+**Review every candidate. Then decide explicitly. Snapshot the evidence state that existed when the decision was made.**