fix(web): stop the sheet-open rule stealing .sectionNav's layout

Review catch, and a bad one: the previous commit anchored its insertion on
`padding: 0.5rem 0.75rem;` and closed .sectionNav there. Everything that
followed in the rule — margin-bottom, box-shadow, display: flex, align-items,
gap — was orphaned into .sectionNavSheetOpen, which is only applied while the
mobile jump sheet is open.

So the sticky nav lost its flex layout, spacing and shadow in the closed
state, which is virtually every page view on every school detail page. A
site-wide regression introduced by a fix for one mobile menu.

Redone by anchoring on the complete rule, closing brace included, so nothing
can be orphaned. .sectionNav is now byte-identical to main and the diff is
purely additive; .sectionNavSheetOpen carries the z-index and nothing else.

The staging experiment that validated this fix set nav.style.zIndex = '1100'
with every other declaration intact, so it was always testing this version
rather than the broken one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

.env.example

@@ -20,7 +20,7 @@ PORT=80
 # =============================================================================
 # CORS
 # =============================================================================
-# Comma-separated list of allowed origins
+# JSON array of allowed origins (pydantic-settings format)
 # In production, only include your actual domain
 ALLOWED_ORIGINS=["https://schoolcompare.co.uk"]
 
@@ -33,13 +33,21 @@ ADMIN_API_KEY=CHANGE_THIS_TO_A_SECURE_RANDOM_KEY
 
 # Rate limiting (requests per minute per IP)
 RATE_LIMIT_PER_MINUTE=60
-RATE_LIMIT_BURST=10
+GLOBAL_RATE_LIMIT_PER_MINUTE=3000
 
 # Maximum request body size in bytes (default 1MB)
 MAX_REQUEST_SIZE=1048576
 
 # =============================================================================
-# API
+# SEARCH AND OPTIONAL FEATURE FLAGS
 # =============================================================================
-DEFAULT_PAGE_SIZE=50
-MAX_PAGE_SIZE=100
+TYPESENSE_URL=http://localhost:8108
+TYPESENSE_API_KEY=CHANGE_THIS_TO_YOUR_TYPESENSE_KEY
+
+# Empty URL disables Unleash-backed flags. Match the managed environment when used.
+UNLEASH_URL=
+UNLEASH_API_TOKEN=
+
+# Page-size limits are currently declared by route Query parameters.
+# DEFAULT_PAGE_SIZE, MAX_PAGE_SIZE and RATE_LIMIT_BURST are not reliable tuning
+# controls in the current routes; see docs/LEGACY_CODE.md.

.gitea/workflows/deploy.yml

@@ -5,6 +5,11 @@ on:
     branches:
       - main
 
+# Serialise the entire build/deploy/test cycle: no other run can move staging tags.
+concurrency:
+  group: staging-release
+  cancel-in-progress: false
+
 env:
   REGISTRY: privaterepo.sitaru.org
   BACKEND_IMAGE_NAME: ${{ gitea.repository }}-backend
@@ -12,7 +17,18 @@ env:
   PIPELINE_IMAGE_NAME: ${{ gitea.repository }}-pipeline
 
 jobs:
+  prepare:
+    runs-on: ubuntu-latest
+    outputs:
+      build_id: ${{ steps.identity.outputs.build_id }}
+    steps:
+      - id: identity
+        run: python3 -c 'import uuid; print("build_id=" + uuid.uuid4().hex)' >> "$GITHUB_OUTPUT"
+
   build-backend:
+    needs: [prepare]
+    outputs:
+      digest: ${{ steps.build.outputs.digest }}
     name: Build Backend (FastAPI)
     runs-on: ubuntu-latest
     steps:
@@ -46,17 +62,24 @@ jobs:
             type=raw,value=staging
 
       - name: Build and push Backend Docker image
+        id: build
         uses: docker/build-push-action@v5
         with:
           context: .
           file: ./Dockerfile
           push: true
+          build-args: |
+            BUILD_SHA=${{ gitea.sha }}
+            BUILD_ID=${{ needs.prepare.outputs.build_id }}
           tags: ${{ steps.meta-backend.outputs.tags }}
           labels: ${{ steps.meta-backend.outputs.labels }}
           cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.BACKEND_IMAGE_NAME }}:buildcache
           cache-to: type=registry,ref=${{ env.REGISTRY }}/${{ env.BACKEND_IMAGE_NAME }}:buildcache,mode=max
 
   build-frontend:
+    needs: [prepare]
+    outputs:
+      digest: ${{ steps.build.outputs.digest }}
     name: Build Frontend (Next.js)
     runs-on: ubuntu-latest
     steps:
@@ -90,18 +113,23 @@ jobs:
             type=raw,value=staging
 
       - name: Build and push Frontend Docker image
+        id: build
         uses: docker/build-push-action@v5
         with:
           context: ./nextjs-app
           file: ./nextjs-app/Dockerfile
           push: true
+          build-args: |
+            BUILD_SHA=${{ gitea.sha }}
+            BUILD_ID=${{ needs.prepare.outputs.build_id }}
           tags: ${{ steps.meta-frontend.outputs.tags }}
           labels: ${{ steps.meta-frontend.outputs.labels }}
-          build-args: |
-            FASTAPI_URL=http://backend:80/api
           # Cache disabled due to registry size limits
 
   build-pipeline:
+    needs: [prepare]
+    outputs:
+      digest: ${{ steps.build.outputs.digest }}
     name: Build Pipeline (Meltano + dbt + Airflow)
     runs-on: ubuntu-latest
     steps:
@@ -135,11 +163,15 @@ jobs:
             type=raw,value=staging
 
       - name: Build and push Pipeline Docker image
+        id: build
         uses: docker/build-push-action@v5
         with:
           context: ./pipeline
           file: ./pipeline/Dockerfile
           push: true
+          build-args: |
+            BUILD_SHA=${{ gitea.sha }}
+            BUILD_ID=${{ needs.prepare.outputs.build_id }}
           tags: ${{ steps.meta-pipeline.outputs.tags }}
           labels: ${{ steps.meta-pipeline.outputs.labels }}
           cache-from: type=registry,ref=${{ env.REGISTRY }}/${{ env.PIPELINE_IMAGE_NAME }}:buildcache
@@ -148,30 +180,23 @@ jobs:
   deploy-staging:
     name: Deploy to Staging
     runs-on: ubuntu-latest
-    needs: [build-backend, build-frontend, build-pipeline]
+    needs: [prepare, build-backend, build-frontend, build-pipeline]
     steps:
       - name: Trigger staging stack update
         run: curl -fsSk -X POST "${{ secrets.PORTAINER_STAGING_WEBHOOK }}"
 
-      - name: Wait for staging to become healthy
-        run: |
-          echo "Polling ${STAGING_BASE_URL} for up to 5 minutes..."
-          for i in $(seq 1 60); do
-            if curl -fsS -o /dev/null --max-time 10 "${STAGING_BASE_URL}/"; then
-              echo "Staging is up (attempt $i)"
-              exit 0
-            fi
-            sleep 5
-          done
-          echo "Staging did not become healthy in time" >&2
-          exit 1
+      - uses: actions/checkout@v4
+      - name: Verify deployed release identity
+        run: python3 scripts/ci/release.py wait
         env:
-          STAGING_BASE_URL: ${{ secrets.STAGING_BASE_URL }}
+          BASE_URL: ${{ secrets.STAGING_BASE_URL }}
+          EXPECTED_SHA: ${{ gitea.sha }}
+          EXPECTED_BUILD_ID: ${{ needs.prepare.outputs.build_id }}
 
   e2e-staging:
     name: E2E Journeys against Staging
     runs-on: ubuntu-latest
-    needs: [deploy-staging]
+    needs: [prepare, deploy-staging, build-backend, build-frontend, build-pipeline]
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4
@@ -187,11 +212,42 @@ jobs:
           npm ci
           npx playwright install --with-deps chromium
 
+      - name: Verify release before journeys
+        run: python3 scripts/ci/release.py wait --timeout 10
+        env:
+          BASE_URL: ${{ secrets.STAGING_BASE_URL }}
+          EXPECTED_SHA: ${{ gitea.sha }}
+          EXPECTED_BUILD_ID: ${{ needs.prepare.outputs.build_id }}
+
       - name: Run E2E journeys
         working-directory: e2e
         run: npx playwright test
         env:
           BASE_URL: ${{ secrets.STAGING_BASE_URL }}
+          EXPECTED_SHA: ${{ gitea.sha }}
+          EXPECTED_BUILD_ID: ${{ needs.prepare.outputs.build_id }}
+
+      - name: Verify release after journeys
+        run: python3 scripts/ci/release.py wait --timeout 10
+        env:
+          BASE_URL: ${{ secrets.STAGING_BASE_URL }}
+          EXPECTED_SHA: ${{ gitea.sha }}
+          EXPECTED_BUILD_ID: ${{ needs.prepare.outputs.build_id }}
+
+      - uses: docker/setup-buildx-action@v3
+      - uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ gitea.actor }}
+          password: ${{ secrets.REGISTRY_TOKEN }}
+      - name: Mark tested image digests as verified
+        run: python3 scripts/ci/release.py verify
+        env:
+          EXPECTED_SHA: ${{ gitea.sha }}
+          EXPECTED_BUILD_ID: ${{ needs.prepare.outputs.build_id }}
+          BACKEND_DIGEST: ${{ needs.build-backend.outputs.digest }}
+          FRONTEND_DIGEST: ${{ needs.build-frontend.outputs.digest }}
+          PIPELINE_DIGEST: ${{ needs.build-pipeline.outputs.digest }}
 
 # Production deployment is a second, manual approval: see promote.yml
 # ("Promote to Production (manual)") and docs/DEPLOY.md.

.gitea/workflows/pr-checks.yml

@@ -68,13 +68,13 @@ jobs:
           python-version: "3.12"
 
       - name: Install dependencies
-        run: pip install -r requirements.txt pytest "httpx<0.28"
+        run: pip install -r requirements.txt pytest "httpx<0.28" pyyaml
 
       - name: Import smoke test
         run: python -c "from backend.app import app; print('backend imports OK')"
 
       - name: Backend unit tests
-        run: python -m pytest backend/tests -q
+        run: python -m pytest backend/tests pipeline/tests scripts/ci/tests -q
 
   build-backend:
     name: Build Backend (no push)

.gitea/workflows/promote.yml

@@ -97,33 +97,15 @@ jobs:
           username: ${{ gitea.actor }}
           password: ${{ secrets.REGISTRY_TOKEN }}
 
-      - name: Retag approved images as prod (keeping rollback pointer)
-        run: |
-          SHORT_SHA="${{ steps.resolve.outputs.short }}"
-          for IMAGE in \
-            "${REGISTRY}/${BACKEND_IMAGE_NAME}" \
-            "${REGISTRY}/${FRONTEND_IMAGE_NAME}" \
-            "${REGISTRY}/${PIPELINE_IMAGE_NAME}"; do
-            # Keep a rollback pointer before moving :prod
-            docker buildx imagetools create -t "${IMAGE}:prod-previous" "${IMAGE}:prod" || true
-            docker buildx imagetools create -t "${IMAGE}:prod" "${IMAGE}:${SHORT_SHA}"
-            echo "Promoted ${IMAGE}:${SHORT_SHA} -> :prod"
-          done
+      - name: Resolve verified digests and promote the complete image set
+        run: python3 scripts/ci/release.py promote --output release.json
+        env:
+          EXPECTED_SHA: ${{ steps.resolve.outputs.full }}
 
       - name: Trigger production stack update
         run: curl -fsSk -X POST "${{ secrets.PORTAINER_PROD_WEBHOOK }}"
 
-      - name: Wait for production to become healthy
-        run: |
-          echo "Polling ${PROD_BASE_URL} for up to 5 minutes..."
-          for i in $(seq 1 60); do
-            if curl -fsS -o /dev/null --max-time 10 "${PROD_BASE_URL}/"; then
-              echo "Production is up (attempt $i)"
-              exit 0
-            fi
-            sleep 5
-          done
-          echo "Production did not become healthy in time" >&2
-          exit 1
+      - name: Verify production release identity
+        run: python3 scripts/ci/release.py wait --release release.json
         env:
-          PROD_BASE_URL: ${{ secrets.PROD_BASE_URL }}
+          BASE_URL: ${{ secrets.PROD_BASE_URL }}

DOCKER_DEPLOY.md

@@ -1,191 +1,17 @@
-# Docker Deployment Guide
+# Docker deployment
 
-## Quick Start
+The maintained deployment runbook is [docs/DEPLOY.md](docs/DEPLOY.md).
 
-Deploy the complete SchoolCompare stack (PostgreSQL + FastAPI + Next.js) with one command:
+- Production: `docker-compose.portainer.yml`, using `:prod` images.
+- Staging: `docker-compose.portainer.staging.yml`, using `:staging` images.
+- Builds and deployment: `.gitea/workflows/deploy.yml`.
+- Human-approved production promotion: `.gitea/workflows/promote.yml`.
 
-```bash
-docker-compose up -d
-```
+The generic `docker-compose.yml` is not a supported one-command onboarding path:
+it still uses `:latest` tags that the release workflow no longer publishes and
+lacks the full current CMS setup. Review the [legacy inventory](docs/LEGACY_CODE.md)
+before using old compose examples. Starting an empty database does not populate
+school marts.
 
-This will start:
-- **PostgreSQL** on port 5432 (database)
-- **FastAPI** on port 8000 (backend API)
-- **Next.js** on port 3000 (frontend)
-
-## Service Details
-
-### PostgreSQL Database
-- **Port**: 5432
-- **Container**: `schoolcompare_db`
-- **Credentials**:
-  - User: `schoolcompare`
-  - Password: `schoolcompare`
-  - Database: `schoolcompare`
-- **Volume**: `postgres_data` (persistent storage)
-
-### FastAPI Backend
-- **Port**: 8000 → 80 (container)
-- **Container**: `schoolcompare_backend`
-- **Built from**: Root `Dockerfile`
-- **API Endpoint**: http://localhost:8000/api
-- **Health Check**: http://localhost:8000/api/data-info
-
-### Next.js Frontend
-- **Port**: 3000
-- **Container**: `schoolcompare_nextjs`
-- **Built from**: `nextjs-app/Dockerfile`
-- **URL**: http://localhost:3000
-- **Connects to**: Backend via internal network
-
-## Commands
-
-### Start all services
-```bash
-docker-compose up -d
-```
-
-### View logs
-```bash
-# All services
-docker-compose logs -f
-
-# Specific service
-docker-compose logs -f nextjs
-docker-compose logs -f backend
-docker-compose logs -f db
-```
-
-### Check status
-```bash
-docker-compose ps
-```
-
-### Stop all services
-```bash
-docker-compose down
-```
-
-### Rebuild after code changes
-```bash
-# Rebuild and restart specific service
-docker-compose up -d --build nextjs
-
-# Rebuild all services
-docker-compose up -d --build
-```
-
-### Clean restart (remove volumes)
-```bash
-docker-compose down -v
-docker-compose up -d
-```
-
-## Initial Database Setup
-
-After first start, you may need to initialize the database:
-
-```bash
-# Enter the backend container
-docker exec -it schoolcompare_backend bash
-
-# Run migrations or data loading
-python -m backend.data_loader
-```
-
-## Accessing Services
-
-Once running:
-- **Frontend**: http://localhost:3000
-- **Backend API**: http://localhost:8000/api
-- **API Docs**: http://localhost:8000/docs (Swagger UI)
-- **Database**: localhost:5432 (use any PostgreSQL client)
-
-## Environment Variables
-
-Create a `.env` file in the root directory to customize:
-
-```env
-# Database
-POSTGRES_USER=schoolcompare
-POSTGRES_PASSWORD=your_secure_password
-POSTGRES_DB=schoolcompare
-
-# Backend
-DATABASE_URL=postgresql://schoolcompare:your_secure_password@db:5432/schoolcompare
-
-# Frontend (for client-side access)
-NEXT_PUBLIC_API_URL=http://localhost:8000/api
-```
-
-Then run:
-```bash
-docker-compose up -d
-```
-
-## Troubleshooting
-
-### Backend not connecting to database
-```bash
-# Check database health
-docker-compose ps
-
-# View backend logs
-docker-compose logs backend
-
-# Restart backend
-docker-compose restart backend
-```
-
-### Frontend not connecting to backend
-```bash
-# Check backend health
-curl http://localhost:8000/api/data-info
-
-# Check Next.js environment variables
-docker exec schoolcompare_nextjs env | grep API
-```
-
-### Port already in use
-```bash
-# Change ports in docker-compose.yml
-# For example, change "3000:3000" to "3001:3000"
-```
-
-### Rebuild from scratch
-```bash
-docker-compose down -v
-docker system prune -a
-docker-compose up -d --build
-```
-
-## Production Deployment
-
-For production, update the following:
-
-1. **Use secure passwords** in `.env` file
-2. **Configure reverse proxy** (Nginx) in front of Next.js
-3. **Enable HTTPS** with SSL certificates
-4. **Set production environment variables**:
-   ```env
-   NODE_ENV=production
-   POSTGRES_PASSWORD=<strong-password>
-   ```
-5. **Backup database** regularly:
-   ```bash
-   docker exec schoolcompare_db pg_dump -U schoolcompare schoolcompare > backup.sql
-   ```
-
-## Network Architecture
-
-```
-Internet
-    ↓
-Next.js (port 3000) ← User browsers
-    ↓ (internal network)
-FastAPI (port 8000) ← API calls
-    ↓ (internal network)
-PostgreSQL (port 5432) ← Data queries
-```
-
-All services communicate via the `schoolcompare-network` Docker network.
+For architecture, configuration and test commands, see
+[ARCHITECTURE.md](docs/ARCHITECTURE.md) and [DEVELOPMENT.md](docs/DEVELOPMENT.md).

Dockerfile

@@ -24,6 +24,12 @@ RUN pip install --no-cache-dir -r requirements.txt
 COPY backend/ ./backend/
 COPY scripts/ ./scripts/
 
+ARG BUILD_SHA=development
+ARG BUILD_ID=development
+LABEL io.schoolcompare.build-id=$BUILD_ID
+LABEL io.schoolcompare.commit=$BUILD_SHA
+RUN python -c 'import json,sys; open("backend/build-info.json", "w").write(json.dumps({"sha":sys.argv[1],"build_id":sys.argv[2]}))' "$BUILD_SHA" "$BUILD_ID"
+
 # Expose the application port
 EXPOSE 80
 

MIGRATION_SUMMARY.md

@@ -1,3 +1,5 @@
+> Historical migration record, retained for context. Setup and architecture claims below may be obsolete. Use [README.md](README.md), [architecture](docs/ARCHITECTURE.md) and [deployment](docs/DEPLOY.md) for current guidance.
+
 # SchoolCompare: Vanilla JS → Next.js Migration Summary
 
 ## Overview

README.md

@@ -1,214 +1,67 @@
-# Primary School Compass 🧒📚
+# SchoolCompare
 
-A modern web application for comparing **primary school (KS2)** performance data in **Wandsworth and Merton** over the last 5 years. Built with FastAPI and vanilla JavaScript with Chart.js visualizations.
+SchoolCompare compares schools across England: primary (KS2), secondary (KS4),
+all-through and post-16 provision, with coverage depending on the source dataset.
+It provides school search, postcode maps, comparisons, rankings, place pages,
+Ofsted information, admissions and destination measures. Editorial content lives
+in a Payload CMS blog.
 
-![Python](https://img.shields.io/badge/Python-3.9+-blue)
-![FastAPI](https://img.shields.io/badge/FastAPI-0.109-green)
-![License](https://img.shields.io/badge/License-MIT-yellow)
+## Start here
 
-## Features
+- [Architecture and data flow](docs/ARCHITECTURE.md)
+- [Development and validation](docs/DEVELOPMENT.md)
+- [Deployment and promotion](docs/DEPLOY.md)
+- [Legacy and unused-code inventory](docs/LEGACY_CODE.md)
+- [Frontend conventions](nextjs-app/README.md)
+- [CMS publishing](nextjs-app/docs/PUBLISHING.md)
 
-- 📊 **Interactive Charts** - Visualize KS2 performance trends over time
-- 🔍 **Smart Search** - Find primary schools by name in Wandsworth & Merton
-- ⚖️ **Side-by-Side Comparison** - Compare up to 5 schools simultaneously
-- 🏆 **Rankings** - View top-performing primary schools by various KS2 metrics
-- 📱 **Responsive Design** - Works beautifully on desktop and mobile
+## Repository map
 
-## Key Metrics (KS2)
+| Path | Responsibility |
+|---|---|
+| `backend/` | FastAPI routes, cached school data, read-only SQLAlchemy mappings, feature flags |
+| `nextjs-app/` | Next.js App Router, React UI, Payload CMS, frontend tests |
+| `pipeline/plugins/extractors/` | Custom Singer taps for GIAS, EES, Ofsted and other datasets |
+| `pipeline/transform/` | dbt staging/intermediate models, marts, seeds and data tests |
+| `pipeline/dags/` | Airflow extraction, transformation and publication workflows |
+| `pipeline/scripts/` | Search indexing, code generation and operational diagnostics |
+| `e2e/` | Playwright journeys against a running environment |
+| `.gitea/workflows/` | PR checks, staging deployment and manual production promotion |
+| `scripts/` | CI review tooling and historical data utilities; see the legacy inventory |
+| `docs/superpowers/`, `mockups/` | Design history and prototypes, not application entry points |
 
-The application tracks these Key Stage 2 performance indicators:
+## Runtime
 
-| Metric | Description |
-|--------|-------------|
-| **Reading Progress** | Progress in reading from KS1 to KS2 |
-| **Writing Progress** | Progress in writing from KS1 to KS2 |
-| **Maths Progress** | Progress in maths from KS1 to KS2 |
-| **Reading Expected %** | Percentage meeting expected standard in reading |
-| **Writing Expected %** | Percentage meeting expected standard in writing |
-| **Maths Expected %** | Percentage meeting expected standard in maths |
-| **Reading, Writing & Maths Combined %** | Percentage meeting expected standard in all three subjects |
+The public site is **Next.js**, not the FastAPI root page. Browser `/api/*`
+requests pass through a Next.js route handler to FastAPI. Server-rendered pages
+call FastAPI directly using `FASTAPI_URL`, including its `/api` suffix.
 
-## Quick Start
+PostgreSQL/PostGIS stores school data. Meltano/Singer extracts source data;
+dbt builds `marts.*`; FastAPI reads those tables. Typesense serves text search
+and autocomplete. Payload runs inside Next.js and owns a separate `payload`
+database schema and uploaded media.
 
-### 1. Clone and Setup
+There is **no automatic CSV import or sample dataset on startup**. A working
+school-data environment needs populated marts from the pipeline or an approved
+database snapshot. See [development](docs/DEVELOPMENT.md) before choosing a setup.
 
