Separate website publishing and prepare fresh knowledge reconstruction

.github/workflows/deploy.yml

@@ -0,0 +1,45 @@
+name: Deploy site (staging)
+
+on:
+  push:
+    branches: [main]
+    paths: ["site/**", "assets/brand/**", ".github/workflows/deploy.yml"]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: deploy
+  cancel-in-progress: false
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    env:
+      CLOUDFLARE_API_TOKEN: ${{ secrets.CLOUDFLARE_API_TOKEN }}
+      CLOUDFLARE_ACCOUNT_ID: ${{ secrets.CLOUDFLARE_ACCOUNT_ID }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 22
+      - uses: astral-sh/setup-uv@v6
+      - uses: actions/cache@v4
+        with:
+          path: ~/.npm
+          key: ${{ runner.os }}-node22-quartz-npm-${{ hashFiles('site/build.sh', 'site/quartz.config.yaml', 'site/plugins/**/package.json') }}
+          restore-keys: |
+            ${{ runner.os }}-node22-quartz-npm-
+      - name: Check and build staging
+        run: bash site/build.sh --target staging
+      - name: Deploy to Cloudflare staging
+        if: env.CLOUDFLARE_API_TOKEN != '' && env.CLOUDFLARE_ACCOUNT_ID != ''
+        uses: cloudflare/wrangler-action@v3
+        with:
+          apiToken: ${{ env.CLOUDFLARE_API_TOKEN }}
+          accountId: ${{ env.CLOUDFLARE_ACCOUNT_ID }}
+          command: deploy --config site/wrangler.jsonc --env staging
+      - name: Cloudflare deployment not configured
+        if: env.CLOUDFLARE_API_TOKEN == '' || env.CLOUDFLARE_ACCOUNT_ID == ''
+        run: echo "::notice::Staging build succeeded. Deployment skipped because Cloudflare secrets are unset."

.github/workflows/validate.yml

@@ -18,3 +18,7 @@ jobs:
         run: uv run .agents/skills/validate/scripts/okf_validate.py knowledge --strict
       - name: House rules
         run: uv run tools/check_bundle.py
+      - name: Source manifest and recovery tests
+        run: |
+          python3 tools/sources.py manifest
+          python3 -m unittest discover -s tests -p test_sources.py

.gitignore

@@ -16,3 +16,12 @@ public/
 # tracked, despite the blanket `dist/` rule above.
 !site/okf-meta-plugin/dist/
 !site/okf-meta-plugin/dist/**
+viz.html
+.playwright-cli/
+
+# Raw source material reviewed for the bundle (may contain unpublished planning); never committed
+/sources
+.wrangler/
+
+# Local review renders, research scratch, and disposable outputs.
+/scratch/

.peruseshow

@@ -0,0 +1,2 @@
+# Local review artifacts stay ignored by Git but are browsable in Peruse.
+scratch/review/

CHANGELOG.md

@@ -4,10 +4,29 @@ All notable reader-facing changes to this knowledgebase. Format follows [Keep a
 
 ## [Unreleased]
 
+### Changed
+
+- Website commands now use `task site:...` from the repository root; knowledge checks and review commands remain separate.
+- The website builds only `site/content/`, with its own navigation and OIDM branding. Dev and staging show a draft banner; production builds require page review stamps.
+
+- Moved the old generated bundle to temporary `archive/knowledge`; `knowledge/` now starts a fresh reconstruction from supported sources.
+
+- Consolidated planning into one implementation plan, preserved owner guidance and source inventories in references, and moved temporary review previews out of tracked documentation.
+
+- Tightened 112 knowledge pages and their navigation descriptions, reducing repetition while preserving examples and source extracts; a five-part review then restored qualifiers, disclaimers, and quoted source material the tightening had dropped, and reverted twelve pages.
+- Diagrams are Excalidraw-rendered SVGs composed for the text column, with click-to-zoom on the site; Mermaid is no longer used.
+
 ### Added
 
+- Full Lucide and Feather license notices in diagram icons and SVGs; PNG exports now include accompanying license files.
+
+- Source manifest with provenance and handling rules, plus commands to recover and checksum-verify selected source files from Tigris without overwriting local changes.
+
+- Source-grounded summaries of 15 project diagram boards, indexed by idea and pillar with links to the source material.
+- Reusable OIDM and Anatomic Locations logos for site content and diagrams, with the original style guide and usage notes.
+
 - First full draft of the knowledgebase: 115 documents in OKF 0.2 covering the overview and vision, a 48-term glossary, the semantic foundation (finding models, common data elements, anatomic locations, exam types, terminologies), the data structures (Observation, Exam Finding List, Imaging Problem List, Imaging Persona, FHIR and IHE alignment), the applications, the roadmap and open questions, project history, a repository map, and verbatim reference extracts.
 - Authoring guide, contributing guide, and migration ledger recording every document migrated from a working repository.
 - Two validation gates (OKF 0.2 conformance and house rules) run in continuous integration.
-- Rendered site built with Quartz and deployed to GitHub Pages, with frontmatter badges showing each document's type, status, provenance, and verification.
+- Website built with Quartz, with separate Cloudflare targets for dev, staging, and production.
 - All documents are status draft pending the project lead's verification.

CONTRIBUTING.md

@@ -1,10 +1,16 @@
 # Contributing
 
-Thank you for helping document the Open Imaging Data Model. This repository holds the canonical high-level documentation; code documentation stays in the working repositories.
+This repository holds source-grounded Open Imaging Data Model knowledge and its presentation pages. Code documentation stays in the working repositories.
 
 ## Before you write
 
-Read the [authoring guide](knowledge/guides/authoring-guide.md). It defines the document types, the frontmatter every file needs, the trust workflow, how links work, and how to migrate a document from a working repository. Agents follow the same guide.
+Start with the [implementation plan](docs/plans/implementation.md) for current destinations, source rules, and review state. Maintain that plan as work progresses rather than adding overlapping task plans.
+
+New knowledge drafting remains paused until the owner completes the remaining execution-plan review. The approved organization and first node do not approve the unanswered review workflow.
+
+Read [Writing the new knowledge graph](docs/knowledge-authoring.md). It describes fresh reconstruction, metadata, links, sources, and review. The old generated bundle is temporarily held in `archive/knowledge`; write new documents from the outline and checked sources, then consult the archive for omissions. Do not edit old prose into replacements. Delete the archive after recovery is complete, as recorded in the implementation plan.
+
+Source collection and recovery instructions are in [Recover and verify source material](docs/sources.md). Run `task sources:manifest` when changing source metadata; recovery code tests run in CI without private data.
 
 ## Ground rules
 
@@ -18,7 +24,7 @@ Read the [authoring guide](knowledge/guides/authoring-guide.md). It defines the
 
 1. Branch from `main`.
 2. Write or edit under `knowledge/`. New agent-written documents start with `status: draft`.
-3. Update the directory `index.md`, add a line to `knowledge/log.md`, and, for migrations, a row in the migration ledger.
+3. Update the directory `index.md` and add a line to `knowledge/log.md`. Record substantive recovery decisions in the implementation plan as needed.
 4. Run both checkers and fix every error:
 
    ```bash