-```bash
-cd school_results
+## Validation
 
-# Create virtual environment
-python -m venv venv
-source venv/bin/activate  # On Windows: venv\Scripts\activate
-
-# Install dependencies
-pip install -r requirements.txt
+```sh
+cd nextjs-app
+npm ci
+npm run typecheck
+npm test -- --runInBand
 ```
 
-### 2. Run the Application
+Backend checks, pipeline validation, runtime versions and E2E requirements are
+listed in [DEVELOPMENT.md](docs/DEVELOPMENT.md). No `npm run lint` script is
+currently defined.
 
-```bash
-# Start the server
-python -m uvicorn backend.app:app --reload --port 8000
-```
-
-Then open http://localhost:8000 in your browser.
-
-The app will run with **sample data** by default, showing **110 primary schools** (66 in Wandsworth, 44 in Merton) with 5 years of KS2 performance data.
-
-### 3. (Optional) Use Real Data
-
-To use real UK school performance data:
-
-1. Visit [Compare School Performance - Download Data](https://www.compare-school-performance.service.gov.uk/download-data)
-
-2. Download **Key Stage 2** data for the years you want (2019-2024)
-   - Select "Key Stage 2" as the data type
-
-3. Place the CSV files in the `data/` folder
-
-4. Restart the server - it will automatically load and filter to Wandsworth & Merton schools
-
-**Note:** The app only displays schools in Wandsworth and Merton. Data from other areas will be filtered out.
-
-See the helper script for more details:
-```bash
-python scripts/download_data.py
-```
-
-## Project Structure
-
-```
-school_results/
-├── backend/
-│   └── app.py           # FastAPI application with all API endpoints
-├── frontend/
-│   ├── index.html       # Main HTML page
-│   ├── styles.css       # Styling (warm, editorial design)
-│   └── app.js           # Frontend JavaScript
-├── data/
-│   └── .gitkeep         # Place CSV data files here
-├── scripts/
-│   └── download_data.py # Helper for downloading/processing data
-├── requirements.txt     # Python dependencies
-└── README.md
-```
-
-## API Endpoints
-
-| Endpoint | Description |
-|----------|-------------|
-| `GET /api/schools` | List schools with optional search/filter |
-| `GET /api/schools/{urn}` | Get detailed data for a specific school |
-| `GET /api/compare?urns=...` | Compare multiple schools |
-| `GET /api/rankings` | Get school rankings by metric |
-| `GET /api/filters` | Get available filter options |
-| `GET /api/metrics` | Get available performance metrics |
-
-### Example API Usage
-
-```bash
-# Search for schools
-curl "http://localhost:8000/api/schools?search=academy"
-
-# Get school details
-curl "http://localhost:8000/api/schools/100001"
-
-# Compare schools
-curl "http://localhost:8000/api/compare?urns=100001,100002,100003"
-
-# Get rankings
-curl "http://localhost:8000/api/rankings?metric=rwm_expected_pct&year=2024"
-```
-
-## Data Format
-
-If using your own CSV data, ensure it includes these columns (or similar):
-
-| Column | Type | Description |
-|--------|------|-------------|
-| URN | Integer | Unique Reference Number |
-| SCHNAME | String | School name |
-| LA | String | Local Authority (must be Wandsworth or Merton) |
-| READPROG | Float | Reading progress score |
-| WRITPROG | Float | Writing progress score |
-| MATPROG | Float | Maths progress score |
-| PTRWM_EXP | Float | % meeting expected standard in reading, writing & maths |
-| PTREAD_EXP | Float | % meeting expected standard in reading |
-| PTWRIT_EXP | Float | % meeting expected standard in writing |
-| PTMAT_EXP | Float | % meeting expected standard in maths |
-
-The application normalizes column names automatically and filters to only show Wandsworth and Merton schools.
-
-## Technology Stack
-
-- **Backend**: FastAPI (Python) - High-performance async API framework
-- **Frontend**: Vanilla JavaScript with Chart.js
-- **Styling**: Custom CSS with CSS variables for theming
-- **Data**: Pandas for CSV processing
-
-## Design Philosophy
-
-The UI features a warm, editorial design inspired by quality publications:
-- **Typography**: DM Sans for body text, Playfair Display for headings
-- **Color Palette**: Warm cream background with coral and teal accents
-- **Interactions**: Smooth animations and hover effects
-- **Charts**: Clean, readable data visualizations
-
-## Development
-
-```bash
-# Run with auto-reload
-python -m uvicorn backend.app:app --reload --port 8000
-
-# Or run directly
-python backend/app.py
-```
-
-## Coverage
-
-This application is specifically designed for:
-
-- **School Phase**: Primary schools only (Key Stage 2)
-- **Geographic Area**: Wandsworth and Merton (London boroughs)
-- **Time Period**: Last 5 years of data (2020-2024)
-
-Note: 2021 data shows as unavailable because SATs were cancelled due to COVID-19.
-
-## Data Source
-
-Data is sourced from the UK Government's [Compare School Performance](https://www.compare-school-performance.service.gov.uk/) service, which provides official school performance data for England.
-
-**Important**: When using real data, please comply with the [terms of use](https://www.compare-school-performance.service.gov.uk/download-data) and data protection regulations.
-
-## Scheduled Jobs
-
-### Geocoding Schools (Cron Job)
-
-School postcodes are geocoded by a scheduled job, not on-demand. This improves performance and reduces API calls.
-
-**Setup the cron job** (runs weekly on Sunday at 2am):
-
-```bash
-# Edit crontab
-crontab -e
-
-# Add this line (adjust paths as needed):
-0 2 * * 0 cd /path/to/school_compare && /path/to/venv/bin/python scripts/geocode_schools.py >> /var/log/geocode_schools.log 2>&1
-```
-
-**Manual run:**
-```bash
-# Geocode only schools missing coordinates
-python scripts/geocode_schools.py
-
-# Force re-geocode all schools
-python scripts/geocode_schools.py --force
-```
-
-## License
-
-MIT License - feel free to use this project for educational purposes.
-
----
-
-Built with ❤️ for Wandsworth & Merton families
+## Deployment
 
+Work on a feature branch and open a PR. Merging to `main` builds images and
+deploys staging. Production promotion is a separate, human-triggered Gitea
+workflow. Use [DEPLOY.md](docs/DEPLOY.md) and the Portainer compose files as the
+operational references. The generic compose examples still reference `:latest`,
+which the current release workflow does not publish.

backend/app.py

@@ -26,7 +26,8 @@ from starlette.middleware.base import BaseHTTPMiddleware
 import asyncio
 from .config import settings
 from .data_loader import (
-    clear_cache,
+    build_latest_school_data,
+    load_school_data_as_dataframe,
     compute_benchmarks,
     load_school_data,
     load_latest_school_data,
@@ -39,20 +40,13 @@ from .data_loader import (
 from .data_loader import get_data_info as get_db_info
 from . import flags
 from .places import build_place_index, build_place_registry, places_for_urn
-from .schemas import METRIC_DEFINITIONS, RANKING_COLUMNS, SCHOOL_COLUMNS
+from .schemas import METRIC_DEFINITIONS, PHASE_GROUPS, RANKING_COLUMNS, SCHOOL_COLUMNS
+from .nearby_schools import select_nearby
 from .utils import clean_for_json, convert_to_native
 
 # Values to exclude from filter dropdowns (empty strings, non-applicable labels)
 EXCLUDED_FILTER_VALUES = {"", "Not applicable", "Does not apply"}
 
-# Maps user-facing phase filter values to the GIAS PhaseOfEducation values they include.
-# All-through schools appear in both primary and secondary results.
-PHASE_GROUPS: dict[str, set[str]] = {
-    "primary": {"primary", "middle deemed primary", "all-through"},
-    "secondary": {"secondary", "middle deemed secondary", "all-through", "16 plus"},
-    "all-through": {"all-through"},
-}
-
 # Must match SITE_URL in nextjs-app/lib/site.ts. The apex 301s to www, and a
 # sitemap <loc> that redirects wastes a crawl on every URL it lists.
 BASE_URL = "https://www.schoolcompare.co.uk"
@@ -272,7 +266,28 @@ def _places_payload(urn: int) -> list[dict]:
     return payload
 
 
-def _place_sitemap_rows(kinds: tuple[str, ...]) -> list[str]:
+def _nearby_schools_payload(urn: int) -> list[dict]:
+    """The nearest eligible schools this page may offer, closest first.
+
+    Phase and reach are read from the school's own row inside select_nearby,
+    so nothing here can hand it a phase that disagrees with the data.
+
+    Wrapped: a failure in selection must never 500 a page that is otherwise
+    complete, which is the posture get_supplementary_data already takes. The
+    section simply does not render.
+    """
+    try:
+        return select_nearby(load_latest_school_data(), int(urn))
+    except Exception:
+        import logging
+
+        logging.getLogger(__name__).exception(
+            "Nearby schools selection failed for urn=%s", urn
+        )
+        return []
+
+
+def _place_sitemap_rows(kinds: tuple[str, ...], registry=None) -> list[str]:
     """A <url> per place, plus a phase variant wherever that phase clears the
     threshold on its own.
 
@@ -282,7 +297,9 @@ def _place_sitemap_rows(kinds: tuple[str, ...]) -> list[str]:
     linked from the place page either.
     """
     rows: list[str] = []
-    for p in sorted(get_place_registry().values(), key=lambda p: (p.kind, p.slug)):
+    if registry is None:
+        registry = get_place_registry()
+    for p in sorted(registry.values(), key=lambda p: (p.kind, p.slug)):
         if p.kind not in kinds:
             continue
         rows.append(_url_element(BASE_URL + _place_url(p)))
@@ -296,9 +313,10 @@ def _place_sitemap_rows(kinds: tuple[str, ...]) -> list[str]:
     return rows
 
 
-def build_sitemaps() -> dict[str, str]:
+def build_sitemaps(df=None, registry=None) -> dict[str, str]:
     """Build the sitemap index and every child, keyed by name."""
-    df = load_school_data()
+    if df is None:
+        df = load_school_data()
 
     children: dict[str, str] = {
         "static.xml": _urlset(
@@ -318,7 +336,7 @@ def build_sitemaps() -> dict[str, str]:
     # measured apart from the school pages'.
     for label, kinds in (("places", ("town", "locality", "authority")),
                          ("outcodes", ("outcode",))):
-        rows = _place_sitemap_rows(kinds)
+        rows = _place_sitemap_rows(kinds, registry)
         chunks = [rows[i:i + SITEMAP_CHUNK_SIZE]
                   for i in range(0, len(rows), SITEMAP_CHUNK_SIZE)] or [[]]
         for n, chunk in enumerate(chunks, start=1):
@@ -700,6 +718,15 @@ async def get_config():
     }
 
 
+@app.get("/api/release")
+async def release_identity():
+    import json
+    from pathlib import Path
+    path = Path(__file__).with_name("build-info.json")
+    identity = json.loads(path.read_text()) if path.exists() else {"sha": "development", "build_id": "development"}
+    return JSONResponse(identity, headers={"Cache-Control": "no-store"})
+
+
 @app.get("/api/schools")
 @limiter.limit(f"{settings.rate_limit_per_minute}/minute")
 async def get_schools(
@@ -736,7 +763,7 @@ async def get_schools(
     df_latest = load_latest_school_data()
 
     if df_latest.empty:
-        return {"schools": [], "total": 0, "page": page, "page_size": 0}
+        raise HTTPException(status_code=503, detail="School data temporarily unavailable")
 
     # Use configured default if not specified
     if page_size is None:
@@ -835,8 +862,8 @@ async def get_schools(
 
     # Apply filters
     if search:
-        ts_urns = search_schools_typesense(search)
-        if ts_urns:
+        ts_urns = await asyncio.to_thread(search_schools_typesense, search)
+        if ts_urns is not None:
             urn_order = {urn: i for i, urn in enumerate(ts_urns)}
             schools_df = schools_df[schools_df["urn"].isin(set(ts_urns))].copy()
             schools_df["_ts_rank"] = schools_df["urn"].map(urn_order)
@@ -844,9 +871,9 @@ async def get_schools(
         else:
             # Fallback: Typesense unavailable, use substring match
             search_lower = search.lower()
-            mask = schools_df["school_name"].str.lower().str.contains(search_lower, na=False)
+            mask = schools_df["school_name"].str.lower().str.contains(search_lower, na=False, regex=False)
             if "address" in schools_df.columns:
-                mask = mask | schools_df["address"].str.lower().str.contains(search_lower, na=False)
+                mask = mask | schools_df["address"].str.lower().str.contains(search_lower, na=False, regex=False)
             schools_df = schools_df[mask]
 
     if local_authority:
@@ -905,7 +932,7 @@ async def get_school_details(request: Request, urn: int):
     df = load_school_data()
 
     if df.empty:
-        raise HTTPException(status_code=404, detail="No data available")
+        raise HTTPException(status_code=503, detail="School data temporarily unavailable")
 
     school_data = df[df["urn"] == urn]
 
@@ -970,6 +997,10 @@ async def get_school_details(request: Request, urn: int):
         # and authority both fall below the publish threshold has nowhere to
         # point, and the page renders without the module.
         "places": _places_payload(urn),
+        # The nearest eligible schools, closest first. Always present on a
+        # build with this code; the frontend treats absent and empty
+        # identically, which is what lets the two images deploy independently.
+        "nearby_schools": _nearby_schools_payload(urn),
         "yearly_data": clean_for_json(school_data),
         # Supplementary data (null if not yet populated by Kestra)
         "ofsted": supplementary.get("ofsted"),
@@ -1542,20 +1573,51 @@ async def get_data_info(request: Request):
     }
 
 
+_publication_lock = asyncio.Lock()
+
+
+def _prepare_publication(df):
+    if df.empty:
+        raise ValueError("Refusing to publish an empty school dataset")
+    if not {"urn", "year", "school_name"}.issubset(df.columns):
+        raise ValueError("School dataset is missing required columns")
+    if df["urn"].isna().any() or df.duplicated(["urn", "year"]).any():
+        raise ValueError("School dataset has missing URNs or duplicate school years")
+    latest = build_latest_school_data(df)
+    registry = build_place_registry(df)
+    index = build_place_index(registry)
+    sitemaps = build_sitemaps(df, registry)
+    return df, latest, registry, index, sitemaps
+
+
+def _publish(prepared):
+    # Called on the event loop with no await: routes cannot observe half a swap.
+    # The application currently runs one worker; replicas require coordination.
+    from . import data_loader
+    global _place_registry, _place_index, _place_index_source, _sitemaps
+    df, latest, registry, index, sitemaps = prepared
+    data_loader._df_cache = df
+    data_loader._df_latest_cache = latest
+    _place_registry = registry
+    _place_index = index
+    _place_index_source = registry
+    _sitemaps = sitemaps
+
+
 @app.post("/api/admin/reload")
 @limiter.limit("5/minute")
-async def reload_data(
-    request: Request,
-    _: bool = Depends(verify_admin_api_key)
-):
-    """
-    Admin endpoint to force data reload (useful after data updates).
-    Requires X-API-Key header with valid admin API key.
-    """
-    clear_cache()
-    await asyncio.to_thread(load_school_data)
-    await asyncio.to_thread(load_latest_school_data)
-    return {"status": "reloaded"}
+async def reload_data(request: Request, _: bool = Depends(verify_admin_api_key)):
+    """Validate a complete replacement before publishing it; retain data on failure."""
+    async with _publication_lock:
+        try:
+            df = await asyncio.to_thread(load_school_data_as_dataframe)
+            prepared = await asyncio.to_thread(_prepare_publication, df)
+        except Exception as exc:
+            import logging
+            logging.getLogger(__name__).exception("Dataset reload failed")
+            raise HTTPException(status_code=503, detail="Dataset reload failed; previous data retained") from exc
+        _publish(prepared)
+    return {"status": "reloaded", "schools": len(prepared[1])}
 
 
 
@@ -1607,15 +1669,16 @@ async def regenerate_sitemap(
     request: Request,
     _: bool = Depends(verify_admin_api_key),
 ):
-    """Rebuild and cache the sitemap from current school data. Called by Airflow after data updates."""
-    global _sitemaps, _place_registry
-    # Places and sitemap are rebuilt together — they read the same marts, and
-    # letting them drift apart would submit URLs for places that no longer
-    # exist.
-    _place_registry = None
-    _sitemaps = build_sitemaps()
-    n = sum(x.count("<url>") for x in _sitemaps.values())
-    return {"status": "ok", "urls": n, "sitemaps": len(_sitemaps)}
+    """Rebuild derived publication data without clearing the live registry."""
+    async with _publication_lock:
+        try:
+            prepared = await asyncio.to_thread(_prepare_publication, load_school_data())
+        except Exception as exc:
+            raise HTTPException(status_code=503, detail="Sitemap rebuild failed; previous data retained") from exc
+        _publish(prepared)
+    n = sum(x.count("<url>") for x in prepared[4].values())
+    return {"status": "ok", "urls": n, "sitemaps": len(prepared[4])}
+
 
 
 # Mount static files directly (must be after all routes to avoid catching API calls)

backend/data_loader.py

@@ -84,21 +84,58 @@ def _get_typesense_client():
         return None
 
 
-def search_schools_typesense(query: str, limit: int = 250) -> List[int]:
-    """Search Typesense. Returns URNs in relevance order, or [] if unavailable."""
+SEARCH_PAGE_SIZE = 250
+# Search results are filtered again by the API (authority, phase, postcode,
+# etc.), so one page is too small for scoped searches. Keep the candidate set
+# bounded, though: a broad query must not turn into an unbounded sequence of
+# Typesense requests. Four pages is enough to preserve useful scoped matches
+# while putting a hard ceiling on latency and upstream load.
+SEARCH_MAX_CANDIDATES = 1_000
+
+
+def search_schools_typesense(query: str) -> Optional[List[int]]:
+    """Return a bounded set of matching URNs in relevance order.
+
+    ``None`` means Typesense is unavailable; ``[]`` is a valid zero-match
+    result. The API applies its remaining filters after this search, so the
+    first few pages are fetched rather than only the first page. Once the
+    candidate ceiling is reached, the relevance-ordered prefix is returned on
+    purpose; fetching every match would make common or adversarial queries
+    unbounded.
+    """
     client = _get_typesense_client()
     if client is None:
-        return []
+        return None
+    urns: list[int] = []
+    fetched = 0
     try:
-        result = client.collections["schools"].documents.search({
-            "q": query,
-            "query_by": "school_name,local_authority,postcode",
-            "per_page": min(limit, 250),
-            "typo_tokens_threshold": 1,
-        })
-        return [int(h["document"]["urn"]) for h in result.get("hits", [])]
+        page = 1
+        while fetched < SEARCH_MAX_CANDIDATES:
+            page_size = min(SEARCH_PAGE_SIZE, SEARCH_MAX_CANDIDATES - fetched)
+            result = client.collections["schools"].documents.search({
+                "q": query,
+                "query_by": "school_name,local_authority,postcode",
+                "per_page": page_size,
+                "page": page,
+                "typo_tokens_threshold": 1,
+            })
+            hits = result.get("hits", [])
+            urns.extend(int(h["document"]["urn"]) for h in hits)
+            fetched += len(hits)
+            if fetched >= result.get("found", fetched):
+                return list(dict.fromkeys(urns))
+            if not hits:
+                raise ValueError("Search pagination ended before all matches arrived")
+            page += 1
+        logging.getLogger(__name__).info(
+            "Typesense search capped at %d candidates for query %r",
+            SEARCH_MAX_CANDIDATES,
+            query,
+        )
+        return list(dict.fromkeys(urns))
     except Exception:
-        return []
+        logging.getLogger(__name__).exception("School search unavailable")
+        return None
 
 
 # The most a public endpoint will return in one response.
@@ -188,16 +225,6 @@ def geocode_single_postcode(postcode: str) -> Optional[Tuple[float, float]]:
     return None
 
 
-def haversine_distance(lat1: float, lon1: float, lat2: float, lon2: float) -> float:
-    """Calculate great-circle distance between two points (miles)."""
-    from math import radians, cos, sin, asin, sqrt
-    lat1, lon1, lat2, lon2 = map(radians, [lat1, lon1, lat2, lon2])
-    dlat = lat2 - lat1
-    dlon = lon2 - lon1
-    a = sin(dlat / 2) ** 2 + cos(lat1) * cos(lat2) * sin(dlon / 2) ** 2
-    return 2 * asin(sqrt(a)) * 3956
-
-
 # =============================================================================
 # MAIN DATA LOAD — joins dim_school + dim_location + fact_performance
 # fact_performance is a merged KS2+KS4 table (one row per URN per year).
@@ -512,7 +539,12 @@ def load_latest_school_data() -> pd.DataFrame:
     if _df_latest_cache is not None:
         return _df_latest_cache
 
-    df = load_school_data()
+    _df_latest_cache = build_latest_school_data(load_school_data())
+    return _df_latest_cache
+
+
+def build_latest_school_data(df: pd.DataFrame) -> pd.DataFrame:
+    """Build a replacement snapshot without mutating the published caches."""
     if df.empty:
         return df
 
@@ -545,8 +577,7 @@ def load_latest_school_data() -> pd.DataFrame:
     df_latest = pd.concat([df_latest, df_no_perf], ignore_index=True)
 
     print(f"Latest-snapshot cache built: {len(df_latest)} schools")
-    _df_latest_cache = df_latest
-    return _df_latest_cache
+    return df_latest
 
 
 def clear_cache():

backend/migration.py

@@ -1,512 +0,0 @@
-"""
-Database migration logic for importing CSV data.
-Used by both CLI script and automatic startup migration.
-"""
-
-import re
-from pathlib import Path
-from typing import Dict, Optional
-
-import numpy as np
-import pandas as pd
-import requests
-
-from .config import settings
-from .database import Base, engine, get_db_session
-from .models import School, SchoolResult
-from .schemas import (
-    COLUMN_MAPPINGS,
-    LA_CODE_TO_NAME,
-    NULL_VALUES,
-    SCHOOL_TYPE_MAP,
-)
-
-
-def parse_numeric(value) -> Optional[float]:
-    """Parse a numeric value, handling special cases."""
-    if pd.isna(value):
-        return None
-    if isinstance(value, (int, float)):
-        return float(value) if not np.isnan(value) else None
-    str_val = str(value).strip().upper()
-    if str_val in NULL_VALUES or str_val == "":
-        return None
-    # Remove percentage signs if present
-    str_val = str_val.replace("%", "")
-    try:
-        return float(str_val)
-    except ValueError:
-        return None
-
-
-def extract_year_from_folder(folder_name: str) -> Optional[int]:
-    """Extract year from folder name like '2023-2024'."""
-    match = re.search(r"(\d{4})-(\d{4})", folder_name)
-    if match:
-        return int(match.group(2))
-    match = re.search(r"(\d{4})", folder_name)
-    if match:
-        return int(match.group(1))
-    return None
-
-
-def geocode_postcodes_bulk(postcodes: list) -> Dict[str, tuple]:
-    """
-    Geocode postcodes in bulk using postcodes.io API.
-    Returns dict of postcode -> (latitude, longitude).
-    """
-    results = {}
-    valid_postcodes = [
-        p.strip().upper()
-        for p in postcodes
-        if p and isinstance(p, str) and len(p.strip()) >= 5
-    ]
-    valid_postcodes = list(set(valid_postcodes))
-
-    if not valid_postcodes:
-        return results
-
-    batch_size = 100
-    total_batches = (len(valid_postcodes) + batch_size - 1) // batch_size
-
-    for i, batch_start in enumerate(range(0, len(valid_postcodes), batch_size)):
-        batch = valid_postcodes[batch_start : batch_start + batch_size]
-        print(
-            f"    Geocoding batch {i + 1}/{total_batches} ({len(batch)} postcodes)..."
-        )
-
-        try:
-            response = requests.post(
-                "https://api.postcodes.io/postcodes",
-                json={"postcodes": batch},
-                timeout=30,
-            )
-            if response.status_code == 200:
-                data = response.json()
-                for item in data.get("result", []):
-                    if item and item.get("result"):
-                        pc = item["query"].upper()
-                        lat = item["result"].get("latitude")
-                        lon = item["result"].get("longitude")
-                        if lat and lon:
-                            results[pc] = (lat, lon)
-        except Exception as e:
-            print(f"    Warning: Geocoding batch failed: {e}")
-
-    return results
-
-
-def load_csv_data(data_dir: Path) -> pd.DataFrame:
-    """Load all CSV data from data directory."""
-    all_data = []
-
-    for folder in sorted(data_dir.iterdir()):
-        if not folder.is_dir():
-            continue
-
-        year = extract_year_from_folder(folder.name)
-        if not year:
-            continue
-
-        # Specifically look for the KS2 results file
-        ks2_file = folder / "england_ks2final.csv"
-        if not ks2_file.exists():
-            continue
-
-        csv_file = ks2_file
-        print(f"  Loading {csv_file.name} (year {year})...")
-
-        try:
-            df = pd.read_csv(csv_file, encoding="latin-1", low_memory=False)
-        except Exception as e:
-            print(f"    Error loading {csv_file}: {e}")
-            continue
-
-        # Rename columns
-        df.rename(columns=COLUMN_MAPPINGS, inplace=True)
-        df["year"] = year
-
-        # Handle local authority name
-        la_name_cols = ["LANAME", "LA (name)", "LA_NAME", "LA NAME"]
-        la_name_col = next((c for c in la_name_cols if c in df.columns), None)
-
-        if la_name_col and la_name_col != "local_authority":
-            df["local_authority"] = df[la_name_col]
-        elif "LEA" in df.columns:
-            df["local_authority_code"] = pd.to_numeric(df["LEA"], errors="coerce")
-            df["local_authority"] = (
-                df["local_authority_code"]
-                .map(LA_CODE_TO_NAME)
-                .fillna(df["LEA"].astype(str))
-            )
-
-        # Store LEA code
-        if "LEA" in df.columns:
-            df["local_authority_code"] = pd.to_numeric(df["LEA"], errors="coerce")
-
-        # Map school type
-        if "school_type_code" in df.columns:
-            df["school_type"] = (
-                df["school_type_code"]
-                .map(SCHOOL_TYPE_MAP)
-                .fillna(df["school_type_code"])
-            )
-
-        # Create combined address
-        addr_parts = ["address1", "address2", "town", "postcode"]
-        for col in addr_parts:
-            if col not in df.columns:
-                df[col] = None
-
-        df["address"] = df.apply(
-            lambda r: ", ".join(
-                str(v)
-                for v in [
-                    r.get("address1"),
-                    r.get("address2"),
-                    r.get("town"),
-                    r.get("postcode"),
-                ]
-                if pd.notna(v) and str(v).strip()
-            ),
-            axis=1,
-        )
-
-        all_data.append(df)
-        print(f"    Loaded {len(df)} records")
-
-    if all_data:
-        result = pd.concat(all_data, ignore_index=True)
-        print(f"\nTotal records loaded: {len(result)}")
-        print(f"Unique schools: {result['urn'].nunique()}")
-        print(f"Years: {sorted(result['year'].unique())}")
-        return result
-
-    return pd.DataFrame()
-
-
-def migrate_data(df: pd.DataFrame, geocode: bool = False, geocode_cache: dict = None):
-    """Migrate DataFrame data to database."""
-
-    if geocode_cache is None:
-        geocode_cache = {}
-
-    # Clean URN column - convert to integer, drop invalid values
-    df = df.copy()
-    df["urn"] = pd.to_numeric(df["urn"], errors="coerce")
-    df = df.dropna(subset=["urn"])
-    df["urn"] = df["urn"].astype(int)
-
-    # Group by URN to get unique schools (use latest year's data)
-    school_data = (
-        df.sort_values("year", ascending=False).groupby("urn").first().reset_index()
-    )
-    print(f"\nMigrating {len(school_data)} unique schools...")
-
-    # Geocode postcodes that aren't already in the cache
-    geocoded = dict(geocode_cache)  # start with preserved coordinates
-    if geocode and "postcode" in df.columns:
-        cached_postcodes = {
-            str(row.get("postcode", "")).strip().upper()
-            for _, row in school_data.iterrows()
-            if int(float(str(row.get("urn", 0) or 0))) in geocode_cache
-        }
-        postcodes_needed = [
-            p for p in df["postcode"].dropna().unique()
-            if str(p).strip().upper() not in cached_postcodes
-        ]
-        if postcodes_needed:
-            print(f"\nGeocoding {len(postcodes_needed)} postcodes ({len(geocode_cache)} restored from cache)...")
-            fresh = geocode_postcodes_bulk(postcodes_needed)
-            geocoded.update(fresh)
-            print(f"  Successfully geocoded {len(fresh)} new postcodes")
-        else:
-            print(f"\nAll {len(geocode_cache)} postcodes restored from cache, skipping geocoding.")
-
-    with get_db_session() as db:
-        # Create schools
-        urn_to_school_id = {}
-        schools_created = 0
-
-        for _, row in school_data.iterrows():
-            # Safely parse URN - handle None, NaN, whitespace, and invalid values
-            urn_val = row.get("urn")
-            urn = None
-            if pd.notna(urn_val):
-                try:
-                    urn_str = str(urn_val).strip()
-                    if urn_str:
-                        urn = int(float(urn_str))  # Handle "12345.0" format
-                except (ValueError, TypeError):
-                    pass
-            if not urn:
-                continue
-
-            # Skip if we've already added this URN (handles duplicates in source data)
-            if urn in urn_to_school_id:
-                continue
-
-            # Get geocoding data
-            postcode = row.get("postcode")
-            lat, lon = None, None
-            if postcode and pd.notna(postcode):
-                coords = geocoded.get(str(postcode).strip().upper())
-                if coords:
-                    lat, lon = coords
-
-            # Safely parse local_authority_code
-            la_code = None
-            la_code_val = row.get("local_authority_code")
-            if pd.notna(la_code_val):
-                try:
-                    la_code_str = str(la_code_val).strip()
-                    if la_code_str:
-                        la_code = int(float(la_code_str))
-                except (ValueError, TypeError):
-                    pass
-
-            school = School(
-                urn=urn,
-                school_name=row.get("school_name")
-                if pd.notna(row.get("school_name"))
-                else "Unknown",
-                local_authority=row.get("local_authority")
-                if pd.notna(row.get("local_authority"))
-                else None,
-                local_authority_code=la_code,
-                school_type=row.get("school_type")
-                if pd.notna(row.get("school_type"))
-                else None,
-                school_type_code=row.get("school_type_code")
-                if pd.notna(row.get("school_type_code"))
-                else None,
-                religious_denomination=row.get("religious_denomination")
-                if pd.notna(row.get("religious_denomination"))
-                else None,
-                age_range=row.get("age_range")
-                if pd.notna(row.get("age_range"))
-                else None,
-                address1=row.get("address1") if pd.notna(row.get("address1")) else None,
-                address2=row.get("address2") if pd.notna(row.get("address2")) else None,
-                town=row.get("town") if pd.notna(row.get("town")) else None,
-                postcode=row.get("postcode") if pd.notna(row.get("postcode")) else None,
-                latitude=lat,
-                longitude=lon,
-            )
-            db.add(school)
-            db.flush()  # Get the ID
-            urn_to_school_id[urn] = school.id
-            schools_created += 1
-
-            if schools_created % 1000 == 0:
-                print(f"  Created {schools_created} schools...")
-
-        print(f"  Created {schools_created} schools")
-
-        # Create results
-        print(f"\nMigrating {len(df)} yearly results...")
-        results_created = 0
-
-        for _, row in df.iterrows():
-            # Safely parse URN
-            urn_val = row.get("urn")
-            urn = None
-            if pd.notna(urn_val):
-                try:
-                    urn_str = str(urn_val).strip()
-                    if urn_str:
-                        urn = int(float(urn_str))
-                except (ValueError, TypeError):
-                    pass
-            if not urn or urn not in urn_to_school_id:
-                continue
-
-            school_id = urn_to_school_id[urn]
-
-            # Safely parse year
-            year_val = row.get("year")
-            year = None
-            if pd.notna(year_val):
-                try:
-                    year = int(float(str(year_val).strip()))
-                except (ValueError, TypeError):
-                    pass
-            if not year:
-                continue
-
-            result = SchoolResult(
-                school_id=school_id,
-                year=year,
-                total_pupils=parse_numeric(row.get("total_pupils")),
-                eligible_pupils=parse_numeric(row.get("eligible_pupils")),
-                # Expected Standard
-                rwm_expected_pct=parse_numeric(row.get("rwm_expected_pct")),
-                reading_expected_pct=parse_numeric(row.get("reading_expected_pct")),
-                writing_expected_pct=parse_numeric(row.get("writing_expected_pct")),
-                maths_expected_pct=parse_numeric(row.get("maths_expected_pct")),
-                gps_expected_pct=parse_numeric(row.get("gps_expected_pct")),
-                science_expected_pct=parse_numeric(row.get("science_expected_pct")),
-                # Higher Standard
-                rwm_high_pct=parse_numeric(row.get("rwm_high_pct")),
-                reading_high_pct=parse_numeric(row.get("reading_high_pct")),
-                writing_high_pct=parse_numeric(row.get("writing_high_pct")),
-                maths_high_pct=parse_numeric(row.get("maths_high_pct")),
-                gps_high_pct=parse_numeric(row.get("gps_high_pct")),
-                # Progress
-                reading_progress=parse_numeric(row.get("reading_progress")),
-                writing_progress=parse_numeric(row.get("writing_progress")),
-                maths_progress=parse_numeric(row.get("maths_progress")),
-                # Averages
-                reading_avg_score=parse_numeric(row.get("reading_avg_score")),
-                maths_avg_score=parse_numeric(row.get("maths_avg_score")),
-                gps_avg_score=parse_numeric(row.get("gps_avg_score")),
-                # Context
-                disadvantaged_pct=parse_numeric(row.get("disadvantaged_pct")),
-                eal_pct=parse_numeric(row.get("eal_pct")),
-                sen_support_pct=parse_numeric(row.get("sen_support_pct")),
-                sen_ehcp_pct=parse_numeric(row.get("sen_ehcp_pct")),
-                stability_pct=parse_numeric(row.get("stability_pct")),
-                # Absence
-                reading_absence_pct=parse_numeric(row.get("reading_absence_pct")),
-                gps_absence_pct=parse_numeric(row.get("gps_absence_pct")),
-                maths_absence_pct=parse_numeric(row.get("maths_absence_pct")),
-                writing_absence_pct=parse_numeric(row.get("writing_absence_pct")),
-                science_absence_pct=parse_numeric(row.get("science_absence_pct")),
-                # Gender
-                rwm_expected_boys_pct=parse_numeric(row.get("rwm_expected_boys_pct")),
-                rwm_expected_girls_pct=parse_numeric(row.get("rwm_expected_girls_pct")),
-                rwm_high_boys_pct=parse_numeric(row.get("rwm_high_boys_pct")),
-                rwm_high_girls_pct=parse_numeric(row.get("rwm_high_girls_pct")),
-                # Disadvantaged
-                rwm_expected_disadvantaged_pct=parse_numeric(
-                    row.get("rwm_expected_disadvantaged_pct")
-                ),
-                rwm_expected_non_disadvantaged_pct=parse_numeric(
-                    row.get("rwm_expected_non_disadvantaged_pct")
-                ),
-                disadvantaged_gap=parse_numeric(row.get("disadvantaged_gap")),
-                # 3-Year
-                rwm_expected_3yr_pct=parse_numeric(row.get("rwm_expected_3yr_pct")),
-                reading_avg_3yr=parse_numeric(row.get("reading_avg_3yr")),
-                maths_avg_3yr=parse_numeric(row.get("maths_avg_3yr")),
-            )
-            db.add(result)
-            results_created += 1
-
-            if results_created % 10000 == 0:
-                print(f"  Created {results_created} results...")
-                db.flush()
-
-        print(f"  Created {results_created} results")
-
-        # Commit all changes
-        db.commit()
-        print("\nMigration complete!")
-
-
-def _apply_schema_alterations():
-    """
-    Add new columns to existing tables using ALTER TABLE … ADD COLUMN IF NOT EXISTS.
-    Safe to run on every migration — no-ops if the column already exists.
-    Add entries here whenever models.py gains new columns on an existing table.
-    """
-    alterations = [
-        # v4: Ofsted Report Card columns
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS framework VARCHAR(20)",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_safeguarding_met BOOLEAN",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_inclusion INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_curriculum_teaching INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_achievement INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_attendance_behaviour INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_personal_development INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_leadership_governance INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_early_years INTEGER",
-        "ALTER TABLE ofsted_inspections ADD COLUMN IF NOT EXISTS rc_sixth_form INTEGER",
-    ]
-    from sqlalchemy import text as sa_text
-    with engine.connect() as conn:
-        for stmt in alterations:
-            try:
-                conn.execute(sa_text(stmt))
-            except Exception as e:
-                print(f"  Warning: alteration skipped ({e})")
-        conn.commit()
-
-
-def _apply_schema_drops():
-    """
-    Drop tables retired from the schema. Idempotent (DROP … IF EXISTS), so it's
-    safe to run on every migration. Add entries here when a model is removed.
-    """
-    drops = [
-        # v6: Ofsted Parent View feature removed
-        "DROP TABLE IF EXISTS marts.fact_parent_view CASCADE",
-    ]
-    from sqlalchemy import text as sa_text
-    with engine.connect() as conn:
-        for stmt in drops:
-            try:
-                conn.execute(sa_text(stmt))
-            except Exception as e:
-                print(f"  Warning: drop skipped ({e})")
-        conn.commit()
-
-
-def run_full_migration(geocode: bool = False) -> bool:
-    """
-    Run a complete migration: drop all tables and reimport from CSV.
-
-    Returns True if successful, False if no data found.
-    Raises exception on error.
-    """
-    # Preserve existing geocoding so a reimport doesn't throw away coordinates
-    # that took a long time to compute.
-    geocode_cache: dict[int, tuple[float, float]] = {}
-    inspector = __import__("sqlalchemy").inspect(engine)
-    if "schools" in inspector.get_table_names():
-        try:
-            with get_db_session() as db:
-                rows = db.execute(
-                    __import__("sqlalchemy").text(
-                        "SELECT urn, latitude, longitude FROM schools "
-                        "WHERE latitude IS NOT NULL AND longitude IS NOT NULL"
-                    )
-                ).fetchall()
-                geocode_cache = {r.urn: (r.latitude, r.longitude) for r in rows}
-                print(f"  Saved {len(geocode_cache)} existing geocoded coordinates.")
-        except Exception as e:
-            print(f"  Warning: could not save geocode cache: {e}")
-
-    # Only drop the core KS2 tables — leave supplementary tables (ofsted, census,
-    # finance, etc.) intact so a reimport doesn't wipe integrator-populated data.
-    # schema_version is NOT dropped: it persists so restarts don't re-trigger migration.
-    ks2_tables = ["school_results", "schools"]
-    print(f"Dropping core tables: {ks2_tables} ...")
-    inspector = __import__("sqlalchemy").inspect(engine)
-    existing = set(inspector.get_table_names())
-    for tname in ks2_tables:
-        if tname in existing:
-            Base.metadata.tables[tname].drop(bind=engine)
-
-    print("Creating all tables...")
-    Base.metadata.create_all(bind=engine)
-
-    # ALTER existing supplementary tables to add any new columns.
-    # create_all() only creates missing tables; it won't add columns to tables
-    # that already exist from an older schema version.  These statements are
-    # idempotent (IF NOT EXISTS) so they're safe to run on every migration.
-    print("Applying column additions to supplementary tables...")
-    _apply_schema_alterations()
-
-    print("Dropping retired tables...")
-    _apply_schema_drops()
-
-    print("\nLoading CSV data...")
-    df = load_csv_data(settings.data_dir)
-
-    if df.empty:
-        print("Warning: No CSV data found to migrate!")
-        return False
-
-    migrate_data(df, geocode=geocode, geocode_cache=geocode_cache)
-    return True

backend/nearby_schools.py

@@ -0,0 +1,259 @@
+"""Which nearby schools a detail page may offer as alternatives.
+
+HARD FILTERS decide eligibility, and encode claims the section is not allowed
+to make. A selective school is not an alternative to a non-selective one, a
+special school is not comparable to a mainstream one, and a Girls school is not
+an option for a Boys school's reader. They never relax, at any distance, even
+where that means the section does not render at all.
+
+DISTANCE decides the order, and nothing else does.
+
+An earlier version ranked by intake similarity first and used distance only as
+a tiebreak. That put a Catholic school 2.9 miles away above the community
+school 0.3 miles down the road, and — because the row filled from the best tier
+before widening — filled all six slots with faith matches while omitting every
+school a parent could actually walk to. For a primary, a school that far is not
+a weaker option; it is not an option. Distance is a constraint and intake is a
+preference, and the ranking now says so.
+
+Similarity survives as `shared`: what a candidate genuinely has in common with
+this school, reported on its card, so a reader applies their own weighting
+instead of having ours applied for them.
+
+Pure functions over a DataFrame: no I/O, no FastAPI, no database.
+"""
+
+from __future__ import annotations
+
+import re
+
+import numpy as np
+import pandas as pd
+
+from .schemas import PHASE_GROUPS
+
+# Three fit the row; the rest are behind the carousel arrows.
+MAX_SCHOOLS = 6
+MINIMUM = 2
+
+# How far the section will reach, in miles, when nothing closer exists.
+#
+# A sanity bound rather than a target: ordering by distance already handles
+# density, so a school in a dense area fills all six slots inside a mile and
+# never sees this. It decides one thing — what happens where the area is
+# sparse — and the answer differs by phase because catchments do. Primary
+# catchments are routinely under a mile; beyond two, a primary is not a weaker
+# option but not an option, and no section is the honest answer.
+PRIMARY_RADIUS_MILES = 2.0
+SECONDARY_RADIUS_MILES = 6.0
+POST16_RADIUS_MILES = 10.0
+
+EARTH_RADIUS_MILES = 3958.8
+
+_SPECIAL = re.compile(r"\bspecial\b|pupil referral|alternative provision", re.I)
+
+# Values that mean "this school has no religious character".
+_NO_FAITH = {"", "none", "does not apply", "not applicable"}
+
+
+def is_special_provision(school_type: str | None) -> bool:
+    """Mirror of isSpecialSchool() in nextjs-app/lib/utils.ts.
+
+    Special schools carry a mainstream phase, so phase alone cannot identify
+    them. The two implementations must agree: a school the frontend treats as
+    special for benchmarking but this treats as mainstream would be dropped
+    from its own England comparison and then offered as a peer to a mainstream
+    school on the next page along.
+    """
+    return bool(_SPECIAL.search(school_type or ""))
+
+
+def is_selective(admissions_policy: str | None) -> bool:
+    """Strictly selective. Unknown counts as non-selective, which is the safe
+    direction: it can only ever exclude a pairing, never invent one."""
+    return (admissions_policy or "").strip().lower() == "selective"
+
+
+def faith_key(denomination: str | None) -> str:
+    value = (denomination or "").strip().lower()
+    return "" if value in _NO_FAITH else value
+
+
+def faith_label(denomination: str | None) -> str:
+    return denomination.strip() if faith_key(denomination) else "No religious character"
+
+
+def genders_compatible(a: str | None, b: str | None) -> bool:
+    single = {"boys", "girls"}
+    left, right = (a or "").strip().lower(), (b or "").strip().lower()
+    return not (left in single and right in single and left != right)
+
+
+def is_secondary_phase(phase: str | None) -> bool:
+    """Whether this phase takes the secondary side: secondary group membership,
+    minus all-through.
+
+    Membership is read from PHASE_GROUPS rather than tested with `"secondary" in
+    phase`, because that substring misses "16 plus" — GIAS phase 6, which
+    PHASE_GROUPS deliberately files as secondary. The substring version fails
+    silently rather than loudly: a sixth-form college is simply handed the
+    primary bucket and offered infant schools as peers.
+
+    All-through is the exception. PHASE_GROUPS lists it on both sides because it
+    belongs on both phases' place pages, but the detail page renders it with the
+    primary template, and the metric follows the template.
+    """
+    text = (phase or "").strip().lower()
+    return text != "all-through" and text in PHASE_GROUPS["secondary"]
+
+
+def radius_miles(phase: str | None) -> float:
+    """How far this phase's section will reach when nothing closer exists."""
+    if (phase or "").strip().lower() == "16 plus":
+        return POST16_RADIUS_MILES
+    return SECONDARY_RADIUS_MILES if is_secondary_phase(phase) else PRIMARY_RADIUS_MILES
+
+
+def _phase_group(is_secondary: bool) -> set[str]:
+    return PHASE_GROUPS["secondary" if is_secondary else "primary"]
+
+
+def _haversine_miles(lat1: float, lon1: float, lat2, lon2):
+    """Vectorised, matching the postcode search in app.py."""
+    lat1_r, lon1_r = np.radians(lat1), np.radians(lon1)
+    lat2_r, lon2_r = np.radians(lat2.astype(float)), np.radians(lon2.astype(float))
+    dlat, dlon = lat2_r - lat1_r, lon2_r - lon1_r
+    a = np.sin(dlat / 2) ** 2 + np.cos(lat1_r) * np.cos(lat2_r) * np.sin(dlon / 2) ** 2
+    return 2 * EARTH_RADIUS_MILES * np.arcsin(np.sqrt(a))
+
+
+def _native(value):
+    """NaN and numpy scalars both reach JSONResponse badly; normalise here so
+    the caller never has to remember to."""
+    if value is None:
+        return None
+    if isinstance(value, np.generic):
+        value = value.item()
+    if isinstance(value, float) and np.isnan(value):
+        return None
+    return value
+
+
+def _mask(series: pd.Series, predicate) -> pd.Series:
+    """A boolean mask that survives an empty frame.
+
+    `Series.apply` on an empty Series returns an empty *DataFrame*, and using
+    that as a mask silently drops every column — so the next column lookup
+    raises KeyError rather than yielding no rows. This is not hypothetical: a
+    special school with no special school near it empties the frame at the
+    provision filter, which is the ordinary case for most special schools.
+    """
+    return pd.Series([predicate(value) for value in series], index=series.index, dtype=bool)
+
+
+def _shared(subject: pd.Series, candidate: pd.Series, is_secondary: bool) -> list[str]:
+    """What this candidate genuinely has in common with the subject.
+
+    Empty is a real answer, and renders no chips at all. A card claiming a
+    shared characteristic it does not have would be worse than a bare one —
+    and since these no longer affect the order, an empty list costs the school
+    nothing but its place in the row, which distance already decided.
+    """
+    shared: list[str] = []
+
+    gender = str(subject.get("gender") or "").strip()
+    if gender and str(candidate.get("gender") or "").strip().lower() == gender.lower():
+        shared.append(gender)
+
+    if is_secondary:
+        policy = str(candidate.get("admissions_policy") or "").strip()
+        subject_policy = str(subject.get("admissions_policy") or "").strip()
+        if (
+            policy
+            and policy.lower() == subject_policy.lower()
+            and policy.lower() not in {"not applicable", "unknown"}
+        ):
+            shared.append(policy)
+
+    if faith_key(candidate.get("religious_denomination")) == faith_key(
+        subject.get("religious_denomination")
+    ):
+        shared.append(faith_label(candidate.get("religious_denomination")))
+
+    return shared
+
+
+def select_nearby(frame: pd.DataFrame, urn: int) -> list[dict]:
+    """The nearest eligible schools, closest first — at most MAX_SCHOOLS, and
+    none at all below MINIMUM.
+
+    The phase is read from the subject's own row rather than passed in, so a
+    caller cannot hand this a phase that disagrees with the data it selects
+    from.
+    """
+    subject_rows = frame[frame["urn"] == urn]
+    if subject_rows.empty:
+        return []
+    subject = subject_rows.iloc[0]
+
+    lat, lon = _native(subject.get("latitude")), _native(subject.get("longitude"))
+    if lat is None or lon is None:
+        return []
+
+    phase = subject.get("phase")
+    is_secondary = is_secondary_phase(phase)
+    reach = radius_miles(phase)
+    metric_key = "attainment_8_score" if is_secondary else "rwm_expected_pct"
+
+    candidates = frame[frame["urn"] != urn].copy()
+    for column in ("latitude", "longitude"):
+        candidates = candidates[candidates[column].notna()]
+    if candidates.empty:
+        return []
+
+    # ── Hard filters ────────────────────────────────────────────────────
+    allowed_phases = _phase_group(is_secondary)
+    candidates = candidates[
+        candidates["phase"].fillna("").str.lower().isin(allowed_phases)
+    ]
+    candidates = candidates[candidates["status"].fillna("").str.lower().str.startswith("open")]
+
+    subject_special = is_special_provision(subject.get("school_type"))
+    special = _mask(candidates["school_type"], is_special_provision)
+    candidates = candidates[special if subject_special else ~special]
+
+    subject_selective = is_selective(subject.get("admissions_policy"))
+    selective = _mask(candidates["admissions_policy"], is_selective)
+    candidates = candidates[selective if subject_selective else ~selective]
+
+    subject_gender = subject.get("gender")
+    candidates = candidates[
+        _mask(candidates["gender"], lambda g: genders_compatible(subject_gender, g))
+    ]
+    if candidates.empty:
+        return []
+
+    candidates["distance_miles"] = _haversine_miles(
+        lat, lon, candidates["latitude"].values, candidates["longitude"].values
+    ).round(1)
+
+    # ── Nearest first, and nothing else has a say ───────────────────────
+    within = candidates[candidates["distance_miles"] <= reach]
+    if len(within) < MINIMUM:
+        return []
+
+    selected = within.sort_values(["distance_miles", "urn"]).head(MAX_SCHOOLS)
+    return [
+        {
+            "urn": int(row["urn"]),
+            "school_name": str(row.get("school_name") or ""),
+            "distance_miles": float(row["distance_miles"]),
+            "school_type": _native(row.get("school_type")),
+            "age_range": _native(row.get("age_range")),
+            "shared": _shared(subject, row, is_secondary),
+            "metric_value": _native(row.get(metric_key)),
+            "metric_key": metric_key,
+            "metric_year": _native(row.get("year")),
+        }
+        for _, row in selected.iterrows()
+    ]

backend/schemas.py

@@ -532,6 +532,18 @@ RANKING_COLUMNS = [
     "gcse_grade_91_pct",
 ]
 
+# Maps user-facing phase filter values to the GIAS PhaseOfEducation values they
+# include. All-through schools appear in both primary and secondary results,
+# which is why this is a set per phase rather than a single string comparison.
+#
+# Lives here rather than in app.py because nearby_schools.py needs it too, and
+# importing app from there would be a cycle.
+PHASE_GROUPS: dict[str, set[str]] = {
+    "primary": {"primary", "middle deemed primary", "all-through"},
+    "secondary": {"secondary", "middle deemed secondary", "all-through", "16 plus"},
+    "all-through": {"all-through"},
+}
+
 # School listing columns
 SCHOOL_COLUMNS = [
     "urn",

backend/tests/test_nearby_schools.py

@@ -0,0 +1,383 @@
+"""Selection rules for the nearby-schools section.
+
+Hard filters encode claims the section is not allowed to make — that a
+selective school is an alternative to a non-selective one, that a special
+school is comparable to a mainstream one, or that a Girls school is an option
+for a Boys school's reader. They decide who is eligible.
+
+Distance decides the order, and nothing else does. An earlier version ranked by
+intake similarity first, which put a Catholic school 2.9 miles away above the
+community school 0.3 miles down the road — for a primary, a school that far is
+not a weaker option, it is not an option. Similarity is now reported on the
+card and never reorders the row.
+"""
+
+import numpy as np
+import pandas as pd
+
+from backend.nearby_schools import (
+    is_secondary_phase,
+    radius_miles,
+    select_nearby,
+)
+
+BASE_LAT, BASE_LON = 51.5000, -0.1000
+
+
+def _row(urn, name, **overrides):
+    base = {
+        "urn": urn,
+        "school_name": name,
+        "local_authority": "Testshire",
+        "school_type": "Community school",
+        "phase": "Primary",
+        "age_range": "4-11",
+        "status": "Open",
+        "gender": "Mixed",
+        "religious_denomination": "None",
+        "admissions_policy": "Not applicable",
+        "latitude": BASE_LAT,
+        "longitude": BASE_LON,
+        "year": 202425,
+        "rwm_expected_pct": 70.0,
+        "attainment_8_score": np.nan,
+    }
+    base.update(overrides)
+    return base
+
+
+def _frame(*rows):
+    return pd.DataFrame(list(rows))
+
+
+def _at(miles):
+    """A latitude `miles` north of BASE_LAT."""
+    return BASE_LAT + miles / 69.0
+
+
+# ---------------------------------------------------------------------------
+# Order: distance, and only distance
+# ---------------------------------------------------------------------------
+
+def test_returns_nearest_first():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "Mid", latitude=_at(1.0)),
+        _row(100003, "Near", latitude=_at(0.4)),
+        _row(100004, "Far", latitude=_at(1.8)),
+    )
+    result = select_nearby(frame, 100001)
+    assert [s["urn"] for s in result] == [100003, 100002, 100004]
+    assert result[0]["distance_miles"] == 0.4
+
+
+def test_a_faith_match_never_outranks_a_closer_school():
+    """The reported defect. A Catholic primary surrounded by Catholic primaries
+    showed six of them and omitted the community school down the road."""
+    frame = _frame(
+        _row(100001, "St Jude's RC Primary", religious_denomination="Roman Catholic"),
+        _row(100002, "Elm Grove Primary", religious_denomination="None", latitude=_at(0.3)),
+        _row(100003, "Holy Cross RC", religious_denomination="Roman Catholic", latitude=_at(0.8)),
+        _row(100004, "Sacred Heart RC", religious_denomination="Roman Catholic", latitude=_at(1.2)),
+        _row(100005, "St Peter's RC", religious_denomination="Roman Catholic", latitude=_at(1.6)),
+    )
+    result = select_nearby(frame, 100001)
+    assert result[0]["urn"] == 100002, "the nearest school leads, whatever its intake"
+    assert [s["distance_miles"] for s in result] == sorted(s["distance_miles"] for s in result)
+
+
+def test_the_nearest_eligible_school_is_always_shown():
+    """Whatever else changes, a section titled "nearby" cannot omit the nearest
+    school while listing one four times further away."""
+    frame = _frame(
+        _row(100001, "Subject", gender="Boys", religious_denomination="Roman Catholic"),
+        _row(100002, "Nearest", gender="Mixed", religious_denomination="None", latitude=_at(0.2)),
+        *[
+            _row(100010 + n, f"Match {n}", gender="Boys",
+                 religious_denomination="Roman Catholic", latitude=_at(0.9 + n * 0.1))
+            for n in range(6)
+        ],
+    )
+    assert select_nearby(frame, 100001)[0]["urn"] == 100002
+
+
+def test_caps_at_six_taking_the_nearest():
+    frame = _frame(
+        _row(100001, "Subject"),
+        *[_row(100010 + n, f"Peer {n}", latitude=_at(0.1 * (n + 1))) for n in range(7)],
+    )
+    result = select_nearby(frame, 100001)
+    assert len(result) == 6
+    assert 100016 not in {s["urn"] for s in result}, "the seventh-nearest is the one dropped"
+
+
+def test_fewer_than_two_matches_returns_empty():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "Only neighbour", latitude=_at(0.5)),
+    )
+    assert select_nearby(frame, 100001) == []
+
+
+def test_excludes_the_subject_school():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "A", latitude=_at(0.5)),
+        _row(100003, "B", latitude=_at(0.6)),
+    )
+    assert 100001 not in {s["urn"] for s in select_nearby(frame, 100001)}
+
+
+def test_a_school_is_never_listed_twice():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "A", latitude=_at(0.5)),
+        _row(100003, "B", latitude=_at(0.6)),
+    )
+    result = select_nearby(frame, 100001)
+    assert len(result) == len({s["urn"] for s in result})
+
+
+# ---------------------------------------------------------------------------
+# Reach: a sanity bound, not a target
+# ---------------------------------------------------------------------------
+
+def test_primary_does_not_reach_past_two_miles():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "Just inside", latitude=_at(1.9)),
+        _row(100003, "Just outside", latitude=_at(2.4)),
+        _row(100004, "Miles away", latitude=_at(4.0)),
+    )
+    # One inside the cap is below the minimum, so nothing renders at all —
+    # a primary with nothing within two miles has no nearby schools.
+    assert select_nearby(frame, 100001) == []
+
+
+def test_secondary_reaches_further_than_primary():
+    frame = _frame(
+        _row(100001, "Subject", phase="Secondary"),
+        _row(100002, "A", phase="Secondary", latitude=_at(3.0)),
+        _row(100003, "B", phase="Secondary", latitude=_at(5.5)),
+    )
+    assert {s["urn"] for s in select_nearby(frame, 100001)} == {100002, 100003}
+
+
+def test_the_cap_follows_the_phase():
+    assert radius_miles("Primary") == 2.0
+    assert radius_miles("Middle deemed primary") == 2.0
+    assert radius_miles("All-through") == 2.0
+    assert radius_miles("Secondary") == 6.0
+    assert radius_miles("Middle deemed secondary") == 6.0
+    # Post-16 is the phase people travel furthest for.
+    assert radius_miles("16 plus") == 10.0
+
+
+# ---------------------------------------------------------------------------
+# Hard filters: eligibility, never order
+# ---------------------------------------------------------------------------
+
+def test_selective_never_meets_non_selective():
+    frame = _frame(
+        _row(100001, "Grammar", phase="Secondary", admissions_policy="Selective"),
+        _row(100002, "Comp A", phase="Secondary", admissions_policy="Non-selective", latitude=_at(0.5)),
+        _row(100003, "Comp B", phase="Secondary", admissions_policy="Non-selective", latitude=_at(0.6)),
+    )
+    assert select_nearby(frame, 100001) == []
+    assert 100001 not in {s["urn"] for s in select_nearby(frame, 100002)}
+
+
+def test_special_schools_match_only_each_other():
+    frame = _frame(
+        _row(100001, "Special", school_type="Community special school"),
+        _row(100002, "Mainstream A", latitude=_at(0.5)),
+        _row(100003, "Mainstream B", latitude=_at(0.6)),
+    )
+    assert select_nearby(frame, 100001) == []
+    assert select_nearby(frame, 100002) == []
+
+
+def test_boys_never_meets_girls():
+    frame = _frame(
+        _row(100001, "Boys School", gender="Boys"),
+        _row(100002, "Girls School", gender="Girls", latitude=_at(0.5)),
+        _row(100003, "Mixed School", gender="Mixed", latitude=_at(0.6)),
+        _row(100004, "Another Mixed", gender="Mixed", latitude=_at(0.7)),
+    )
+    urns = {s["urn"] for s in select_nearby(frame, 100001)}
+    assert 100002 not in urns
+    assert urns == {100003, 100004}
+
+
+def test_closed_schools_and_missing_coordinates_are_dropped():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "Closed", status="Closed", latitude=_at(0.5)),
+        _row(100003, "No coords", latitude=np.nan, longitude=np.nan),
+        _row(100004, "Good A", latitude=_at(0.6)),
+        _row(100005, "Good B", latitude=_at(0.7)),
+    )
+    assert {s["urn"] for s in select_nearby(frame, 100001)} == {100004, 100005}
+
+
+def test_all_through_is_offered_on_both_phase_sides():
+    frame = _frame(
+        _row(100001, "Primary subject", phase="Primary"),
+        _row(100002, "All through", phase="All-through", latitude=_at(0.5)),
+        _row(100003, "Primary peer", phase="Primary", latitude=_at(0.6)),
+    )
+    assert 100002 in {s["urn"] for s in select_nearby(frame, 100001)}
+
+    secondary = _frame(
+        _row(100010, "Secondary subject", phase="Secondary"),
+        _row(100002, "All through", phase="All-through", latitude=_at(0.5)),
+        _row(100011, "Secondary peer", phase="Secondary", latitude=_at(0.6)),
+    )
+    assert 100002 in {s["urn"] for s in select_nearby(secondary, 100010)}
+
+
+def test_sixteen_plus_is_matched_against_secondary_not_primary():
+    """GIAS phase 6 is "16 plus", and PHASE_GROUPS puts it in the secondary
+    group — a sixth-form college's peers are secondaries and other colleges,
+    never primary schools. A substring test for "secondary" misses it silently:
+    no crash, just a page offering infant schools to a sixth form."""
+    frame = _frame(
+        _row(100001, "Sixth Form College", phase="16 plus", age_range="16-19"),
+        _row(100002, "Nearby Secondary", phase="Secondary", latitude=_at(0.5),
+             attainment_8_score=52.0),
+        _row(100003, "Nearby College", phase="16 plus", latitude=_at(0.6)),
+        _row(100004, "Nearby Primary", phase="Primary", latitude=_at(0.1)),
+    )
+    result = select_nearby(frame, 100001)
+    urns = {s["urn"] for s in result}
+    assert 100004 not in urns, "a primary school is not a peer for a sixth form"
+    assert urns == {100002, 100003}
+    assert all(s["metric_key"] == "attainment_8_score" for s in result)
+
+
+def test_is_secondary_phase_agrees_with_the_phase_groups_it_selects_from():
+    for phase in ("Secondary", "Middle deemed secondary", "16 plus"):
+        assert is_secondary_phase(phase) is True, phase
+    for phase in ("Primary", "Middle deemed primary", "Nursery", "", None):
+        assert is_secondary_phase(phase) is False, phase
+    # In PHASE_GROUPS an all-through school is on both sides, but it renders
+    # with the primary template, and the metric follows the phase side.
+    assert is_secondary_phase("All-through") is False
+
+
+# ---------------------------------------------------------------------------
+# What the card reports
+# ---------------------------------------------------------------------------
+
+def test_shared_lists_only_what_is_actually_shared():
+    frame = _frame(
+        _row(100001, "Subject", phase="Secondary", gender="Mixed",
+             religious_denomination="None", admissions_policy="Non-selective"),
+        _row(100002, "Full match", phase="Secondary", gender="Mixed",
+             religious_denomination="None", admissions_policy="Non-selective", latitude=_at(0.5)),
+        _row(100003, "Faith differs", phase="Secondary", gender="Mixed",
+             religious_denomination="Church of England", admissions_policy="Non-selective", latitude=_at(0.6)),
+    )
+    by_urn = {s["urn"]: s for s in select_nearby(frame, 100001)}
+    assert by_urn[100002]["shared"] == ["Mixed", "Non-selective", "No religious character"]
+    assert by_urn[100003]["shared"] == ["Mixed", "Non-selective"]
+
+
+def test_a_shared_faith_is_named():
+    frame = _frame(
+        _row(100001, "Subject", religious_denomination="Roman Catholic"),
+        _row(100002, "Also RC", religious_denomination="Roman Catholic", latitude=_at(0.4)),
+        _row(100003, "Secular", religious_denomination="None", latitude=_at(0.5)),
+    )
+    by_urn = {s["urn"]: s for s in select_nearby(frame, 100001)}
+    assert "Roman Catholic" in by_urn[100002]["shared"]
+    assert by_urn[100003]["shared"] == ["Mixed"]
+
+
+def test_shared_is_empty_when_nothing_is_shared():
+    frame = _frame(
+        _row(100001, "Subject", gender="Boys", religious_denomination="Roman Catholic"),
+        _row(100002, "A", gender="Mixed", religious_denomination="None", latitude=_at(0.4)),
+        _row(100003, "B", gender="Mixed", religious_denomination="Church of England", latitude=_at(0.5)),
+    )
+    assert all(s["shared"] == [] for s in select_nearby(frame, 100001))
+
+
+def test_no_tier_is_reported_because_there_are_no_tiers():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "A", latitude=_at(0.4)),
+        _row(100003, "B", latitude=_at(0.5)),
+    )
+    assert all("tier" not in s for s in select_nearby(frame, 100001))
+
+
+def test_metric_follows_the_phase_side_not_the_neighbour():
+    frame = _frame(
+        _row(100001, "Subject", phase="Secondary", attainment_8_score=50.0),
+        _row(100002, "A", phase="Secondary", attainment_8_score=52.8, latitude=_at(0.5)),
+        _row(100003, "B", phase="Secondary", attainment_8_score=np.nan, latitude=_at(0.6)),
+    )
+    by_urn = {s["urn"]: s for s in select_nearby(frame, 100001)}
+    assert by_urn[100002]["metric_key"] == "attainment_8_score"
+    assert by_urn[100002]["metric_value"] == 52.8
+    assert by_urn[100002]["metric_year"] == 202425
+    assert by_urn[100003]["metric_value"] is None
+
+
+def test_values_are_json_safe_native_types():
+    frame = _frame(
+        _row(100001, "Subject"),
+        _row(100002, "A", latitude=_at(0.5)),
+        _row(100003, "B", latitude=_at(0.6)),
+    )
+    for school in select_nearby(frame, 100001):
+        assert isinstance(school["urn"], int)
+        assert isinstance(school["distance_miles"], float)
+        assert not isinstance(school["metric_value"], np.generic)
+
+
+# ---------------------------------------------------------------------------
+# The endpoint
+# ---------------------------------------------------------------------------
+
+import pytest
+from fastapi.testclient import TestClient
+
+
+def _endpoint_frame():
+    return _frame(
+        _row(100001, "Subject Primary"),
+        _row(100002, "Neighbour A", latitude=_at(0.5)),
+        _row(100003, "Neighbour B", latitude=_at(0.6)),
+    )
+
+
+@pytest.fixture()
+def client(monkeypatch):
+    from backend import app as app_module
+
+    monkeypatch.setattr(app_module, "load_latest_school_data", _endpoint_frame)
+    monkeypatch.setattr(app_module, "load_school_data", _endpoint_frame)
+    monkeypatch.setattr(app_module, "get_supplementary_data", lambda db, urn: {})
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_detail_payload_carries_nearby_schools(client):
+    resp = client.get("/api/schools/100001")
+    assert resp.status_code == 200, resp.text
+    similar = resp.json()["nearby_schools"]
+    assert [s["school_name"] for s in similar] == ["Neighbour A", "Neighbour B"]
+    assert similar[0]["metric_key"] == "rwm_expected_pct"
+
+
+def test_a_failure_in_selection_does_not_break_the_page(client, monkeypatch):
+    from backend import app as app_module
+
+    def _explode(*args, **kwargs):
+        raise ValueError("selection blew up")
+
+    monkeypatch.setattr(app_module, "select_nearby", _explode)
+    resp = client.get("/api/schools/100001")
+    assert resp.status_code == 200, resp.text
+    assert resp.json()["nearby_schools"] == []

backend/tests/test_publication.py

@@ -0,0 +1,68 @@
+"""Publication must preserve the current dataset until every replacement is ready."""
+import asyncio
+import pandas as pd
+import pytest
+from fastapi.testclient import TestClient
+from backend import app as api, data_loader
+from backend.tests.test_sixth_form_flag import _schools_df
+
+
+@pytest.fixture
+def client(monkeypatch):
+    old = _schools_df()
+    monkeypatch.setattr(data_loader, '_df_cache', old)
+    monkeypatch.setattr(data_loader, '_df_latest_cache', old)
+    monkeypatch.setattr(api, '_place_registry', {'old': 'registry'})
+    monkeypatch.setattr(api, '_place_index', {'old': 'index'})
+    monkeypatch.setattr(api, '_place_index_source', api._place_registry)
+    monkeypatch.setattr(api, '_sitemaps', {'old.xml': 'old sitemap'})
+    monkeypatch.setattr(api, '_publication_lock', asyncio.Lock())
+    monkeypatch.setattr(api.limiter, 'enabled', False)
+    api.app.dependency_overrides[api.verify_admin_api_key] = lambda: True
+    yield TestClient(api.app, raise_server_exceptions=False)
+    api.app.dependency_overrides.clear()
+
+
+def state():
+    return (data_loader._df_cache, data_loader._df_latest_cache, api._place_registry,
+            api._place_index, api._place_index_source, api._sitemaps)
+
+
+@pytest.mark.parametrize('failure', ['empty', 'database', 'sitemap', 'duplicate'])
+def test_failed_reload_preserves_every_published_object(client, monkeypatch, failure):
+    before = state()
+    df = _schools_df()
+    if failure == 'empty':
+        df = pd.DataFrame()
+    if failure == 'duplicate':
+        df = pd.concat([df, df.iloc[:1]], ignore_index=True)
+    def load():
+        if failure == 'database':
+            raise RuntimeError('database unavailable')
+        return df
+    monkeypatch.setattr(api, 'load_school_data_as_dataframe', load)
+    if failure == 'sitemap':
+        monkeypatch.setattr(api, 'build_sitemaps', lambda *args: (_ for _ in ()).throw(RuntimeError('bad XML')))
+    response = client.post('/api/admin/reload')
+    assert response.status_code == 503
+    assert all(a is b for a, b in zip(before, state()))
+
+
+def test_success_publishes_school_data_places_and_sitemaps(client, monkeypatch):
+    df = _schools_df()
+    df.loc[0, 'school_name'] = 'Replacement School'
+    monkeypatch.setattr(api, 'load_school_data_as_dataframe', lambda: df)
+    response = client.post('/api/admin/reload')
+    assert response.status_code == 200
+    assert data_loader.load_school_data() is df
+    assert data_loader.load_latest_school_data().iloc[0].school_name == 'Replacement School'
+    assert api._place_index_source is api._place_registry
+    assert 'old.xml' not in api._sitemaps
+    assert 'replacement-school' in api._sitemaps['schools-1.xml']
+
+
+def test_failed_sitemap_regeneration_keeps_existing_publication(client, monkeypatch):
+    before = state()
+    monkeypatch.setattr(api, 'build_sitemaps', lambda *args: (_ for _ in ()).throw(RuntimeError('bad XML')))
+    assert client.post('/api/admin/regenerate-sitemap').status_code == 503
+    assert all(a is b for a, b in zip(before, state()))

backend/tests/test_search_completeness.py

@@ -0,0 +1,114 @@
+from types import SimpleNamespace
+import pytest
+from fastapi.testclient import TestClient
+from backend import app as api, data_loader
+from backend.tests.test_sixth_form_flag import _schools_df
+
+
+def client_for(monkeypatch, search):
+    client = SimpleNamespace(collections={'schools': SimpleNamespace(documents=SimpleNamespace(search=search))})
+    monkeypatch.setattr(data_loader, '_get_typesense_client', lambda: client)
+
+
+def test_search_returns_matches_beyond_first_page(monkeypatch):
+    pages = []
+    def search(params):
+        pages.append(params['page'])
+        urns = range(100000, 100250) if params['page'] == 1 else [100999]
+        return {'found': 251, 'hits': [{'document': {'urn': u}} for u in urns]}
+    client_for(monkeypatch, search)
+    result = data_loader.search_schools_typesense('academy')
+    assert len(result) == 251
+    assert result[-1] == 100999
+    assert pages == [1, 2]
+
+
+def test_search_caps_broad_queries_at_a_bounded_number_of_pages(monkeypatch):
+    requests = []
+
+    def search(params):
+        requests.append(params)
+        return {
+            'found': 10_000,
+            'hits': [
+                {'document': {'urn': 100000 + params['page'] * 1000 + i}}
+                for i in range(params['per_page'])
+            ],
+        }
+
+    client_for(monkeypatch, search)
+    result = data_loader.search_schools_typesense('school')
+
+    assert len(result) == data_loader.SEARCH_MAX_CANDIDATES
+    assert len(requests) == data_loader.SEARCH_MAX_CANDIDATES // data_loader.SEARCH_PAGE_SIZE
+    assert all(request['per_page'] == data_loader.SEARCH_PAGE_SIZE for request in requests)
+    assert requests[-1]['page'] == len(requests)
+
+
+def test_search_uses_a_smaller_final_page_when_the_cap_is_not_a_page_multiple(monkeypatch):
+    monkeypatch.setattr(data_loader, 'SEARCH_MAX_CANDIDATES', 251)
+    requests = []
+
+    def search(params):
+        requests.append(params)
+        return {
+            'found': 10_000,
+            'hits': [{'document': {'urn': 100000 + len(requests) * 1000 + i}}
+                     for i in range(params['per_page'])],
+        }
+
+    client_for(monkeypatch, search)
+    result = data_loader.search_schools_typesense('school')
+
+    assert len(result) == 251
+    assert [request['per_page'] for request in requests] == [250, 1]
+
+
+def test_later_page_failure_does_not_return_partial_results(monkeypatch):
+    def search(params):
+        if params['page'] == 2:
+            raise RuntimeError('timeout')
+        return {'found': 251, 'hits': [{'document': {'urn': u}} for u in range(100000, 100250)]}
+    client_for(monkeypatch, search)
+    assert data_loader.search_schools_typesense('academy') is None
+
+
+def test_zero_matches_are_distinct_from_unavailable(monkeypatch):
+    client_for(monkeypatch, lambda _: {'found': 0, 'hits': []})
+    assert data_loader.search_schools_typesense('academy') == []
+    monkeypatch.setattr(data_loader, '_get_typesense_client', lambda: None)
+    assert data_loader.search_schools_typesense('academy') is None
+
+
+@pytest.mark.parametrize('matches, expected', [([], []), (None, [100001])])
+def test_fallback_only_on_dependency_failure(monkeypatch, matches, expected):
+    monkeypatch.setattr(api.limiter, 'enabled', False)
+    monkeypatch.setattr(api, 'load_latest_school_data', _schools_df)
+    monkeypatch.setattr(api, 'search_schools_typesense', lambda _: matches)
+    response = TestClient(api.app).get('/api/schools?search=Alpha')
+    assert response.status_code == 200
+    assert [s['urn'] for s in response.json()['schools']] == expected
+
+
+def test_filtered_api_keeps_match_from_second_search_page(monkeypatch):
+    monkeypatch.setattr(api.limiter, 'enabled', False)
+    df = _schools_df()
+    monkeypatch.setattr(api, 'load_latest_school_data', lambda: df)
+    def search(params):
+        urns = range(200000, 200250) if params['page'] == 1 else [100001]
+        return {'found': 251, 'hits': [{'document': {'urn': u}} for u in urns]}
+    client_for(monkeypatch, search)
+    response = TestClient(api.app).get('/api/schools?search=Alpha&local_authority=Testshire')
+    assert response.status_code == 200
+    assert response.json()['total'] == 1
+    assert response.json()['schools'][0]['urn'] == 100001
+
+
+def test_unavailable_dataset_is_not_a_missing_school_or_empty_search(monkeypatch):
+    import pandas as pd
+    monkeypatch.setattr(api.limiter, 'enabled', False)
+    monkeypatch.setattr(api, 'load_school_data', lambda: pd.DataFrame())
+    monkeypatch.setattr(api, 'load_latest_school_data', lambda: pd.DataFrame())
+    client = TestClient(api.app)
+    assert client.get('/api/schools/100001').status_code == 503
+    assert client.get('/api/schools?search=school').status_code == 503

backend/version.py

@@ -1,26 +0,0 @@
-"""
-Schema versioning for database migrations.
-
-HOW TO USE:
-- Bump SCHEMA_VERSION when making changes to database models
-- This triggers an automatic full data reimport on next app startup
-
-WHEN TO BUMP:
-- Adding/removing columns in models.py
-- Changing column types or constraints
-- Modifying CSV column mappings in schemas.py
-- Any change that requires fresh data import
-"""
-
-# Current schema version - increment when models change
-SCHEMA_VERSION = 6
-
-# Changelog for documentation
-SCHEMA_CHANGELOG = {
-    1: "Initial schema with School and SchoolResult tables",
-    2: "Added pupil absence fields (reading, maths, gps, writing, science)",
-    3: "Added supplementary data tables: ofsted, parent_view, census, admissions, sen_detail, phonics, deprivation, finance; GIAS columns on schools",
-    4: "Added Ofsted Report Card columns to ofsted_inspections (new framework from Nov 2025)",
-    5: "Apply ALTER TABLE additions for RC columns missed by create_all on existing tables",
-    6: "Removed the Ofsted Parent View feature: dropped fact_parent_view table and model",
-}

claude.md

@@ -1,180 +1,37 @@
-# SchoolCompare.co.uk - Project Context
+# SchoolCompare project context
 
-## Overview
+## Maintained documentation
 
-SchoolCompare is a web application for comparing UK primary school (KS2) performance data. It allows users to:
-- Search and browse schools by name, location (postcode), or local authority
-- Compare multiple schools side-by-side with charts and tables
-- View school rankings by various KS2 metrics
-- See historical performance trends across years
+Read [README.md](README.md), [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) and
+[docs/DEVELOPMENT.md](docs/DEVELOPMENT.md) for the current implementation.
+[docs/LEGACY_CODE.md](docs/LEGACY_CODE.md) records obsolete paths and deliberate
+compatibility code. Historical design documents are not current setup instructions.
 
-## Architecture
+## Architecture constraints
 
-### Backend (Python/FastAPI)
-- **Framework**: FastAPI with uvicorn
-- **Database**: PostgreSQL with SQLAlchemy ORM
-- **Data Source**: UK Government "Compare School Performance" CSV downloads
-
-Key files:
-- `backend/app.py` - Main FastAPI application, API routes
-- `backend/config.py` - Configuration via pydantic-settings (env vars, .env file)
-- `backend/database.py` - SQLAlchemy engine, session management
-- `backend/models.py` - Database models (School, SchoolResult)
-- `backend/data_loader.py` - Data queries, geocoding, legacy DataFrame compatibility
-- `backend/schemas.py` - Column mappings, metric definitions, LA code mappings
-
-### Content / CMS (Payload)
-
-Payload CMS runs **inside** the Next.js app — one image, one container, no
-separate service. It powers `/blog`; `/about` is a plain coded page.
-
-- **Admin panel:** `/admin`. The only authenticated surface on the site.
-  `noindex` via both `robots.txt` and `X-Robots-Tag`.
-- **CMS API:** `/cms-api`, **not** `/api`. `/api/*` is a catch-all proxy to
-  FastAPI (`app/(frontend)/api/[...path]`) which would silently swallow every
-  admin call and forward it to the backend. Mount points are defined once in
-  `lib/payloadRoutes.ts`.
-- **Database:** the existing Postgres, in its own `payload` schema, so no
-  pipeline operation on `public` — including
-  `scripts/migrate_csv_to_db.py --drop` — can reach blog content.
-- **Uploads:** the `payload_media` Docker volume at `/app/media`. Not
-  reproducible from the pipeline; must be backed up.
-- **New env vars:** `DATABASE_URL` and `PAYLOAD_SECRET` on the frontend service.
-  Staging must use a different `PAYLOAD_SECRET` from production.
-- Publishing workflow and house style: `nextjs-app/docs/PUBLISHING.md`.
-- **Admin field components resolve through a generated import map**
-  (`app/(payload)/admin/importMap.js`). Payload hands the client a *path* per
-  field and looks it up there; a missing entry renders no field and reports no
-  error, while `required` still blocks the save. After adding or changing any
-  field, editor or lexical feature, run `npm run generate:importmap` in
-  `nextjs-app/` and commit the result.
-
-### Two route groups
-
-`nextjs-app/app/` has no root `layout.tsx`. It cannot: Payload's admin panel
-ships its own root layout rendering `<html>`/`<body>`, and Next permits
-multiple root layouts only when no `app/layout.tsx` exists.
-
-- `app/(frontend)/` — the site. Its `layout.tsx` is the site's root layout.
-- `app/(payload)/` — the admin panel and `/cms-api`.
-
-Route groups are invisible to routing, so every public URL is unchanged.
-
-**The metadata file conventions stay at the `app/` root** — `robots.ts`,
-`opengraph-image.tsx`, `icon.png`, `apple-icon.png`. Inside a route group Next
-treats them as segment-scoped: it renames `/icon.png` to `/icon-<hash>.png` and
-drops `/robots.txt` entirely. Route handlers are unaffected.
-
-The build must succeed with `DATABASE_URL` unset, because CI builds it that
-way. Never call `getCachedPayload()` at module scope, and never add
-`generateStaticParams` to a DB-backed route.
-
-### Frontend (Vanilla JS)
-- Single-page application with hash-based routing
-- Chart.js for data visualization
-- No build step required
-
-Key files:
-- `frontend/index.html` - Main HTML structure
-- `frontend/app.js` - All application logic, API calls, rendering
-- `frontend/styles.css` - Styling (CSS variables, responsive design)
-
-### Database Schema
-
-```
-schools                          school_results
-├── id (PK)                      ├── id (PK)
-├── urn (unique, indexed)        ├── school_id (FK → schools.id)
-├── school_name                  ├── year (indexed)
-├── local_authority              ├── rwm_expected_pct
-├── school_type                  ├── reading_expected_pct
-├── postcode                     ├── ... (all KS2 metrics)
-├── latitude, longitude          └── unique(school_id, year)
-└── results → SchoolResult[]
-```
-
-## Configuration
-
-Environment variables (or `.env` file):
-- `DATABASE_URL` - PostgreSQL connection string (default: `postgresql://schoolcompare:schoolcompare@localhost:5432/schoolcompare`)
-- `HOST`, `PORT` - Server binding (default: `0.0.0.0:80`)
-- `ALLOWED_ORIGINS` - CORS origins
-
-## Running Locally
-
-1. Start PostgreSQL:
-   ```bash
-   docker compose up -d db
-   ```
-
-2. Run migration to import CSV data:
-   ```bash
-   python scripts/migrate_csv_to_db.py --drop
-   # Add --geocode to geocode postcodes (slower, adds lat/long)
-   ```
-
-3. Start the app:
-   ```bash
-   uvicorn backend.app:app --host 0.0.0.0 --port 8000
-   ```
-
-## Docker Deployment
-
-```bash
-docker compose up -d
-```
-
-This starts:
-- `db` - PostgreSQL 16 with persistent volume
-- `app` - FastAPI application on port 80
-
-## Data
-
-- Source: UK Government Compare School Performance downloads
-- Location: `data/` directory with year folders (e.g., `2023-2024/england_ks2final.csv`)
-- The `scripts/download_data.py` can fetch data from the government website
-
-## Key Features
-
-- **Location Search**: Enter postcode to find nearby schools (uses postcodes.io API)
-- **Multi-school Comparison**: Select multiple schools, view metrics across years
-- **Rankings**: Top schools by any KS2 metric, filterable by local authority
-- **Variability Analysis**: Shows standard deviation of scores across years
-
-## API Endpoints
-
-- `GET /api/schools` - List/search schools (supports pagination, location search)
-- `GET /api/schools/{urn}` - School details with all yearly data
-- `GET /api/compare?urns=123,456` - Compare multiple schools
-- `GET /api/rankings` - School rankings by metric
-- `GET /api/filters` - Available filter options (LAs, types, years)
-- `GET /api/metrics` - Metric definitions (single source of truth)
-- `GET /api/data-info` - Database stats
+- Next.js serves the public UI. FastAPI serves school data from dbt-built `marts.*`.
+  The backend does not create school tables or import CSVs at startup.
+- School coverage spans England and multiple phases, not only primary schools in
+  Wandsworth and Merton.
+- `/api/*` belongs to the FastAPI proxy. Payload uses `/cms-api` and `/admin`.
+- Payload runs inside Next.js, with its own `payload` schema and persistent media.
+  Keep CMS migrations independent of school-data transformations.
+- Public and Payload route groups have separate root layouts. Do not introduce
+  `app/layout.tsx`. Keep site-wide metadata files at the `app/` root.
+- Builds must succeed with `DATABASE_URL` unset. Do not call `getCachedPayload()`
+  at module scope or add DB-backed `generateStaticParams`.
+- After changing CMS fields/editors, run `npm run generate:importmap` and commit
+  the generated import map. See `nextjs-app/docs/PUBLISHING.md`.
+- The backend and pipeline GIAS dictionary copies are generated together; preserve
+  their parity. Tests enforce it.
 
 ## SDLC
 
-Full details in `docs/DEPLOY.md`. The short version:
+Follow [docs/DEPLOY.md](docs/DEPLOY.md).
 
-- **Never push to `main` directly.** Work on a feature branch and open a PR;
-  branch protection requires the PR checks (typecheck, tests, builds, AI review)
-  to pass before merge.
-- Merging to `main` deploys automatically **to staging only**: images are
-  built once, deployed to the staging Portainer stack, and verified by the
-  Playwright journeys in `e2e/`. Production is a second, manual approval:
-  the "Promote to Production (manual)" workflow in Gitea Actions, run after
-  testing the feature on staging. It refuses commits whose staging E2E gate
-  isn't green. Never trigger it yourself — promotion is the human's call.
-- If you change user-facing behaviour, update or extend the `e2e/` journey
-  tests in the same PR — they gate whether staging is fit for human testing
-  and whether a commit is promotable.
-
-## Recent Changes
-
-- Added staging environment + automated staging→prod pipeline (Gitea Actions)
-- Migrated from CSV file storage to PostgreSQL database
-- Added location-based search using postcode geocoding
-- Added local authority filter to rankings
-- Improved frontend with featured schools, loading states, API caching
-
-# Important
-- Do not attempt to start a local server to test the application, it does not work
+- Never push directly to `main`. Use a feature branch and a PR with passing checks.
+- Merges deploy staging only. Production promotion is a separate human decision;
+  do not trigger the promotion workflow yourself.
+- Update E2E journeys in the same PR when changing user-facing behaviour.
+- Do not attempt to start a local server to test the application; use unit checks
+  and the configured integration environment.

docs/ARCHITECTURE.md

@@ -0,0 +1,125 @@
+# Architecture
+
+This describes the implementation as reviewed on 2026-09-14. It distinguishes
+current behaviour from improvements still to be implemented.
+
+## Request flow
+
+```text
+Browser → Next.js public routes
+          ├─ /api/* proxy → FastAPI → cached DataFrames / PostgreSQL marts
+          │                         ├─ Typesense (search and suggestions)
+          │                         └─ postcodes.io (postcode lookup)
+          └─ /admin, /cms-api, /blog → Payload → payload schema + media volume
+
+Next.js server rendering → FastAPI directly through FASTAPI_URL
+```
+
+`nextjs-app/lib/api.ts` contains typed fetch wrappers and revalidation defaults.
+The proxy is `nextjs-app/app/(frontend)/api/[...path]/route.ts`. Payload uses
+`/cms-api` so its routes do not collide with the FastAPI proxy. The proxy denies
+`/api/flags`; server-side rendering reads flags directly from FastAPI.
+
+## Data ownership
+
+| Layer | Owner and role |
+|---|---|
+| Source data | GIAS, DfE EES, Ofsted, finance, deprivation and council admission-distance sources |
+| `raw` | Singer taps and the PostgreSQL target configured in `pipeline/meltano.yml` |
+| Staging/intermediate/marts | dbt models in `pipeline/transform`; marts are materialized tables |
+| `marts.dim_school`, `marts.dim_location` | School identity and location, filtered to supported England establishments |
+| `marts.fact_*` | Performance and supplementary datasets; coverage and years vary |
+| Typesense `schools` alias | Search documents built by `pipeline/scripts/sync_typesense.py` |
+| `payload` | CMS collections and migrations in `nextjs-app/`; independent of dbt |
+| Media volume | Uploaded blog media; requires backup and cannot be regenerated from school datasets |
+
+`backend/models.py` maps existing marts for reading. It does not create the school
+schema. There is no startup schema-version migration or CSV reimport. Payload's
+`nextjs-app/migrations/` is active and must not be confused with the removed
+legacy backend migration code.
+
+Coordinates normally come from GIAS British National Grid coordinates transformed
+by PostGIS in `dim_location.sql`. `pipeline/scripts/geocode_postcodes.py` is a
+manual fallback utility, not a task wired into the current school-data DAG.
+Backend postcode searches also use postcodes.io; that lookup does not populate
+school coordinates in the database.
+
+## Backend boundaries
+
+- `app.py`: routes, middleware, search filtering, sitemap/place publication and response assembly.
+- `data_loader.py`: SQL loading, process-local DataFrame caches, Typesense calls,
+  postcode lookups, supplementary queries and benchmark calculation.
+- `database.py`: synchronous SQLAlchemy engine and sessions.
+- `schemas.py`: metric definitions, column mappings and display metadata; despite
+  its name this is not a collection of Pydantic API response models.
+- `places.py` and `localities.py`: place registry and curated locality information.
+- `flags.py`: Unleash-backed feature flags, disabled when no server is configured.
+- `gias_codes.py` / `ofsted_codes.py`: source-code translation and display rules.
+
+Search starts from a cached latest-row-per-school snapshot. Detail pages read
+history from the full DataFrame and supplementary data from marts. Comparisons
+batch supplementary queries across selected URNs. Async routes still contain
+synchronous dependency calls; a fully asynchronous database layer is not present.
+
+## Frontend boundaries
+
+`app/(frontend)` owns the public root layout and pages. `app/(payload)` owns the
+CMS root layout. Do not add a shared `app/layout.tsx`: these groups deliberately
+have separate root layouts. Root metadata files remain in `app/`.
+
+Server pages fetch initial data and pass it to client views. Client state uses
+React hooks, URL search parameters and the comparison context/localStorage.
+There is no SWR dependency. Leaflet maps are loaded through dynamic wrappers;
+Chart.js renders performance and comparison charts.
+
+`components/school/` contains detail sections, with section decisions and data
+preparation in `lib/schoolSections.ts`. The nearby-schools section is selected in
+`backend/nearby_schools.py` — hard filters decide eligibility (phase, provision,
+selectivity, gender) and distance alone decides the order, capped per phase —
+and served on `/api/schools/{urn}`. Its rules are presentation logic,
+deliberately kept out of `marts.*` so they can be tuned by deploy rather than by
+pipeline run. `lib/types.ts` contains manually maintained
+API types. `payload-types.ts` and the Payload import map are generated artifacts.
+
+## Publication and caching today
+
+1. Airflow DAGs extract and validate source data, then run selected dbt builds.
+2. Relevant DAGs rebuild Typesense and swap the `schools` alias.
+3. They call `POST /api/admin/reload` with `X-API-Key`. It builds and validates
+   replacement DataFrames, places, reverse membership and sitemaps off the request
+   loop, then publishes them together. Failure returns 503 and preserves live data.
+4. A separate weekly sitemap DAG can regenerate the derived publication from the
+   current DataFrame without clearing the live registry first.
+
+GIAS is scheduled daily, Ofsted monthly, and annual datasets are manually
+triggered. The DAG definitions are authoritative for selectors and dependencies.
+
+Caches exist in several independent layers: backend DataFrames and registries,
+backend HTTP Cache-Control/ETags, Next.js fetch/page revalidation, and browser or
+shared HTTP caches where configured. Place fetches request a one-week revalidation
+interval. HTTP ETags are computed after route execution, not before database work.
+
+Typesense publication validates every import response and the final document
+count before switching aliases. A session-scoped PostgreSQL advisory lock
+serialises index reads/publication across DAGs. The previous collection remains
+available for rollback; old unaliased collections are pruned after success.
+Failed drafts are retained until a later successful cleanup, because an uncertain
+alias-update response must never cause deletion of a potentially live index.
+
+The backend snapshot swap is process-local and assumes the current single-worker
+deployment. It is not an atomic transaction spanning PostgreSQL marts, Typesense
+and Next.js caches. Next.js caches are not explicitly purged by the pipeline.
+School search retrieves a relevance-ordered candidate prefix (currently capped at
+1,000 URNs) before applying API filters. This keeps scoped searches useful while
+putting a hard ceiling on Typesense round trips; only a dependency failure invokes
+substring fallback, not a valid empty match set.
+
+## Deployment references
+
+See [DEPLOY.md](DEPLOY.md). PR checks include frontend typechecking/tests, backend
+unit tests, image builds and AI review. Staging journeys run after merging.
+Staging runs are serialised across builds, deployment and E2E. Build-stamped
+frontend/backend identities are checked before and after journeys. Only then are
+the captured image digests marked verified. Promotion resolves and validates the
+complete verified image set before retagging production. See the runbook for
+first-rollout requirements and remaining integration checks.

docs/DEPLOY.md

@@ -19,8 +19,9 @@ PR checks (.gitea/workflows/pr-checks.yml)
    ▼
 Stage pipeline (.gitea/workflows/deploy.yml) — automatic
    1. build & push images        → tags sha-<sha>, staging
-   2. staging Portainer webhook  → wait for staging health
+   2. staging Portainer webhook  → verify frontend/backend SHA + build ID
    3. Playwright E2E journeys against staging      ← gate before human testing
+   4. verify identity again; tag tested digests verified-<full-sha>
    ▼
 Manual testing on staging (stx.schoolcompare.co.uk)
    │  Actions → "Promote to Production (manual)"   ← approval #2
@@ -28,14 +29,15 @@ Manual testing on staging (stx.schoolcompare.co.uk)
 Promote pipeline (.gitea/workflows/promote.yml) — manual dispatch
    1. resolve target sha (input, or latest main if empty)
    2. REFUSE unless that commit's "E2E Journeys against Staging" status is green
-   3. retag sha-<sha> → :prod    (same bytes — build once, promote the image)
+   3. resolve verified-<full-sha> digests, validate labels, retag digests → :prod
       previous :prod saved as :prod-previous
-   4. prod Portainer webhook     → wait for prod health
+   4. prod Portainer webhook     → verify expected SHA + build ID
 ```
 
 Key principle: **build once, promote the exact image**. Production pins `:prod`,
 which only moves when a human runs the promote workflow — and the workflow
-only accepts commits that passed the staging E2E gate. Nothing tags `:latest`
+only accepts commits that passed the staging E2E gate and have a complete verified
+image set. Nothing tags `:latest`
 anymore.
 
 ## Branch & PR workflow
@@ -256,3 +258,50 @@ how long any feature is exposed to this.
 If `UNLEASH_URL` is unset, every flag is `False` and no connection is
 attempted. That is the correct behaviour for local development and CI, and it
 means the test suites need no flag server.
+
+## Release identity and the P1 reliability gate
+
+Every staging run creates a random build ID before building its three images.
+Each image carries the commit and build ID as labels. Frontend/backend images
+also contain a build-time JSON file; environment overrides cannot rewrite it.
+`/release.json` returns both identities with `Cache-Control: no-store`. It fails
+with 503 when either identity cannot be read. FastAPI's internal endpoint is
+`/api/release`.
+
+The entire staging workflow shares one concurrency group, with cancellation
+disabled. This needs Gitea 1.26 or newer, where workflow concurrency is supported
+([release notes](https://blog.gitea.com/release-of-1.26.0/)); the configured server
+reported 1.27.3 during this change. Do not run the workflow on an older server
+that ignores the concurrency key. Manual deployments outside this workflow must
+also avoid changing staging during journeys.
+
+The gate checks both identities before and after Playwright. It then validates
+labels on the captured build output digests and tags them `verified-<full-sha>`.
+The manual promotion script resolves all three verified tags to immutable digests
+and confirms one matching commit/build ID before moving any `:prod` tag. It polls
+production for that same identity using a locally saved release manifest.
+A registry error can still interrupt the three tag writes; the Portainer webhook
+only runs after successful promotion, and rerunning promotion resolves the full
+verified set again. There is no cross-registry atomic tag transaction.
+
+**First rollout:** old green commits without verified tags/build identities are
+not promotable through this gate. Build and test a commit containing the new
+workflow first. The release route must be reachable through the configured
+`STAGING_BASE_URL`/`PROD_BASE_URL`; it deliberately avoids the public staging
+`/api` proxy limitation. No new deployment secret is required.
+
+`scripts/ci/release.py` implements identity polling and digest verification.
+The poller identifies itself as `SchoolCompare-Release-Check/1.0`: the public
+staging proxy has returned HTTP 403 to Python's default urllib user agent even
+while the release endpoint was healthy. It logs changes in HTTP/connection
+failures or observed release identities, and includes the last observation in
+the timeout error. If verification fails, use that observation to distinguish
+proxy rejection (403), an unavailable release endpoint (503), and containers
+still reporting an older SHA/build ID. Check the configured base URL from the
+CI runner; a successful request from another machine does not establish runner
+connectivity. Do not bypass identity verification to unblock a deployment.
+Its mocked tests run in PR checks alongside backend and index-publication tests.
+The new Playwright journeys also check deployed identity and stale pagination.
+Local unit checks do not validate registry credentials, Portainer behaviour,
+proxy routing or a deployed image; those require the staging run. Production
+promotion remains a separate human action.

docs/DEVELOPMENT.md

@@ -0,0 +1,94 @@
+# Development and validation
+
+## Prerequisites and environment boundaries
+
+Use a feature branch. The deployed stack is the integration environment; do not
+assume a local server can run from a fresh checkout. This cleanup did not start
+local servers or provision databases. Unit tests use fixtures and mocks.
+
+The current versions are not yet aligned:
+
+| Component | Container | PR checks |
+|---|---|---|
+| Backend | Python 3.11 | Python 3.12 |
+| Frontend | Node 24 | Node 22 |
+| Pipeline | Python 3.13 | Pipeline image build |
+
+Use the component's container version when reproducing deployment behaviour.
+The backend dependency pins predate Python 3.14; do not assume the system Python
+can install or run them. Version alignment is a separate maintenance task.
+
+## Frontend checks
+
+```sh
+cd nextjs-app
+npm ci
+npm run typecheck
+npm test -- --runInBand
+```
+
+`npm run build` is the production build check. There is no `lint` script.
+Tests live in `__tests__/` and use Jest/React Testing Library. These checks do not
+prove that live PostgreSQL queries, Typesense or a deployed proxy work.
+
+The frontend `.env.example` documents runtime variables. Browser traffic normally
+uses `/api`; `FASTAPI_URL` is an absolute server-side URL ending in `/api`.
+Payload additionally needs `DATABASE_URL` and `PAYLOAD_SECRET` when used at runtime.
+Never commit credentials or real `.env` files.
+
+## Backend checks
+
+From the repository root, using an available Python 3.11 or 3.12 interpreter:
+
+```sh
+python3.11 -m venv /tmp/schoolcompare-backend-venv
+/tmp/schoolcompare-backend-venv/bin/python -m pip install -r requirements.txt pytest 'httpx<0.28' pyyaml
+/tmp/schoolcompare-backend-venv/bin/python -m pytest backend/tests pipeline/tests scripts/ci/tests -q
+```
+
+Substitute `python3.12` if matching PR CI. The test dependencies above match the
+current workflow; they are not yet captured in a dedicated development lockfile.
+Backend configuration is defined in `backend/config.py`; `.env.example` documents
+commonly used values. `ALLOWED_ORIGINS` uses a JSON array, not a comma-separated string.
+
+## Data and pipeline work
+
+The app needs populated `marts.*` tables. A new Postgres instance alone is not a
+working school-data environment. Use the existing managed pipeline or an approved
+snapshot; the removed CSV importer cannot build the current schema.
+
+The pipeline container includes Meltano, dbt/Postgres, Airflow and the custom taps.
+Airflow commands/selectors live in `pipeline/dags/`. Schema tests live in
+`pipeline/transform/tests/` and model YAML files. Run the relevant `dbt build`
+selector in an isolated data environment for model changes; it writes tables and
+is not a read-only smoke test. Prefer `python -m dbt.cli.main` as the DAGs do.
+
+GIAS dictionaries are generated together by
+`pipeline/scripts/generate_gias_codes.py`. The backend and pipeline copies are
+intentional; `backend/tests/test_gias_codes.py` checks that they stay identical.
+
+For Payload collection/editor changes, run `npm run generate:importmap` in
+`nextjs-app/` and include the generated map. Preserve CMS migrations and the
+separate `payload` schema. See [publishing](../nextjs-app/docs/PUBLISHING.md).
+
+## End-to-end checks
+
+Against an existing, authorised test environment:
+
+```sh
+cd e2e
+npm ci
+npx playwright install chromium
+BASE_URL=https://your-test-environment.example npx playwright test
+```
+
+The suite does not start a web server. CI installs Chromium with system dependencies
+and runs against staging. Use the configured staging target: `docs/DEPLOY.md`
+records the public staging proxy limitation. User-visible behaviour changes should
+update the corresponding journeys.
+
+## Before requesting review
+
+Run checks relevant to the change, inspect `git diff --check`, and report checks
+that could not run. Do not publish or promote as part of local validation.
+[DEPLOY.md](DEPLOY.md) documents the PR and human promotion gates.

docs/LEGACY_CODE.md

@@ -0,0 +1,89 @@
+# Legacy and unused-code inventory
+
+Reviewed 2026-09-14. This inventory records source evidence, not production usage
+telemetry. A command with no repository caller may still be run manually or from
+an external scheduler. Historical specs and prototypes are not runtime imports.
+
+## Method and scope
+
+Searched backend imports, tests, CLI scripts, Airflow DAGs, Meltano configuration,
+Gitea workflows, Dockerfiles and documentation. For frontend candidates, inspected
+TypeScript imports, re-exports, literal dynamic imports and `require` calls,
+resolving relative and `@/` paths while excluding tests, dependencies and build
+output. Checked candidates again with text searches including tests.
+
+Next.js route files, generated Payload import-map entries and plugin discovery
+are entry points even without ordinary imports. This is why a zero-import count
+alone is not sufficient grounds for deletion. Computed imports and external
+operators are outside this static audit.
+
+## Removed in this cleanup
+
+These names are recorded for Git-history lookup; they are no longer file links.
+
+| Removed path or symbol | Evidence and replacement |
+|---|---|
+| `backend/migration.py` | Imported `School` and `SchoolResult`, which no longer exist in `backend/models.py`. Only the legacy CSV CLI imported it. Current tables are built by dbt. |
+| `backend/version.py` | Only the legacy importer consumed `SCHEMA_VERSION`. FastAPI lifespan does not perform version-triggered imports. This is unrelated to active Payload migrations. |
+| `scripts/migrate_csv_to_db.py` | Imported removed `init_db`/`set_db_schema_version` helpers and the obsolete models indirectly. No runtime, DAG or workflow calls it. Use the managed pipeline for current marts. |
+| `scripts/geocode_schools.py` | Imported the removed `School` ORM model. No pipeline/workflow calls it. Coordinates now come from GIAS/PostGIS; a separate mart-aware manual utility remains under `pipeline/scripts/`. |
+| `backend.data_loader.haversine_distance` | No callers. Search uses its inline vectorised NumPy calculation. |
+| `nextjs-app/lib/api.ts: fetcher` | No callers; SWR is not installed. Application fetches use the named API wrappers. |
+| `nextjs-app/lib/api.ts: kmToMiles` | No callers. `calculateDistance` remains because `CutoffMapPanel` uses it. |
+
+The removed command files could not import successfully against the current
+backend. This cleanup does not run replacements, migrate data or modify databases.
+Their previous implementations remain recoverable from Git history.
+
+## Unused candidates retained for a separate cleanup
+
+| Candidate | Evidence | Recommended next step |
+|---|---|---|
+| `nextjs-app/components/LoadingSkeleton.tsx` and its CSS | No application or test imports found. | Remove together after confirming no planned use. |
+| `nextjs-app/components/Pagination.tsx` and its CSS | No application or test imports found; HomeView implements load-more behaviour. | Remove as a pair if numbered pagination will not return. |
+| `nextjs-app/components/SchoolCard.tsx` and its CSS | Imported by its own tests, not application code. HomeView uses SchoolRow/SecondarySchoolRow. | Decide whether to retire the card design; if removed, remove its dedicated tests as well. Passing tests do not establish runtime use. |
+| `backend/database.py: get_db`, `get_db_session` | No remaining callers after removing the importer. Current code creates SessionLocal directly. | Either adopt these helpers during session-lifecycle cleanup or remove them; do not rewrite active sessions in a documentation change. |
+| `backend/schemas.py: COLUMN_MAPPINGS`, `NULL_VALUES`, `LA_CODE_TO_NAME` | No remaining Python consumers found after importer removal. Other constants in this module are active. | Remove individual constants after checking external data utilities; retain the module. |
+| `backend/config.py: data_dir`, `max_page_size`, `rate_limit_burst` | No active consumers found. `default_page_size` appears only in a branch that expects None, although the route supplies a concrete default. | Reconcile settings with route validation in a focused API change. |
+
+## Legacy/manual paths requiring operational verification
+
+| Path | Status and reason to retain for now |
+|---|---|
+| FastAPI `/`, `/compare`, `/rankings`, `/favicon.svg`, `/robots.txt`, and conditional `/static` | Old frontend-serving routes reference a `frontend/` directory absent from the checkout and backend image. Next.js owns these public surfaces. Removal changes externally callable routes, so first check proxy/operator usage and define replacement responses. |
+| `scripts/fetch_real_data.py`, `scripts/download_data.py` | Historical standalone CSV utilities. The fetch script targets Wandsworth/Merton; neither is wired into the managed pipeline. Marked historical, retained pending confirmation of manual use. |
+| `pipeline/scripts/geocode_postcodes.py` | Mart-aware postcode fallback, not called by the current DAGs. Do not confuse it with the removed legacy ORM geocoder. Verify the target schema before manual use. |
+| `docker-compose.yml` | Uses unpublished `:latest` release tags and lacks frontend Payload DB/secret/media configuration. Retained as an old development topology, not recommended onboarding. |
+| `nextjs-app/docker-compose.yml` | Standalone legacy recipe with old backend port assumptions and no CMS persistence setup. Retained until its consumers are checked. |
+| `MIGRATION_SUMMARY.md`, `docs/superpowers/`, `mockups/` | Historical designs and prototypes. Retain as history; do not follow as current deployment instructions. |
+| `scripts/sql/drop_fact_parent_view.sql` | One-off maintenance SQL. Not an application entry point; repository call-site searches cannot establish whether it is still needed operationally. |
+
+## Active code that can look obsolete
+
+- `backend/data_loader.py` older-mart query fallbacks are covered by backend tests
+  and support databases at different migration stages. Remove only after verifying
+  the deployed schemas in every supported environment.
+- `backend/gias_codes.py` and `pipeline/scripts/gias_codes.py` are intentionally
+  generated copies for separate runtime images. Their parity is tested.
+- `nextjs-app/migrations/`, `payload-types.ts` and the Payload import map are active
+  CMS artifacts, not remnants of the removed school importer.
+- `get_available_years`, `get_available_local_authorities` and `get_schools_count`
+  in `data_loader.py` are called through `get_data_info`, which serves the backend
+  data-info endpoint. They are not dead functions.
+- `get_supplementary_data` is an intentional single-school wrapper around the
+  batch implementation.
+- `pipeline/transform` models named `legacy` can be active data sources: annual
+  DAG selectors explicitly include legacy KS2/KS4 lineage. Names alone do not
+  establish obsolescence.
+
+## Suggested next passes
+
+1. Decide the fate of the three unused UI components and remove paired assets/tests.
+2. Consolidate backend session usage and remove abandoned settings/constants.
+3. Verify external consumers, then retire static-serving API routes and old compose recipes.
+4. Audit manual data utilities with pipeline operators before deleting them.
+5. Revisit compatibility fallbacks only after documenting supported schema versions.
+
+Validation for this cleanup should include frontend typechecking/tests, Python
+syntax checks, reference searches and documentation link checks. Live database,
+external scheduler and deployed route usage require separate integration evidence.

docs/superpowers/specs/2026-09-21-similar-schools-nearby-design.md

@@ -0,0 +1,432 @@
+# Other Schools Nearby — Design
+
+**Date:** 2026-09-21, revised 2026-09-22
+**Status:** revised after staging review
+**Scope:** school detail pages, both phase templates
+
+> **Revision, 2026-09-22.** The first build ranked by intake similarity and used
+> distance as a tiebreak. On staging a Catholic primary showed six Catholic
+> primaries, none of them close enough to be a real option, and omitted the
+> community school down the road. Distance now decides the order and nothing
+> else does; the tier system is gone. The reasoning is kept below rather than
+> quietly overwritten, because the mistake is the instructive part.
+
+## Goal
+
+Give a school detail page an answer to the question every reader arrives with
+after the results tables: *and what else is around here?*
+
+Today a school page links outward to its place pages through
+`components/school/NearbyPlaces.tsx` and nowhere else. It never links to another
+school. This section adds that edge — up to six nearby schools of the same phase
+and a comparable intake, three at a time in a carousel, each a crawlable link and
+each addable to the comparison basket in one click.
+
+Mockup, in both themes (drawn against the original tiered design, so its ledes
+and chip fallbacks are one revision behind the copy specified below):
+<https://claude.ai/artifact/168KdUMcfkUeGWW2FGjuec>
+
+Source of the same page in the repo: `mockups/similar-schools-nearby.html`.
+
+## The constraint that shapes everything
+
+**A nearby school is not automatically a comparable school.**
+
+The section's whole value is that a reader treats what it shows as a shortlist.
+That makes every card an implicit claim that the school is a realistic
+alternative, and there are three ways that claim goes wrong:
+
+1. A **selective** school beside a non-selective one. Their intakes are
+   different by construction, so putting their Attainment 8 figures side by side
+   invites a conclusion the data cannot support.
+2. A **special school, PRU or AP** beside a mainstream school. This is the same
+   error PR #70 fixed for the England benchmark, where Greenmead (URN 101099)
+   rendered "0% — 62 below England".
+3. A **single-sex** school of the opposite sex. Not a weak match — not an option
+   at all.
+
+So the design separates two kinds of fact, and never confuses them:
+
+- **Hard filters** encode the claims above. They decide eligibility, and are
+  never relaxed at any distance, even if that means the section does not render.
+- **Shared characteristics** — gender, religious character, selectivity —
+  describe how closely an intake resembles this school's. They are *reported on
+  the card and never ranked on*, so the reader weighs them rather than having
+  them weighed for them.
+
+Everything below follows from that split. The revision at the top of this
+document is what happens when the second kind is treated as the first.
+
+## Selection algorithm
+
+A backend helper, `_nearby_schools_payload(urn)` in `backend/app.py`, modelled
+on the existing `_places_payload(urn)` and delegating to
+`backend/nearby_schools.select_nearby(frame, urn)`, which operates on the cached
+`load_latest_school_data()` frame — one row per URN, already carrying
+`latitude`, `longitude`, `phase`, `gender`, `religious_denomination`,
+`admissions_policy`, `school_type` and `status`.
+
+### Hard filters
+
+| Filter | Rule |
+|---|---|
+| Self | `urn` is excluded |
+| Status | GIAS status must be open |
+| Coordinates | both `latitude` and `longitude` present on both schools |
+| Phase | same phase group via the existing `PHASE_GROUPS` map |
+| Provision | special/PRU/AP match only each other |
+| Selectivity | selective matches selective; non-selective matches non-selective |
+| Gender | Boys never matches Girls; Mixed is compatible with both |
+
+`PHASE_GROUPS` is reused rather than re-derived so an all-through school is
+offered correctly on both the primary and secondary sides, exactly as it already
+behaves in search.
+
+The provision filter needs a backend counterpart to the frontend's
+`isSpecialSchool()` in `nextjs-app/lib/utils.ts:897`, reading the same GIAS
+establishment types through `backend/gias_codes.py`. The two must agree: a
+school the frontend treats as special for benchmarking but the backend treats as
+mainstream for matching would be dropped from its own England comparison and
+then offered as a peer to a mainstream school on the next page along.
+
+**Up to six cards, three visible.** Three fit the row; the rest are reached with
+the carousel arrows. Two is the minimum that renders at all.
+
+### Order: distance, and nothing else
+
+The nearest eligible schools, closest first. Similarity does not enter the
+ranking at any point.
+
+**Why not, having built it the other way first.** The original design ranked by
+tiers — same gender and faith within 3 miles, then same gender within 5, then
+anything within 10 — and used distance only to order the result. Two things
+followed, and both showed up on the first Catholic primary anyone looked at:
+
+- A faith match at 2.9 miles outranked a community school at 0.3 miles. For a
+  primary, whose catchment is routinely under a mile, the far school is not a
+  weaker option; it is not an option.
+- Because the row filled from the best tier before widening, three Catholic
+  schools within 3 miles were enough to fill all six slots with Catholic
+  schools. The stopping rule that produced this had been added to prevent the
+  *opposite* failure — padding a row with weak distant matches — and made this
+  one certain.
+
+The premise was backwards. **Distance is a constraint and intake is a
+preference.** A parent cannot act on a school outside their reach however well
+it matches, and they are perfectly capable of noticing a shared denomination
+for themselves if we show it to them. So similarity moved from the ranking to
+the card: `shared` reports what a school genuinely has in common, and the reader
+applies their own weighting.
+
+The hard filters above were always where the defensibility lived. They are
+untouched.
+
+### Reach: a sanity bound, not a target
+
+| Phase | Reach |
+|---|---|
+| Primary, middle deemed primary, all-through | 2 miles |
+| Secondary, middle deemed secondary | 6 miles |
+| 16 plus | 10 miles |
+
+Ordering by distance already handles density — a school in inner London fills
+all six slots inside a mile and never approaches the cap. The cap decides one
+thing: what happens where the area is sparse. It differs by phase because
+catchments do, and because people travel furthest for post-16.
+
+**A primary with nothing inside two miles renders no section**, and that is the
+intended answer rather than a gap. The alternative is a section headed "nearby"
+listing a school four miles from a five-year-old.
+
+**Past the sixth school, the rest are dropped without a count.** The section
+does not try to be the list: `NearbyPlaces` sits directly beneath and already
+leads to the place pages, which are built for browsing a full set and which the
+school page exists to feed.
+
+**Fewer than two results renders nothing.** Not an empty state, not a single
+lonely card. The section is absent, the nav item is absent, and the page is
+unchanged from before it existed.
+
+### Distance
+
+Straight-line, from the vectorised haversine already used for postcode search at
+`backend/app.py:831`, computed over the ~27k-row frame in numpy. Reported to one
+decimal place in miles, consistent with the rest of the site.
+
+Straight-line distance is not road distance and is not measured from the
+reader's home. The section says so in its disclosure rather than leaving the
+reader to assume otherwise.
+
+## API
+
+`/api/schools/{urn}` gains a `similar_schools` array. Each row:
+
+| Field | Notes |
+|---|---|
+| `urn` | for the link and the compare basket |
+| `school_name` | link text |
+| `distance_miles` | one decimal place |
+| `school_type` | GIAS type, translated, for the card's meta line |
+| `age_range` | for the meta line |
+| `shared` | what this school genuinely shares with the subject; may be empty |
+| `metric_value` | the phase-appropriate headline figure, or null |
+| `metric_key` | `rwm_expected_pct` or `attainment_8_score` — see below |
+| `metric_year` | the year the figure is from |
+
+The metric follows the subject school's phase side, not the neighbour's own
+phase, so a row of cards never mixes two scales. The secondary
+side uses `attainment_8_score`; the primary side uses `rwm_expected_pct`. Where
+the neighbour has no value for that key, the card reads "Not published" rather
+than falling back to the other key.
+
+Which side a school takes is decided once, in
+`similar_schools.is_secondary_phase`, by membership of `PHASE_GROUPS["secondary"]`
+minus all-through — never by testing for the substring "secondary", which misses
+`16 plus` (GIAS phase 6) and hands a sixth-form college the primary bucket.
+All-through is the exception in the other direction: `PHASE_GROUPS` lists it on
+both sides, but it takes the primary metric.
+
+This is usually the same thing as "the template the page renders", but not
+always. `computeSchoolFlags` decides the template with that same substring test,
+so a `16 plus` school renders `PrimarySchoolSections` while being matched — 
+correctly — against secondaries. The section therefore takes its lede noun from
+the school's own phase rather than from its template, or it would print "Other
+primary schools near <sixth form college>" above a row of secondaries.
+
+Up to six rows of roughly 130 bytes each. It rides in the existing detail payload
+rather than a new endpoint because the page already makes exactly one server
+fetch for its data, and `/school/[slug]` regenerates at most weekly
+(`revalidate = 604800`), so the per-request cost is paid once per school per
+week.
+
+**The key is absent, not null, on a backend that does not have this code.** The
+frontend treats absent and empty identically, which is what allowed
+`NearbyPlaces` to ship without a lockstep deploy of the two images.
+
+`shared` is computed on the backend, beside the data it is derived from, not
+re-derived on the frontend. Deriving it twice is how a card comes to claim
+something the selection never established. An empty list is a real answer and
+renders no chips: a bare card costs a school nothing but the likeness it does
+not have, since the order was already settled by distance.
+
+## Frontend
+
+### Components
+
+`components/school/SimilarSchoolsSection.tsx` — a server component wrapped in
+the shared `Section` shell from `sectionShared.tsx`. It renders the heading,
+the lede, the card grid, the footer CTA and one caption line. Every
+card's title is an `<a>` to the school's canonical slug URL via `schoolUrl()`.
+
+`components/school/AddToCompareButton.tsx` — calls `addSchool` from
+`ComparisonProvider` and reports the selection with a `from: 'similar_schools'`
+attribution, mirroring `addSchoolFromSearch` in `HomeView.tsx:442`.
+
+`components/school/SimilarSchoolsCarousel.tsx` — the scroller and its arrows. It
+takes the server-rendered cards as `children` and the server-rendered heading and
+lede as a `header` prop, so those stay server components while the client
+component owns only the ref, the scroll handler and the arrows' disabled state.
+
+The split matters: the links — the part with SEO value and the part that must
+work without JavaScript — are server-rendered into the initial HTML, and only
+the basket interaction and the arrows are hydrated.
+
+### The carousel
+
+**Every card is in the initial HTML.** The arrows scroll a list; they never swap
+a view. Six `<a>` elements are in the markup whether or not anything is
+hydrated, which is the whole reason the section exists — a paginated widget that
+mounts cards on click would put four of the six links beyond a crawler and
+beyond a reader with no JavaScript.
+
+So the scroller is a plain overflowing `<ul>` with `scroll-snap-type: x
+mandatory`, and the arrows call `scrollBy` on it. With no JavaScript it
+degrades to a horizontally scrollable row that still works by touch and by
+trackpad. Three cards are visible at desktop width and two below 820px.
+
+**Arrows appear only when there is somewhere to go** — that is, only when more
+than three schools were found. Each disables itself at its own end of the
+travel.
+
+#### Below 640px the arrows go away
+
+This follows [MOBILE.md](../../../MOBILE.md), which makes 360px the design
+floor and mobile the primary target at ≥55% of traffic.
+
+Kept in the heading's flex row at 360px, the two arrow buttons take 96px from a
+328px card and crush the lede into a four-line column — measured, not guessed.
+And swiping already does what they do. So below 640px the header becomes a
+single column, the arrows are not rendered, one card shows at 86% width so the
+next one peeks, and the affordance is carried by the right-edge scroll-fade that
+MOBILE.md documents for exactly this case:
+
+```css
+mask-image: linear-gradient(to right, #000 calc(100% - 28px), transparent);
+```
+
+The fade lifts at the end of the travel, where there is nothing left to hint
+at. That means the at-end state must be computed whether or not an arrow exists
+to consume it — on mobile it drives the mask alone.
+
+**Every interactive element clears 44×44px**, per MOBILE.md's iOS HIG check: the
+arrow buttons and the add-to-compare button are both 44px, up from the 40px they
+were first drawn at. A card title's own box is shorter than that, but its hit
+area is the whole card through the `::after` overlay, so it passes on the target
+that actually receives the tap.
+
+**The edge test needs a tolerance, and this is not fussiness.** The scroller
+carries 2px of padding so focus rings are not clipped, and scroll-snap treats
+that padding as the first card's snap position: a scroller sitting at its start
+reports `scrollLeft` of 2, not 0. Sub-pixel rounding moves it again at other
+zoom levels. Testing `scrollLeft === 0` therefore leaves the back arrow live and
+pointing nowhere on first paint — confirmed in the mockup before it was fixed.
+Both ends compare against an 8px tolerance.
+
+**Selecting a school must not move the row.** Adding to the basket re-renders
+the footer; the scroll offset lives in the DOM rather than in React state, so
+the carousel must not remount or reset on that render. A reader who ticks the
+fifth school and is thrown back to the first has been punished for using the
+feature.
+
+### Placement and navigation
+
+Rendered as the last section **inside** `SchoolDetailShell`, from both
+`PrimarySchoolSections` and `SecondarySchoolSections`. Inside, not after, because
+the sticky nav's scroll-spy locates sections with `document.getElementById` and
+can only reach a section that lives in the shell.
+
+`NearbyPlaces` stays where it is, outside the shell, immediately below. The
+resulting order — this school, then similar schools, then the places containing
+them — narrows before it widens, which is the order a reader leaves a page in.
+
+`buildNavItems` and `buildSecondaryNavItems` both gain
+`{ id: 'similar', label: 'Similar schools' }`, gated on the section rendering.
+The id must match the `Section` id or the scroll-spy silently breaks.
+
+### The comparison CTA
+
+A plain `<a href="/compare?urns=…">`, built from this school's URN plus the
+selected ones. `/compare` already parses `urns` from the query string
+(`app/(frontend)/compare/page.tsx:55`), so this needs no new compare plumbing.
+With nothing selected the CTA is disabled; the button also adds to the shared
+basket so the site-wide comparison state stays consistent with what the page
+shows.
+
+## Copy, and what the section is allowed to claim
+
+**The lede never claims an intake.** It reads "Other primary schools near X." —
+one sentence, no variants. The earlier version varied the wording by tier, which
+only existed to soften a claim the section should not have been making.
+
+**The heading is "Other schools nearby", not "Similar schools nearby".** The
+hard filters do guarantee a comparable set — same phase, same selectivity,
+mainstream never beside special — but nothing ranks on likeness, so the heading
+does not say it does. The nav item reads "Nearby schools" and the section id is
+`nearby`.
+
+**Chips state only what is shared, and may be absent entirely.** A card with
+nothing in common renders no chip row rather than falling back to a filler.
+Since chips no longer affect the order, an empty one costs that school nothing
+except a claim it cannot support — and a Catholic parent scanning the row still
+spots "Roman Catholic" on the card that carries it, and weighs it themselves.
+
+**The neighbour's metric carries no valence colour.** Green and terracotta are
+reserved site-wide for comparison against the England average. Colouring a
+neighbour's figure against this school's would read as ranking the neighbours
+against each other, which is precisely the endorsement this section must not
+make. The figure sits in neutral ink above a plain "72% at this school"
+reference line, and the reader draws their own conclusion.
+
+**A missing figure reads "Not published".** Never 0, never blank, never an
+em dash. This follows the same rule the rest of the detail page uses: a school
+with no published result has not scored zero.
+
+**There is no "how these are chosen" disclosure.** The method is visible in what
+the section already shows — the phase in the lede, the shared characteristics on
+each card, the distance above each name — and a collapsed panel restating it
+earns less than the space it costs.
+
+**One caption line survives, and only one:** that distances are straight-line
+from the school and not road distance. This is not a method note. A reader who
+sees "0.6 miles away" and takes it for the walk has been misled by us, and no
+other element on the card corrects that. The remaining notes — that listing is
+not a recommendation, that special schools only meet special schools — are
+statements the selection rules already keep true without being narrated.
+
+## Degradation
+
+| Condition | Behaviour |
+|---|---|
+| `similar_schools` absent (older backend image) | no section, no nav item |
+| fewer than 2 qualifying schools | no section, no nav item |
+| this school has no coordinates | no section |
+| the helper raises | returns `[]`; the page renders without the section |
+
+The helper is wrapped so a failure inside it never 500s a page that is otherwise
+complete — the posture `get_supplementary_data` already takes for its own
+queries.
+
+## Testing
+
+**Backend**, in a new `backend/tests/test_similar_schools.py`, against a
+synthetic frame rather than live marts:
+
+- a selective school never returns a non-selective one, and vice versa
+- a special school returns only special schools; a mainstream school returns none
+- a Boys school never returns a Girls school; Mixed matches both
+- closed schools and schools without coordinates are never returned
+- results are ordered by distance ascending, always
+- a faith match never outranks a closer school (the staging defect, pinned)
+- the nearest eligible school is always present
+- more than six qualifying schools returns the six nearest
+- reach is capped per phase, and a primary beyond two miles returns `[]`
+- an all-through school is offered on both phase sides
+- a `16 plus` school is matched against secondaries and colleges, never primaries
+- `is_secondary_phase` and `PHASE_GROUPS` agree on every GIAS phase value
+- fewer than two qualifying schools returns `[]`
+- distances match a hand-computed haversine for a known pair
+
+**Frontend**, in `nextjs-app/__tests__`:
+
+- the section renders nothing for absent, empty and single-row inputs
+- the lede never claims a similar intake
+- an empty `shared` renders no chips rather than a filler
+- a null metric renders "Not published"
+- the nav item appears only alongside the section
+- every card is in the DOM, including the ones scrolled out of view
+- arrows render only when more than three schools were found
+
+jsdom has no layout, so `scrollWidth` and `clientWidth` are both 0 there and the
+arrows' disabled state cannot be meaningfully asserted in Jest. That behaviour is
+covered in the journey instead, against a real engine, rather than by a unit test
+that would pass on a measurement that does not exist.
+
+**E2E**, added to the existing journeys in `e2e/tests` in the same PR, per the
+repository's rule on user-facing behaviour:
+
+- the section renders on a known staging URN, with resolving links
+- where arrows are present, the back arrow starts disabled and the forward arrow
+  moves the row
+- selecting a school does not reset the scroll position
+- add-to-compare reaches `/compare` with the expected `urns`
+- at 360, 390 and 430px: no horizontal overflow, every interactive element in the
+  section clears 44×44px, and no arrows are rendered
+
+The E2E gate runs after merge on this project, so these journeys are not
+provable in the PR checks; the PR is verified on the unit tests, and the
+journeys are confirmed on the post-merge staging run.
+
+## Out of scope
+
+- A map of the nearby schools. The section is a list; the page already has a map.
+- Autoplay, dots, or an infinite loop on the carousel. It is a short list a
+  reader scans deliberately, not a banner competing for attention, and a row
+  that moves on its own is a row that moves while someone is reading it.
+- Statistical neighbours on deprivation, size or cohort profile. If plain
+  distance proves too blunt, that is the trigger to move this computation into a
+  dbt mart — `select_nearby` is a deliberate seam for exactly that swap.
+- Precomputing neighbours in `marts.*`. Rejected for now: a new mart is inert
+  until Airflow runs, so the feature would ship dark, and every tuning change to
+  the rules would become a pipeline round-trip instead of a deploy. The revision
+  at the top of this document is the argument for keeping that loop short.
+- Any change to `/api/compare`, the compare page, or the comparison basket.

e2e/tests/journeys.spec.ts

@@ -1,4 +1,4 @@
-import { test, expect, Page } from '@playwright/test';
+import { test, expect, Locator, Page } from '@playwright/test';
 
 /**
  * Journey tests for SchoolCompare, run against the staging environment as the
@@ -19,6 +19,31 @@ function schoolLinks(page: Page) {
   return page.locator('a[href^="/school/"]');
 }
 
+/**
+ * A scroll offset that has stopped moving.
+ *
+ * The carousel arrows scroll with `behavior: 'smooth'`, so a reading taken
+ * straight after a click lands mid-animation. Measured against staging: the
+ * animation runs ~700ms, and a poll for "has it moved at all" is satisfied
+ * 50ms in, at 13px of a 1300px journey. A test that then records an offset,
+ * does something, and records again is measuring the tail of the arrow's
+ * animation rather than the effect of whatever it did in between.
+ *
+ * Two identical readings in a row is the cheapest sound definition of settled.
+ */
+async function settledScrollLeft(scroller: Locator): Promise<number> {
+  let previous = -1;
+  await expect
+    .poll(async () => {
+      const current = await scroller.evaluate((node: HTMLElement) => Math.round(node.scrollLeft));
+      const settled = current === previous;
+      previous = current;
+      return settled;
+    }, { timeout: 10_000 })
+    .toBe(true);
+  return previous;
+}
+
 /**
  * Two URNs guaranteed to be pure-primary (same phase). The compare page's
  * phase tabs split all-through schools (which carry KS4 data) onto the
@@ -2734,3 +2759,166 @@ test('the content sitemap lists the about page and is advertised in robots', asy
   expect(body).toContain('/sitemap.xml');
   expect(body).toContain('/content-sitemap.xml');
 });
+
+/**
+ * Other schools nearby.
+ *
+ * The section is absent by design where fewer than two schools qualify, and the
+ * arrows are absent where three cards fit, so this asserts each part of the
+ * contract only where it applies.
+ *
+ * Two things here cannot be tested anywhere else: the arrows' disabled state,
+ * which jsdom cannot measure because it has no layout, and the scroll position
+ * surviving a selection, which is DOM state rather than React state.
+ */
+test('nearby schools link on to other schools and into compare', async ({ page }) => {
+  await searchByName(page, 'Primary');
+  await schoolLinks(page).first().click();
+  await page.waitForURL(/\/school\//);
+
+  const section = page.locator('#nearby');
+  if ((await section.count()) === 0) {
+    test.skip(true, 'No qualifying similar schools for this school');
+  }
+
+  // Every card is a real link to another school page — including the ones
+  // behind the arrows, which is the whole reason this is a scroller and not a
+  // paginated widget.
+  const links = section.locator('a[href^="/school/"]');
+  const linkCount = await links.count();
+  expect(linkCount).toBeGreaterThanOrEqual(2);
+  expect(linkCount).toBeLessThanOrEqual(6);
+  expect(await links.first().getAttribute('href')).toMatch(/^\/school\/\d{6}-/);
+
+  await expect(section.getByText(/miles away/).first()).toBeVisible();
+
+  const scroller = section.locator('ul').first();
+
+  // The carousel, where this school had more than three matches.
+  const forward = section.getByRole('button', { name: 'More schools' });
+  if (await forward.count()) {
+    const back = section.getByRole('button', { name: 'Previous schools' });
+    await expect(back).toBeDisabled();
+
+    await forward.click();
+    expect(await settledScrollLeft(scroller)).toBeGreaterThan(8);
+    await expect(back).toBeEnabled();
+  }
+
+  // The compare hand-off, and the row must not jump back to the start when the
+  // footer re-renders underneath it.
+  //
+  // Click the LAST card's button, not the first. Playwright scrolls a target
+  // into view before clicking it, so clicking card one while the row is paged
+  // to the end scrolls the container back to the start — and the assertion
+  // below then measures Playwright's own scrolling rather than the app's.
+  // That is what this test did on its first staging run: 537 → 2, reproduced
+  // afterwards on a static page with no React on it at all.
+  //
+  // A few pixels of snap or sub-pixel adjustment are fine; a reset to the
+  // start is not, which is the whole point of the check.
+  const offsetBefore = await settledScrollLeft(scroller);
+  await section.getByRole('button', { name: /Add to compare/ }).last().click();
+  await expect(
+    section.getByRole('button', { name: /Added to compare/ }).first(),
+  ).toBeVisible();
+  const offsetAfter = await settledScrollLeft(scroller);
+  expect(Math.abs(offsetAfter - offsetBefore)).toBeLessThanOrEqual(8);
+});
+
+/**
+ * Every section in the mobile jump sheet can actually be reached.
+ *
+ * The sheet is a fixed bottom sheet, and the app has a fixed bottom tab bar.
+ * `position: sticky` with a z-index on the sticky nav makes it a stacking
+ * context, so the sheet's own z-index orders it only within that context —
+ * against the tab bar, the nav's value is what counts. The last item in the
+ * sheet was therefore painted over and untappable as soon as the list grew
+ * long enough to reach the bar, which adding "Nearby schools" is what did.
+ *
+ * Bounding boxes are not enough to catch this: the item is in the viewport and
+ * the right size, it is simply underneath something. So this asks the question
+ * a thumb asks — what is on top at this point.
+ */
+test('every section in the mobile jump sheet is tappable, not under the tab bar', async ({ page }) => {
+  await page.setViewportSize({ width: 390, height: 844 });
+  await searchByName(page, 'Primary');
+  await schoolLinks(page).first().click();
+  await page.waitForURL(/\/school\//);
+
+  // Scroll down so the sticky nav is docked and the sheet has somewhere to open.
+  await page.evaluate(() => window.scrollTo({ top: 1200 }));
+  // Two controls carry aria-haspopup: the mobile "Section" button and the
+  // desktop "All" one, which is display:none here but still in the DOM.
+  await page.locator('[aria-haspopup="menu"]:visible').click();
+
+  const sheet = page.locator('[role="menu"]');
+  await expect(sheet).toBeVisible();
+
+  const covered = await sheet.evaluate((panel: HTMLElement) =>
+    Array.from(panel.querySelectorAll('[role="menuitem"]'))
+      .map((el) => {
+        const box = el.getBoundingClientRect();
+        const hit = document.elementFromPoint(
+          Math.round(box.left + box.width / 2),
+          Math.round(box.top + box.height / 2),
+        );
+        return { label: (el as HTMLElement).innerText.trim().replace(/\s+/g, ' '), reachable: !!(hit && hit.closest('[role="menuitem"]')) };
+      })
+      .filter((item) => !item.reachable)
+      .map((item) => item.label),
+  );
+
+  expect(covered).toEqual([]);
+});
+
+/**
+ * The nearby-schools section at MOBILE.md's three reference widths.
+ *
+ * MOBILE.md asks for exactly this check and records that it was not written
+ * because "Playwright isn't currently in the project dependency set". That is
+ * no longer true — this suite is Playwright — so the check exists now, scoped
+ * to the page this feature touches.
+ */
+for (const width of [360, 390, 430]) {
+  test(`nearby schools survives a ${width}px viewport`, async ({ page }) => {
+    await page.setViewportSize({ width, height: 800 });
+    await searchByName(page, 'Primary');
+    await schoolLinks(page).first().click();
+    await page.waitForURL(/\/school\//);
+
+    const section = page.locator('#nearby');
+    if ((await section.count()) === 0) {
+      test.skip(true, 'No qualifying similar schools for this school');
+    }
+
+    // 1. Nothing bleeds past the right edge.
+    expect(
+      await page.evaluate(() => document.documentElement.scrollWidth - window.innerWidth),
+    ).toBe(0);
+
+    // 2. No arrows on touch widths — swiping does the job, and they would take
+    //    96px from a 328px card.
+    await expect(section.getByRole('button', { name: 'More schools' })).toHaveCount(0);
+
+    // 3. Every tap target in the section clears 44px. A card title's own box is
+    //    shorter, but its hit area is the whole card via ::after.
+    const failing = await section.evaluate((root: HTMLElement) =>
+      Array.from(root.querySelectorAll('a, button'))
+        .filter((el) => (el as HTMLElement).offsetParent)
+        .map((el) => {
+          const card = el.closest('li');
+          const box = el.matches('h3 a') && card
+            ? card.getBoundingClientRect()
+            : el.getBoundingClientRect();
+          return {
+            text: (el as HTMLElement).innerText.trim().slice(0, 24),
+            w: box.width,
+            h: box.height,
+          };
+        })
+        .filter((o) => o.w < 44 || o.h < 44),
+    );
+    expect(failing).toEqual([]);
+  });
+}

e2e/tests/reliability.spec.ts

@@ -0,0 +1,41 @@
+import { test, expect, Route } from '@playwright/test';
+
+test('the deployed frontend and backend report the tested build', async ({ request }) => {
+  const response = await request.get('/release.json');
+  expect(response.ok()).toBeTruthy();
+  expect(response.headers()['cache-control']).toContain('no-store');
+  const identity = await response.json();
+  expect(identity.frontend).toEqual(identity.backend);
+  expect(identity.frontend.sha).toMatch(/^[a-f0-9]{40}$/);
+  expect(identity.frontend.build_id).toMatch(/^[a-f0-9]{32}$/);
+  if (process.env.EXPECTED_SHA) expect(identity.frontend.sha).toBe(process.env.EXPECTED_SHA);
+  if (process.env.EXPECTED_BUILD_ID) expect(identity.frontend.build_id).toBe(process.env.EXPECTED_BUILD_ID);
+});
+
+test('changing search while loading another page does not append old results', async ({ page }) => {
+  await page.goto('/?phase=primary');
+  await expect(page.getByRole('button', { name: 'Load more schools' })).toBeVisible();
+  let received!: (route: Route) => void;
+  const pending = new Promise<Route>(resolve => { received = resolve; });
+  await page.route('**/api/schools?**', async route => {
+    if (new URL(route.request().url()).searchParams.get('page') === '2') {
+      received(route);
+      return;
+    }
+    await route.continue();
+  });
+  await page.getByRole('button', { name: 'Load more schools' }).click();
+  const oldRequest = await pending;
+  const search = page.getByPlaceholder('School name or postcode').first();
+  await search.fill('secondary');
+  await search.press('Enter');
+  await page.waitForURL(/search=secondary/);
+  // A cancelled fetch may prevent route fulfilment altogether; either way,
+  // this deliberately late response must not become part of the new results.
+  await oldRequest.fulfill({ json: {
+    schools: [{ urn: 999998, school_name: 'P1 stale result sentinel', phase: 'Primary' }],
+    total: 2, page: 2, page_size: 1, total_pages: 2,
+  } }).catch(() => {});
+  await expect(page.getByText('P1 stale result sentinel')).toHaveCount(0);
+  await expect(page.getByRole('button', { name: 'Loading...' })).toHaveCount(0);
+});

mockups/similar-schools-nearby.html

@@ -0,0 +1,373 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>Similar Schools Nearby</title>
+<link rel="preconnect" href="https://fonts.googleapis.com">
+<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
+<link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600&family=Manrope:wght@500;600;700&display=swap" rel="stylesheet">
+<style>
+/* Tokens copied verbatim from nextjs-app/app/(frontend)/globals.css so this
+   mockup cannot drift from the shipped palette. Light values first, dark
+   under prefers-color-scheme, both overridable by the theme switch. */
+:root {
+  color-scheme: light;
+  --bg-primary:#FAFAF8; --bg-secondary:#F5EFE6; --bg-card:#FFFFFF;
+  --text-primary:#1C2731; --text-secondary:#4A5560; --text-muted:#5F6A75;
+  --border:#E5E7EB; --border-strong:#D3D7DD;
+  --brand:#0F766E; --brand-strong:#0C5F58; --brand-bg:rgba(15,118,110,.10); --brand-on:#FFFFFF;
+  --action:#BE3C27; --action-strong:#A33320; --action-on:#FFFFFF;
+  --sand:#F5EFE6;
+  --font-display:Manrope,-apple-system,BlinkMacSystemFont,sans-serif;
+  --font-ui:Inter,-apple-system,BlinkMacSystemFont,sans-serif;
+  --radius-md:8px; --radius-lg:16px;
+  --shadow:0 1px 2px rgba(28,39,49,.06),0 1px 3px rgba(28,39,49,.05);
+}
+@media (prefers-color-scheme: dark) {
+  :root:not([data-theme="light"]) {
+    color-scheme: dark;
+    --bg-primary:#111A20; --bg-secondary:#16222A; --bg-card:#18242C;
+    --text-primary:#E9EEF0; --text-secondary:#B4C2C7; --text-muted:#8B9AA1;
+    --border:#26343D; --border-strong:#35454F;
+    --brand:#5FC7BB; --brand-strong:#7BD6CC; --brand-bg:rgba(95,199,187,.14); --brand-on:#0A1418;
+    --action:#F08A72; --action-strong:#F5A492; --action-on:#241009;
+    --sand:#1B2730;
+    --shadow:0 1px 2px rgba(0,0,0,.3),0 1px 3px rgba(0,0,0,.25);
+  }
+}
+:root[data-theme="dark"] {
+  color-scheme: dark;
+  --bg-primary:#111A20; --bg-secondary:#16222A; --bg-card:#18242C;
+  --text-primary:#E9EEF0; --text-secondary:#B4C2C7; --text-muted:#8B9AA1;
+  --border:#26343D; --border-strong:#35454F;
+  --brand:#5FC7BB; --brand-strong:#7BD6CC; --brand-bg:rgba(95,199,187,.14); --brand-on:#0A1418;
+  --action:#F08A72; --action-strong:#F5A492; --action-on:#241009;
+  --sand:#1B2730;
+  --shadow:0 1px 2px rgba(0,0,0,.3),0 1px 3px rgba(0,0,0,.25);
+}
+* { box-sizing:border-box; }
+body {
+  margin:0; padding:32px 16px 80px; background:var(--bg-primary);
+  color:var(--text-primary); font:15px/1.55 var(--font-ui);
+  -webkit-font-smoothing:antialiased;
+}
+.page { max-width:960px; margin:0 auto; }
+.page > header { margin-bottom:28px; display:flex; flex-wrap:wrap; gap:16px; align-items:flex-start; justify-content:space-between; }
+.page > header h1 { font:700 25px/1.25 var(--font-display); letter-spacing:-.6px; margin:0 0 6px; }
+.page > header p { margin:0; color:var(--text-muted); font-size:14px; max-width:60ch; }
+.theme-switch { border:1px solid var(--border-strong); background:var(--bg-card); color:var(--text-secondary); border-radius:999px; padding:8px 14px; font:500 13px var(--font-ui); cursor:pointer; min-height:44px; }
+.theme-switch:hover { border-color:var(--brand); color:var(--brand); }
+
+/* ── The page context each variant is shown inside ─────────────────── */
+.variant { margin-bottom:40px; }
+.variant > .context { padding:0 4px 12px; }
+.variant .eyebrow { margin:0 0 4px; font-size:12px; letter-spacing:.04em; text-transform:uppercase; color:var(--text-muted); }
+.variant .context h2 { font:600 18px/1.35 var(--font-display); margin:0; color:var(--text-secondary); }
+.variant .note { margin:10px 4px 0; font-size:12.5px; color:var(--text-muted); }
+.variant .note b { color:var(--text-secondary); font-weight:600; }
+
+/* ── The section itself — mirrors components/school/Section ────────── */
+.card {
+  background:var(--bg-card); border:1px solid var(--border);
+  border-radius:var(--radius-lg); padding:28px; box-shadow:var(--shadow);
+}
+.top { display:flex; align-items:flex-start; justify-content:space-between; gap:16px; }
+.top h2 { font:700 22px/1.25 var(--font-display); letter-spacing:-.4px; margin:0; }
+.lede { margin:8px 0 20px; color:var(--text-secondary); font-size:14.5px; max-width:64ch; }
+
+/* ── Carousel ───────────────────────────────────────────────────────
+   Every card is in the DOM and in the initial HTML — the arrows scroll a
+   list, they do not swap a view. That keeps all six links crawlable and
+   keeps the section usable with no JavaScript, where it degrades to a
+   plain horizontally scrollable row. */
+.arrows { display:flex; gap:8px; flex:none; }
+.arrow {
+  width:44px; height:44px; display:grid; place-items:center; cursor:pointer;
+  border:1px solid var(--border-strong); border-radius:999px;
+  background:var(--bg-card); color:var(--brand);
+}
+.arrow:hover:not(:disabled) { border-color:var(--brand); background:var(--brand-bg); }
+.arrow:disabled { opacity:.35; cursor:default; }
+.arrow:focus-visible { outline:2px solid var(--brand); outline-offset:2px; }
+.arrow svg { width:17px; height:17px; }
+
+.scroller {
+  display:grid; grid-auto-flow:column;
+  grid-auto-columns:calc((100% - 28px) / 3);
+  gap:14px; overflow-x:auto; scroll-snap-type:x mandatory;
+  padding:2px; margin:-2px;                /* room for focus rings */
+  scrollbar-width:none; -ms-overflow-style:none;
+  list-style:none;
+}
+.scroller::-webkit-scrollbar { display:none; }
+.scroller:focus-visible { outline:2px solid var(--brand); outline-offset:4px; border-radius:var(--radius-md); }
+@media (max-width:820px) { .scroller { grid-auto-columns:calc((100% - 14px) / 2); } }
+/* Touch widths: the arrows would squeeze the lede into a four-line column for a
+   control that swiping already provides, so they go and the documented
+   right-edge fade carries the affordance instead (MOBILE.md). The fade lifts at
+   the end of the travel, where there is nothing more to hint at. */
+@media (max-width:640px) {
+  .top { display:block; }
+  .arrows { display:none; }
+  .scroller { grid-auto-columns:86%; mask-image:linear-gradient(to right, #000 calc(100% - 28px), transparent); }
+  .scroller[data-at-end=true] { mask-image:none; }
+  .card { padding:20px; }
+}
+
+.school {
+  position:relative; display:flex; flex-direction:column; scroll-snap-align:start;
+  border:1px solid var(--border); border-radius:var(--radius-md);
+  padding:16px; background:var(--bg-card);
+}
+.school:has(.add[aria-pressed=true]) { border-color:var(--brand); background:var(--brand-bg); }
+.distance { display:flex; align-items:center; gap:5px; font-size:12px; color:var(--text-muted); margin:0 0 10px; }
+.distance svg { width:13px; height:13px; flex:none; }
+.school h3 { font:600 16px/1.35 var(--font-display); margin:0 0 6px; }
+/* The whole card is the link target; the button sits above it on z-index so
+   it stays independently clickable. */
+.school h3 a { color:var(--text-primary); text-decoration:none; }
+.school h3 a::after { content:""; position:absolute; inset:0; border-radius:var(--radius-md); }
+.school:hover { border-color:var(--border-strong); }
+.school h3 a:hover { color:var(--brand); text-decoration:underline; }
+.school h3 a:focus-visible { outline:none; }
+.school:has(h3 a:focus-visible) { outline:2px solid var(--brand); outline-offset:2px; }
+.meta { margin:0 0 12px; font-size:12.5px; color:var(--text-muted); }
+.shared { display:flex; flex-wrap:wrap; gap:6px; margin:0 0 14px; padding:0; list-style:none; }
+.shared li { font-size:11.5px; line-height:1.4; padding:4px 8px; border-radius:999px; background:var(--brand-bg); color:var(--brand); border:1px solid transparent; }
+.shared li.loose { background:transparent; color:var(--text-muted); border-color:var(--border); }
+.metric { margin-top:auto; padding-top:13px; border-top:1px solid var(--border); }
+.value { font:700 26px/1.1 var(--font-display); letter-spacing:-.6px; margin:0; }
+.value.absent { font-size:15px; font-weight:600; color:var(--text-muted); letter-spacing:0; }
+.metric .label { margin:4px 0 0; font-size:12px; color:var(--text-secondary); }
+.metric .ref { margin:2px 0 0; font-size:12px; color:var(--text-muted); }
+.add {
+  position:relative; z-index:1; margin-top:14px; width:100%; min-height:44px;
+  font:500 13px var(--font-ui); cursor:pointer; border-radius:var(--radius-md);
+  border:1px solid var(--border-strong); background:var(--bg-card); color:var(--brand);
+}
+.add:hover { border-color:var(--brand); background:var(--brand-bg); }
+.add[aria-pressed=true] { border-color:var(--brand); background:var(--brand-bg); font-weight:600; }
+.add:focus-visible { outline:2px solid var(--brand); outline-offset:2px; }
+
+.footer {
+  display:flex; flex-wrap:wrap; align-items:center; justify-content:space-between;
+  gap:14px; margin-top:20px; padding-top:18px; border-top:1px solid var(--border);
+}
+.footer p { margin:0; font-size:12.5px; color:var(--text-muted); }
+.footer strong { display:block; font:600 14px var(--font-ui); color:var(--text-primary); }
+/* Coral: the one decisive action in this section, and there is only one. */
+.compare {
+  min-height:44px; padding:0 20px; border-radius:var(--radius-md); cursor:pointer;
+  font:600 14px var(--font-ui); background:var(--action); color:var(--action-on);
+  border:1px solid var(--action); text-decoration:none; display:inline-flex; align-items:center; gap:8px;
+}
+.compare:hover { background:var(--action-strong); border-color:var(--action-strong); }
+.compare[aria-disabled=true] { opacity:.45; pointer-events:none; }
+.caption { margin:16px 0 0; font-size:11.5px; color:var(--text-muted); }
+.sr { position:absolute; width:1px; height:1px; padding:0; margin:-1px; overflow:hidden; clip:rect(0 0 0 0); white-space:nowrap; border:0; }
+</style>
+</head>
+<body>
+<div class="page">
+<header>
+  <div>
+    <h1>Similar schools nearby</h1>
+    <p>A new section on the school detail page. Fictional schools and figures; shipped
+       colour, type and section shell taken from <code>globals.css</code>.</p>
+  </div>
+  <button class="theme-switch" type="button" id="theme">Dark theme</button>
+</header>
+<div id="variants"></div>
+</div>
+
+<script>
+const PIN = '<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><path d="M20 10c0 6-8 12-8 12s-8-6-8-12a8 8 0 0 1 16 0Z"/><circle cx="12" cy="10" r="3"/></svg>';
+const CHEV = (dir) => `<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><path d="${dir === 'prev' ? 'M15 18l-6-6 6-6' : 'M9 18l6-6-6-6'}"/></svg>`;
+
+const variants = [
+  {
+    id: 'dense',
+    eyebrow: 'Variant 1 · Dense urban primary — six matches, carousel active',
+    context: 'Meadowbrook Primary School — Ages 4–11 · Mixed · No religious character · Community school',
+    lede: 'Other primary schools near Meadowbrook Primary School, with a similar intake.',
+    metric: 'Reading, writing & maths',
+    caption: 'Meeting the expected standard at key stage 2, 2025.',
+    thisValue: '72%',
+    note: 'Fourteen schools cleared <b>tier 1</b> within three miles, so the section takes the six nearest and stops there. The arrows scroll a list that is entirely in the HTML — all six links are crawlable, and with JavaScript off the row still scrolls.',
+    schools: [
+      { name:'Willow Lane Primary School', distance:'0.4', meta:'Community school · Ages 4–11', shared:['Mixed','No religious character'], value:'74%' },
+      { name:'Oakfield Primary School', distance:'0.6', meta:'Academy converter · Ages 3–11', shared:['Mixed','No religious character'], value:'69%' },
+      { name:'Brookside Primary School', distance:'0.9', meta:'Community school · Ages 4–11', shared:['Mixed','No religious character'], value:'Not published' },
+      { name:'Hollytree Primary School', distance:'1.3', meta:'Academy converter · Ages 4–11', shared:['Mixed','No religious character'], value:'81%' },
+      { name:'Marsh Green Primary School', distance:'1.8', meta:'Community school · Ages 3–11', shared:['Mixed','No religious character'], value:'64%' },
+      { name:'Kingsway Primary School', distance:'2.2', meta:'Foundation school · Ages 4–11', shared:['Mixed','No religious character'], value:'77%' },
+    ],
+  },
+  {
+    id: 'secondary',
+    eyebrow: 'Variant 2 · Secondary — four matches, mixed tiers',
+    context: 'Meadowbrook High School — Ages 11–18 · Mixed · Non-selective · Academy',
+    lede: 'Other secondary schools near Meadowbrook High School, with a similar intake.',
+    metric: 'Attainment 8',
+    caption: 'Average GCSE attainment score across eight qualifications, 2025.',
+    thisValue: '51.2',
+    note: 'Only two schools cleared tier 1, so the search widened to <b>tier 2</b> and found two more. It stops there rather than widening again to reach six — tiers relax to reach a usable set, never to fill the last slots. Selectivity never relaxes, so no grammar school can appear here.',
+    schools: [
+      { name:'Rivermead High School', distance:'0.9', meta:'Academy converter · Ages 11–18', shared:['Mixed','Non-selective','No religious character'], value:'52.8' },
+      { name:'Oakfield Academy', distance:'1.7', meta:'Academy sponsor led · Ages 11–16', shared:['Mixed','Non-selective','No religious character'], value:'49.6' },
+      { name:'St Aidan’s Catholic High School', distance:'2.4', meta:'Voluntary aided · Ages 11–18', shared:['Mixed','Non-selective'], value:'53.4' },
+      { name:'Parkside Community School', distance:'3.8', meta:'Community school · Ages 11–16', shared:['Mixed','Non-selective'], value:'50.9' },
+    ],
+  },
+  {
+    id: 'sparse',
+    eyebrow: 'Variant 3 · Rural — two matches, no arrows',
+    context: 'Little Ashby Church of England Primary School — Ages 4–11 · Mixed · Church of England · Voluntary controlled',
+    lede: 'Other primary schools near Little Ashby Church of England Primary School.',
+    metric: 'Reading, writing & maths',
+    caption: 'Meeting the expected standard at key stage 2, 2025.',
+    thisValue: '66%',
+    note: 'Nothing matched on religious character within range. At <b>tier 3</b> the lede drops the phrase “with a similar intake” and the chips fall back to the plain phase. Two cards fit the row, so the arrows are not rendered at all. One school fewer and the section would not render either.',
+    schools: [
+      { name:'Great Marden Primary School', distance:'4.2', meta:'Community school · Ages 4–11', shared:['Primary school'], loose:true, value:'71%' },
+      { name:'Ashby Vale Academy', distance:'7.8', meta:'Academy converter · Ages 4–11', shared:['Primary school'], loose:true, value:'58%' },
+    ],
+  },
+];
+
+const selected = Object.fromEntries(variants.map((v) => [v.id, new Set()]));
+
+function card(v, s, i) {
+  const on = selected[v.id].has(i);
+  const absent = s.value === 'Not published';
+  return `
+  <li class="school">
+    <p class="distance">${PIN}${s.distance} miles away</p>
+    <h3><a href="#">${s.name}</a></h3>
+    <p class="meta">${s.meta}</p>
+    <ul class="shared">${s.shared.map((c) => `<li class="${s.loose ? 'loose' : ''}">${c}</li>`).join('')}</ul>
+    <div class="metric">
+      <p class="value ${absent ? 'absent' : ''}">${s.value}</p>
+      <p class="label">${v.metric}</p>
+      <p class="ref">${v.thisValue} at this school</p>
+    </div>
+    <button class="add" type="button" data-variant="${v.id}" data-index="${i}" aria-pressed="${on}">
+      ${on ? '✓ Added to compare' : '+ Add to compare'}<span class="sr"> — ${s.name}</span>
+    </button>
+  </li>`;
+}
+
+function render() {
+  document.getElementById('variants').innerHTML = variants.map((v) => {
+    const count = selected[v.id].size;
+    // Three fit the row, so anything more is what the arrows are for.
+    const scrollable = v.schools.length > 3;
+    return `
+    <section class="variant">
+      <div class="context">
+        <p class="eyebrow">${v.eyebrow}</p>
+        <h2>${v.context}</h2>
+      </div>
+      <div class="card">
+        <div class="top">
+          <div>
+            <h2 id="h-${v.id}">Similar schools nearby</h2>
+            <p class="lede">${v.lede}</p>
+          </div>
+          ${scrollable ? `<div class="arrows">
+            <button class="arrow" type="button" data-scroll="prev" data-variant="${v.id}" aria-label="Previous schools" aria-controls="sc-${v.id}">${CHEV('prev')}</button>
+            <button class="arrow" type="button" data-scroll="next" data-variant="${v.id}" aria-label="More schools" aria-controls="sc-${v.id}">${CHEV('next')}</button>
+          </div>` : ''}
+        </div>
+
+        <ul class="scroller" id="sc-${v.id}" ${scrollable ? `tabindex="0" role="group" aria-labelledby="h-${v.id}"` : ''}>
+          ${v.schools.map((s, i) => card(v, s, i)).join('')}
+        </ul>
+
+        <div class="footer">
+          <p aria-live="polite">
+            <strong>${count ? `${count} school${count === 1 ? '' : 's'} selected` : 'Compare side by side'}</strong>
+            ${count ? 'This school is included automatically.' : 'Add a school to compare it with this one.'}
+          </p>
+          <a class="compare" href="#" aria-disabled="${count ? 'false' : 'true'}">
+            ${count ? `Compare ${count + 1} schools` : 'Compare'} →
+          </a>
+        </div>
+        <p class="caption">Distances are straight-line from this school, not road distance.
+           ${v.caption} Fictional schools and figures for this mockup.</p>
+      </div>
+      <p class="note">${v.note}</p>
+    </section>`;
+  }).join('');
+
+  variants.forEach((v) => {
+    const scroller = document.getElementById(`sc-${v.id}`);
+    if (scroller) syncArrows(v.id, scroller);
+  });
+}
+
+/** An arrow that scrolls nowhere is a dead control, so each end disables its own.
+ *
+ *  EDGE is not paranoia. The scroller carries 2px of padding so focus rings are
+ *  not clipped, and scroll-snap treats that padding as the first card's snap
+ *  position — so a scroller sitting at its start reports scrollLeft 2, not 0.
+ *  Sub-pixel rounding at other zoom levels moves it again. Testing against an
+ *  exact 0 leaves the back arrow live at the start, pointing nowhere. */
+const EDGE = 8;
+
+function syncArrows(id, scroller) {
+  const max = scroller.scrollWidth - scroller.clientWidth;
+  const atStart = scroller.scrollLeft <= EDGE;
+  const atEnd = scroller.scrollLeft >= max - EDGE;
+
+  // Drives the mobile scroll-fade, so it is computed even where no arrow is
+  // rendered to consume it.
+  scroller.dataset.atEnd = String(atEnd);
+
+  const prev = document.querySelector(`.arrow[data-scroll="prev"][data-variant="${id}"]`);
+  const next = document.querySelector(`.arrow[data-scroll="next"][data-variant="${id}"]`);
+  if (!prev || !next) return;
+  prev.disabled = atStart;
+  next.disabled = atEnd;
+}
+
+document.getElementById('variants').addEventListener('click', (event) => {
+  const arrow = event.target.closest('.arrow');
+  if (arrow) {
+    const scroller = document.getElementById(`sc-${arrow.dataset.variant}`);
+    // A page is what the reader can see, so the viewport is the step.
+    scroller.scrollBy({ left: (arrow.dataset.scroll === 'next' ? 1 : -1) * scroller.clientWidth, behavior: 'smooth' });
+    return;
+  }
+  const button = event.target.closest('.add');
+  if (!button) return;
+  const { variant, index } = button.dataset;
+  const set = selected[variant];
+  const i = Number(index);
+  // Scroll position is DOM state, not React state; keep it across the re-render.
+  const offset = document.getElementById(`sc-${variant}`).scrollLeft;
+  set.has(i) ? set.delete(i) : set.add(i);
+  render();
+  const scroller = document.getElementById(`sc-${variant}`);
+  scroller.scrollLeft = offset;
+  syncArrows(variant, scroller);
+  document.querySelector(`.add[data-variant="${variant}"][data-index="${index}"]`).focus();
+}, true);
+
+document.getElementById('variants').addEventListener('scroll', (event) => {
+  const scroller = event.target.closest('.scroller');
+  if (scroller) syncArrows(scroller.id.replace('sc-', ''), scroller);
+}, true);
+
+const themeButton = document.getElementById('theme');
+themeButton.addEventListener('click', () => {
+  const dark = document.documentElement.dataset.theme === 'dark';
+  document.documentElement.dataset.theme = dark ? 'light' : 'dark';
+  themeButton.textContent = dark ? 'Dark theme' : 'Light theme';
+});
+
+render();
+</script>
+</body>
+</html>

nextjs-app/.env.example

@@ -1,8 +1,16 @@
-# API Configuration
-NEXT_PUBLIC_API_URL=http://localhost:8000/api
+# Browser requests use the same-origin Next.js proxy.
+NEXT_PUBLIC_API_URL=/api
 
-# Production API URL (for deployment)
-# NEXT_PUBLIC_API_URL=https://api.schoolcompare.co.uk/api
+# Absolute URL for server-side fetching and the proxy; include /api.
+# In the managed container network this is http://backend:80/api (staging differs).
+FASTAPI_URL=http://localhost:8000/api
+
+# Payload CMS runtime configuration. Use the managed environment's database;
+# Payload owns the payload schema, independently of the school marts.
+DATABASE_URL=postgresql://schoolcompare:CHANGE_THIS_PASSWORD@localhost:5432/schoolcompare
+# Generate a secret: python -c "import secrets; print(secrets.token_urlsafe(32))"
+# Use distinct secrets for staging and production.
+PAYLOAD_SECRET=CHANGE_THIS_TO_A_SECURE_RANDOM_SECRET
 
 # Node Environment
 NODE_ENV=development

nextjs-app/DEPLOYMENT.md

@@ -1,291 +1,15 @@
-# Deployment Guide
+# Frontend deployment
 
-This guide covers deployment options for the SchoolCompare Next.js application.
+Next.js and Payload run in the same frontend container. The maintained deployment
+procedure is [docs/DEPLOY.md](../docs/DEPLOY.md), with the production and staging
+Portainer compose files at the repository root.
 
-## Deployment Options
+The frontend Dockerfile builds a standalone Next.js image. Runtime configuration
+supplies `FASTAPI_URL`, `DATABASE_URL` and `PAYLOAD_SECRET`; uploaded CMS media is
+persisted in a volume. Promote the built image through the repository's Gitea
+workflow after human staging approval.
 
-### Option 1: Vercel (Recommended for Next.js)
-
-Vercel is the easiest and most optimized platform for Next.js applications.
-
-#### Steps:
-
-1. **Install Vercel CLI**:
-   ```bash
-   npm install -g vercel
-   ```
-
-2. **Login to Vercel**:
-   ```bash
-   vercel login
-   ```
-
-3. **Deploy**:
-   ```bash
-   vercel --prod
-   ```
-
-4. **Configure Environment Variables** in Vercel dashboard:
-   - `NEXT_PUBLIC_API_URL`: Your FastAPI endpoint (e.g., `https://api.schoolcompare.co.uk/api`)
-   - `FASTAPI_URL`: Same as above for server-side requests
-
-#### Benefits:
-- Automatic HTTPS
-- Global CDN
-- Zero-config deployment
-- Automatic preview deployments
-- Built-in analytics
-
----
-
-### Option 2: Docker (Self-hosted)
-
-Deploy using Docker containers for full control.
-
-#### Prerequisites:
-- Docker 20+
-- Docker Compose 2+
-
-#### Steps:
-
-1. **Build Docker Image**:
-   ```bash
-   docker build -t schoolcompare-nextjs:latest .
-   ```
-
-2. **Run with Docker Compose**:
-   ```bash
-   # Create .env file with production variables
-   echo "NEXT_PUBLIC_API_URL=https://api.schoolcompare.co.uk/api" > .env
-   echo "FASTAPI_URL=http://backend:8000/api" >> .env
-
-   # Start services
-   docker-compose up -d
-   ```
-
-3. **Verify Deployment**:
-   ```bash
-   curl http://localhost:3000
-   ```
-
-#### Environment Variables:
-- `NEXT_PUBLIC_API_URL`: Public API endpoint (client-side)
-- `FASTAPI_URL`: Internal API endpoint (server-side)
-- `NODE_ENV`: `production`
-
----
-
-### Option 3: PM2 (Node.js Process Manager)
-
-Deploy directly on a Node.js server using PM2.
-
-#### Prerequisites:
-- Node.js 24+
-- PM2 (`npm install -g pm2`)
-
-#### Steps:
-
-1. **Build Application**:
-   ```bash
-   npm run build
-   ```
-
-2. **Create PM2 Ecosystem File** (`ecosystem.config.js`):
-   ```javascript
-   module.exports = {
-     apps: [{
-       name: 'schoolcompare-nextjs',
-       script: 'npm',
-       args: 'start',
-       cwd: '/path/to/nextjs-app',
-       instances: 'max',
-       exec_mode: 'cluster',
-       env: {
-         NODE_ENV: 'production',
-         PORT: 3000,
-         NEXT_PUBLIC_API_URL: 'https://api.schoolcompare.co.uk/api',
-         FASTAPI_URL: 'http://localhost:8000/api',
-       },
-     }],
-   };
-   ```
-
-3. **Start with PM2**:
-   ```bash
-   pm2 start ecosystem.config.js
-   pm2 save
-   pm2 startup
-   ```
-
----
-
-### Option 4: Nginx Reverse Proxy
-
-Use Nginx as a reverse proxy in front of Next.js.
-
-#### Nginx Configuration:
-
-```nginx
-server {
-    listen 80;
-    server_name schoolcompare.co.uk;
-
-    # Redirect to HTTPS
-    return 301 https://$server_name$request_uri;
-}
-
-server {
-    listen 443 ssl http2;
-    server_name schoolcompare.co.uk;
-
-    # SSL Configuration
-    ssl_certificate /etc/ssl/certs/schoolcompare.crt;
-    ssl_certificate_key /etc/ssl/private/schoolcompare.key;
-
-    # Security Headers
-    # frame-ancestors replaces X-Frame-Options so the analytics subdomain
-    # (Umami heatmap/recorder) can embed the site in an iframe.
-    add_header Content-Security-Policy "frame-ancestors 'self' https://analytics.schoolcompare.co.uk" always;
-    add_header X-Content-Type-Options "nosniff" always;
-    add_header X-XSS-Protection "1; mode=block" always;
-
-    # Proxy to Next.js
-    location / {
-        proxy_pass http://localhost:3000;
-        proxy_http_version 1.1;
-        proxy_set_header Upgrade $http_upgrade;
-        proxy_set_header Connection 'upgrade';
-        proxy_set_header Host $host;
-        proxy_cache_bypass $http_upgrade;
-        proxy_set_header X-Real-IP $remote_addr;
-        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-    }
-
-    # Proxy to FastAPI
-    location /api/ {
-        proxy_pass http://localhost:8000;
-        proxy_http_version 1.1;
-        proxy_set_header Host $host;
-        proxy_set_header X-Real-IP $remote_addr;
-        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-    }
-
-    # Cache static files
-    location /_next/static/ {
-        proxy_pass http://localhost:3000;
-        add_header Cache-Control "public, max-age=31536000, immutable";
-    }
-}
-```
-
----
-
-## Pre-Deployment Checklist
-
-- [ ] Run `npm run build` successfully
-- [ ] Run `npm test` - all tests pass
-- [ ] Environment variables configured
-- [ ] FastAPI backend accessible
-- [ ] Database migrations applied
-- [ ] SSL certificates configured (production)
-- [ ] Domain DNS configured
-- [ ] Monitoring/logging set up
-- [ ] Backup strategy in place
-
----
-
-## Post-Deployment Verification
-
-1. **Health Check**:
-   ```bash
-   curl https://schoolcompare.co.uk
-   ```
-
-2. **Test Routes**:
-   - Home: `https://schoolcompare.co.uk/`
-   - School Page: `https://schoolcompare.co.uk/school/100001`
-   - Compare: `https://schoolcompare.co.uk/compare`
-   - Rankings: `https://schoolcompare.co.uk/rankings`
-
-3. **Check SEO**:
-   - Sitemap: `https://schoolcompare.co.uk/sitemap.xml`
-   - Robots: `https://schoolcompare.co.uk/robots.txt`
-
-4. **Performance Audit**:
-   - Run Lighthouse in Chrome DevTools
-   - Target scores: 90+ for Performance, Accessibility, Best Practices, SEO
-
----
-
-## Monitoring
-
-### Recommended Tools:
-- **Vercel Analytics** (if using Vercel)
-- **Sentry** for error tracking
-- **Google Analytics** for user analytics
-- **Uptime Robot** for uptime monitoring
-
-### Health Check Endpoint:
-The application automatically serves health data at the root route.
-
----
-
-## Rollback Procedure
-
-### Vercel:
-```bash
-vercel rollback
-```
-
-### Docker:
-```bash
-docker-compose down
-docker-compose up -d --force-recreate
-```
-
-### PM2:
-```bash
-pm2 stop schoolcompare-nextjs
-# Restore previous build
-pm2 start schoolcompare-nextjs
-```
-
----
-
-## Troubleshooting
-
-### Issue: API requests failing
-- **Solution**: Check `NEXT_PUBLIC_API_URL` and `FASTAPI_URL` environment variables
-- **Verify**: FastAPI backend is accessible from Next.js container/server
-
-### Issue: Build fails
-- **Solution**: Check Node.js version (requires 24+)
-- **Clear cache**: `rm -rf .next node_modules && npm install && npm run build`
-
-### Issue: Slow page loads
-- **Solution**: Enable caching in API calls
-- **Check**: Network latency to FastAPI backend
-- **Verify**: CDN is serving static assets
-
----
-
-## Security Considerations
-
-- ✅ HTTPS enabled
-- ✅ Security headers configured (X-Frame-Options, CSP, etc.)
-- ✅ API keys in environment variables (never in code)
-- ✅ CORS properly configured
-- ✅ Rate limiting on API endpoints
-- ✅ Regular security updates
-- ✅ Dependency vulnerability scanning
-
----
-
-## Support
-
-For deployment issues, contact the DevOps team or refer to:
-- [Next.js Deployment Docs](https://nextjs.org/docs/deployment)
-- [Vercel Documentation](https://vercel.com/docs)
-- [Docker Documentation](https://docs.docker.com/)
+Earlier Vercel and standalone deployment recipes have been retired from this file
+because they do not describe the current CMS, persistence and promotion setup.
+See [development](../docs/DEVELOPMENT.md) for checks and
+[publishing](docs/PUBLISHING.md) for CMS operations.

nextjs-app/Dockerfile

@@ -28,6 +28,10 @@ ENV NODE_ENV=production
 ARG FASTAPI_URL=http://backend:80/api
 ENV FASTAPI_URL=${FASTAPI_URL}
 
+ARG BUILD_SHA=development
+ARG BUILD_ID=development
+RUN node -e 'require("fs").writeFileSync("build-info.json", JSON.stringify({sha:process.argv[1],build_id:process.argv[2]}))' "$BUILD_SHA" "$BUILD_ID"
+
 # Build application
 RUN npm run build
 
@@ -70,6 +74,12 @@ USER nextjs
 EXPOSE 3000
 
 # Set environment variables
+ARG BUILD_SHA=development
+ARG BUILD_ID=development
+LABEL io.schoolcompare.build-id=$BUILD_ID
+LABEL io.schoolcompare.commit=$BUILD_SHA
+COPY --from=builder /app/build-info.json ./build-info.json
+
 ENV PORT=3000
 ENV HOSTNAME="0.0.0.0"
 

nextjs-app/README.md

@@ -1,156 +1,58 @@
-# SchoolCompare Next.js Application
+# SchoolCompare frontend and CMS
 
-Modern Next.js application for comparing primary school KS2 performance across England.
+Next.js App Router with React, TypeScript, CSS Modules, Chart.js, Leaflet and
+Payload CMS. It serves school search, comparisons, rankings, school/place detail
+pages and editorial content across England.
 
-## Features
+Start with the [repository overview](../README.md),
+[architecture](../docs/ARCHITECTURE.md) and [development checks](../docs/DEVELOPMENT.md).
 
-- **Server-Side Rendering (SSR)**: Fast initial page loads with pre-rendered content
-- **Individual School Pages**: Dedicated pages for each school with full SEO optimization
-- **Side-by-Side Comparison**: Compare up to 5 schools simultaneously
-- **School Rankings**: Top-performing schools by various metrics
-- **Interactive Maps**: Leaflet integration for geographic visualization
-- **Performance Charts**: Chart.js visualizations for historical data
-- **Responsive Design**: Mobile-first approach with full responsive support
-- **SEO Optimized**: Dynamic sitemaps, meta tags, and structured data
+## Source map
 
-## Tech Stack
+| Path | Purpose |
+|---|---|
+| `app/(frontend)/` | Public root layout, server pages and FastAPI proxy |
+| `app/(payload)/` | Payload root layout, `/admin` and `/cms-api` |
+| `app/robots.ts`, `app/opengraph-image.tsx`, root icons | Site-wide metadata endpoints |
+| `components/` | Client views and reusable display components |
+| `components/school/` | School detail sections |
+| `lib/api.ts`, `lib/types.ts` | Fetch wrappers and manual school API types |
+| `lib/schoolSections.ts`, `lib/compareLogic.ts` | Presentation decisions and data preparation |
+| `context/`, `hooks/` | Comparison state, suggestion state and responsive behaviour |
+| `collections/`, `blocks/`, `migrations/` | CMS schema and production migrations |
+| `__tests__/` | Jest and React Testing Library tests |
 
-- **Framework**: Next.js 16 (App Router)
-- **Language**: TypeScript 5
-- **Styling**: CSS Modules + CSS Variables
-- **State Management**: React Context API + URL state
-- **Data Fetching**: SWR (client-side) + Next.js fetch (server-side)
-- **Charts**: Chart.js + react-chartjs-2
-- **Maps**: Leaflet + react-leaflet
-- **Testing**: Jest + React Testing Library
-- **Validation**: Zod
+Do not introduce a shared `app/layout.tsx`: public pages and Payload have separate
+root layouts. Keep root metadata files outside the route groups.
 
-## Getting Started
+## Data and state
 
-### Prerequisites
+Server pages fetch initial data directly from `FASTAPI_URL`. Browser fetches use
+`/api` by default, forwarded by `app/(frontend)/api/[...path]/route.ts`.
+`FASTAPI_URL` must include `/api`. See `.env.example` for CMS and API settings.
 
-- Node.js 24+ (using nvm recommended)
-- FastAPI backend running on port 8000
+State uses React hooks/context, URL search parameters and localStorage for the
+comparison basket. SWR is not installed. Maps use dynamic Leaflet wrappers.
+Revalidation intervals are configured in fetch wrappers and pages; they vary by
+resource. Backend reloads do not automatically invalidate every Next.js cache.
 
-### Installation
+## Commands
 
-```bash
-# Install dependencies
-npm install
-
-# Copy environment variables
-cp .env.example .env.local
-
-# Update .env.local with your configuration
-```
-
-### Development
-
-```bash
-# Start development server
-npm run dev
-
-# Open http://localhost:3000
-```
-
-### Building
-
-```bash
-# Build for production
+```sh
+npm ci
+npm run typecheck
+npm test -- --runInBand
 npm run build
-
-# Start production server
-npm start
 ```
 
-### Testing
+`test:watch` and `test:coverage` are also available. There is no `lint` script.
+A running application needs the backend/data environment described in the
+[development guide](../docs/DEVELOPMENT.md).
 
-```bash
-# Run tests
-npm test
+After CMS field or editor changes, run `npm run generate:importmap`. Keep
+`payload-types.ts` generated from the CMS schema rather than editing it by hand.
+The build must work without a database connection; avoid module-scope CMS queries
+and DB-backed `generateStaticParams` functions.
 
-# Run tests in watch mode
-npm run test:watch
-
-# Run tests with coverage
-npm run test:coverage
-```
-
-### Linting
-
-```bash
-# Run ESLint
-npm run lint
-```
-
-## Project Structure
-
-```
-nextjs-app/
-├── app/                      # App Router pages
-│   ├── layout.tsx           # Root layout
-│   ├── page.tsx             # Home page
-│   ├── compare/             # Compare page
-│   ├── rankings/            # Rankings page
-│   ├── school/[urn]/        # Individual school pages
-│   ├── sitemap.ts           # Dynamic sitemap
-│   └── robots.ts            # Robots.txt
-├── components/              # React components
-│   ├── SchoolCard.tsx       # School card component
-│   ├── FilterBar.tsx        # Search/filter controls
-│   ├── ComparisonView.tsx   # Comparison interface
-│   ├── RankingsView.tsx     # Rankings table
-│   └── ...
-├── lib/                     # Utility libraries
-│   ├── api.ts              # API client
-│   ├── types.ts            # TypeScript types
-│   └── utils.ts            # Helper functions
-├── hooks/                   # Custom React hooks
-├── context/                 # React Context providers
-├── styles/                  # Global styles
-├── public/                  # Static assets
-└── __tests__/              # Test files
-```
-
-## Environment Variables
-
-| Variable | Description | Default |
-|----------|-------------|---------|
-| `NEXT_PUBLIC_API_URL` | Public API endpoint (client-side) | `http://localhost:8000/api` |
-| `FASTAPI_URL` | Server-side API endpoint | `http://localhost:8000/api` |
-| `NODE_ENV` | Environment mode | `development` |
-
-## Performance Optimizations
-
-- **Server-Side Rendering**: Initial HTML rendered on server
-- **Static Generation**: Where possible, pages are pre-generated
-- **Image Optimization**: Next.js Image component with AVIF/WebP support
-- **Code Splitting**: Automatic route-based code splitting
-- **Dynamic Imports**: Heavy components loaded on demand
-- **API Caching**: Configurable revalidation for data fetching
-- **Bundle Optimization**: Tree shaking and minification
-- **Compression**: Gzip compression enabled
-
-## SEO Features
-
-- **Dynamic Meta Tags**: Generated per page with Next.js Metadata API
-- **Open Graph**: Social media optimization
-- **JSON-LD**: Structured data for search engines
-- **Sitemap**: Auto-generated from database
-- **Robots.txt**: Search engine crawling rules
-- **Canonical URLs**: Duplicate content prevention
-
-## Browser Support
-
-- Chrome (latest)
-- Firefox (latest)
-- Safari (latest)
-- Edge (latest)
-
-## License
-
-Proprietary - SchoolCompare
-
-## Support
-
-For issues and questions, please contact the development team.
+See [publishing](docs/PUBLISHING.md) for CMS operations and
+[deployment](../docs/DEPLOY.md) for staging and production promotion.

nextjs-app/__tests__/app/dataFailures.test.tsx

@@ -0,0 +1,51 @@
+import { fireEvent, render, screen } from '@testing-library/react';
+import HomePage from '@/app/(frontend)/page';
+import SchoolPage from '@/app/(frontend)/school/[slug]/page';
+import ErrorPage from '@/app/(frontend)/error';
+import { APIFetchError, fetchSchools, fetchFilters, fetchSchoolDetails } from '@/lib/api';
+import { fetchPlace, fetchPlaces } from '@/lib/places';
+
+jest.mock('@/lib/api', () => ({
+  ...jest.requireActual('@/lib/api'),
+  fetchSchools: jest.fn(),
+  fetchSchoolDetails: jest.fn(),
+  fetchFilters: jest.fn(async () => ({})),
+  fetchDataInfo: jest.fn(async () => null),
+  fetchNationalAverages: jest.fn(async () => null),
+}));
+jest.mock('@/lib/flags', () => ({ getFlags: jest.fn(async () => ({})) }));
+jest.mock('next/navigation', () => ({
+  notFound: () => { throw new Error('NEXT_NOT_FOUND'); },
+  redirect: jest.fn(),
+}));
+const realFetch = global.fetch;
+afterEach(() => { global.fetch = realFetch; jest.clearAllMocks(); });
+
+test('school outages propagate; only a real 404 becomes not found', async () => {
+  const request = { params: Promise.resolve({ slug: '100001-school' }) };
+  const outage = new APIFetchError('unavailable', 503);
+  jest.mocked(fetchSchoolDetails).mockRejectedValueOnce(outage);
+  await expect(SchoolPage(request)).rejects.toBe(outage);
+  jest.mocked(fetchSchoolDetails).mockRejectedValueOnce(new APIFetchError('missing', 404));
+  await expect(SchoolPage(request)).rejects.toThrow('NEXT_NOT_FOUND');
+});
+
+test('homepage search failure is not returned as an empty successful page', async () => {
+  jest.mocked(fetchSchools).mockRejectedValueOnce(new APIFetchError('unavailable', 503));
+  await expect(HomePage({ searchParams: Promise.resolve({ search: 'school' }) })).rejects.toThrow('unavailable');
+});
+
+test('a place is absent only on 404; other failures propagate', async () => {
+  global.fetch = jest.fn().mockResolvedValue({ ok: false, status: 404 });
+  await expect(fetchPlace('town', 'example')).resolves.toBeNull();
+  jest.mocked(global.fetch).mockResolvedValue({ ok: false, status: 503 } as Response);
+  await expect(fetchPlace('town', 'example')).rejects.toMatchObject({ status: 503 });
+  await expect(fetchPlaces()).rejects.toMatchObject({ status: 503 });
+});
+
+test('the error boundary offers a retry without showing an empty search', () => {
+  const reset = jest.fn();
+  render(<ErrorPage error={new Error('offline')} reset={reset} />);
+  fireEvent.click(screen.getByRole('button', { name: 'Try again' }));
+  expect(reset).toHaveBeenCalledTimes(1);
+});

nextjs-app/__tests__/app/releaseIdentity.test.ts

@@ -0,0 +1,30 @@
+/** @jest-environment node */
+import { GET } from '@/app/(frontend)/release.json/route';
+import { readFile } from 'node:fs/promises';
+
+jest.mock('node:fs/promises', () => ({ readFile: jest.fn() }));
+const realFetch = global.fetch;
+const identity = { sha: 'a'.repeat(40), build_id: 'b'.repeat(32) };
+beforeEach(() => {
+  jest.mocked(readFile).mockResolvedValue(JSON.stringify(identity));
+  global.fetch = jest.fn(async () => Response.json(identity));
+});
+afterEach(() => { global.fetch = realFetch; jest.resetAllMocks(); });
+
+test('reports immutable file identity and backend identity without caching', async () => {
+  const response = await GET();
+  expect(response.status).toBe(200);
+  expect(response.headers.get('Cache-Control')).toBe('no-store');
+  expect(await response.json()).toEqual({ frontend: identity, backend: identity });
+  expect(fetch).toHaveBeenCalledWith(expect.stringMatching(/\/api\/release$/), expect.objectContaining({ cache: 'no-store', signal: expect.anything() }));
+});
+
+test('missing build metadata cannot pass the release gate', async () => {
+  jest.mocked(readFile).mockRejectedValueOnce(new Error('missing file'));
+  expect((await GET()).status).toBe(503);
+});
+
+test('backend failure cannot pass the release gate', async () => {
+  jest.mocked(fetch).mockResolvedValueOnce(new Response('', { status: 503 }));
+  expect((await GET()).status).toBe(503);
+});

nextjs-app/__tests__/components/HomeView.staleFetch.test.tsx

@@ -0,0 +1,81 @@
+import { act, fireEvent, render, screen } from '@testing-library/react';
+import { HomeView } from '@/components/HomeView';
+import { fetchSchools } from '@/lib/api';
+import { primaryFixture } from '../support/schoolFixtures';
+import type { SchoolsResponse, School } from '@/lib/types';
+
+let params = new URLSearchParams('postcode=SW1A+1AA');
+jest.mock('next/navigation', () => ({
+  useSearchParams: () => params,
+  usePathname: () => '/',
+  useRouter: () => ({ push: jest.fn(), replace: jest.fn() }),
+}));
+jest.mock('@/context/ComparisonContext', () => ({
+  useComparisonContext: () => ({ addSchool: jest.fn(), removeSchool: jest.fn(), selectedSchools: [] }),
+}));
+jest.mock('@/lib/api', () => ({
+  fetchSchools: jest.fn(),
+  fetchNationalAverages: jest.fn(async () => ({})),
+  fetchLAaverages: jest.fn(async () => ({ secondary: { attainment_8_by_la: {} } })),
+}));
+jest.mock('@/components/FilterBar', () => ({ FilterBar: () => null }));
+jest.mock('@/components/SchoolRow', () => ({ SchoolRow: ({ school }: {school: School}) => <div>{school.school_name}</div> }));
+jest.mock('@/components/SchoolMap', () => ({ SchoolMap: ({ schools }: {schools: School[]}) => <div data-testid="map">{schools.map(s => s.school_name).join(',')}</div> }));
+
+const filters = { local_authorities: [], school_types: [], years: [], phases: [], genders: [], admissions_policies: [] };
+function response(name: string): SchoolsResponse {
+  return { schools: [{ ...primaryFixture.schoolInfo, school_name: name }],
+    total: 2, page: 1, page_size: 1, total_pages: 2 };
+}
+function deferred() {
+  let resolve!: (value: SchoolsResponse) => void;
+  let reject!: (error: Error) => void;
+  const promise = new Promise<SchoolsResponse>((yes, no) => { resolve = yes; reject = no; });
+  return { promise, resolve, reject };
+}
+beforeEach(() => {
+  params = new URLSearchParams('postcode=SW1A+1AA');
+  jest.mocked(fetchSchools).mockReset();
+});
+
+test('load-more results from an old search are discarded, even after returning to it', async () => {
+  const pending = deferred();
+  jest.mocked(fetchSchools).mockReturnValueOnce(pending.promise);
+  const view = render(<HomeView initialSchools={response('Initial A')} filters={filters} />);
+  fireEvent.click(screen.getByRole('button', { name: 'Load more schools' }));
+  const signal = jest.mocked(fetchSchools).mock.calls[0][1]?.signal;
+  params = new URLSearchParams('postcode=SW2+1AA');
+  view.rerender(<HomeView initialSchools={response('Initial B')} filters={filters} />);
+  expect(signal?.aborted).toBe(true);
+  params = new URLSearchParams('postcode=SW1A+1AA');
+  view.rerender(<HomeView initialSchools={response('Fresh A')} filters={filters} />);
+  await act(async () => pending.resolve(response('Stale append')));
+  expect(screen.queryByText('Stale append')).not.toBeInTheDocument();
+  expect(screen.getByRole('button', { name: 'Load more schools' })).toBeEnabled();
+});
+
+test('an older map response cannot overwrite the current search', async () => {
+  const first = deferred(), second = deferred();
+  jest.mocked(fetchSchools).mockReturnValueOnce(first.promise).mockReturnValueOnce(second.promise);
+  const initial = response('Initial A');
+  const view = render(<HomeView initialSchools={initial} filters={filters} />);
+  fireEvent.click(screen.getByRole('button', { name: 'Map' }));
+  params = new URLSearchParams('postcode=SW2+1AA');
+  view.rerender(<HomeView initialSchools={response('Initial B')} filters={filters} />);
+  await act(async () => second.resolve(response('Current map')));
+  await act(async () => first.resolve(response('Stale map')));
+  expect(screen.getByTestId('map')).toHaveTextContent('Current map');
+  expect(screen.getByTestId('map')).not.toHaveTextContent('Stale map');
+});
+
+test('failed map requests can be retried by reopening the map', async () => {
+  const pending = deferred();
+  jest.mocked(fetchSchools).mockReturnValueOnce(pending.promise).mockResolvedValue(response('Retry result'));
+  render(<HomeView initialSchools={response('Initial')} filters={filters} />);
+  fireEvent.click(screen.getByRole('button', { name: 'Map' }));
+  await act(async () => pending.reject(new Error('offline')));
+  fireEvent.click(screen.getByRole('button', { name: 'List' }));
+  await act(async () => fireEvent.click(screen.getByRole('button', { name: 'Map' })));
+  expect(fetchSchools).toHaveBeenCalledTimes(2);
+  expect(screen.getByTestId('map')).toHaveTextContent('Retry result');
+});

nextjs-app/__tests__/components/NearbySchoolsSection.test.tsx

@@ -0,0 +1,192 @@
+/**
+ * The section's job is to be honest about what it is showing. These tests pin
+ * the ways it could mislead: rendering below the minimum, claiming a likeness
+ * it does not rank on, showing a missing figure as a number, or hiding a card
+ * behind an arrow where a crawler cannot reach it.
+ */
+
+import { render, screen } from '@testing-library/react';
+
+import {
+  nearbyNoun,
+  NearbySchoolsSection,
+  shouldRenderNearby,
+} from '@/components/school/NearbySchoolsSection';
+import type { NearbySchool } from '@/lib/types';
+
+jest.mock('@/components/school/AddToCompareButton', () => ({
+  AddToCompareButton: ({ school }: { school: NearbySchool }) => (
+    <button type="button">Add {school.school_name} to compare</button>
+  ),
+}));
+
+jest.mock('@/components/school/NearbySchoolsCompareBar', () => ({
+  NearbySchoolsCompareBar: ({ thisUrn }: { thisUrn: number }) => (
+    <div data-testid="compare-bar">bar for {thisUrn}</div>
+  ),
+}));
+
+function school(overrides: Partial<NearbySchool> = {}): NearbySchool {
+  return {
+    urn: 100002,
+    school_name: 'Willow Lane Primary School',
+    distance_miles: 0.6,
+    school_type: 'Community school',
+    age_range: '4-11',
+    shared: ['Mixed', 'No religious character'],
+    metric_value: 74,
+    metric_key: 'rwm_expected_pct',
+    metric_year: 202425,
+    ...overrides,
+  };
+}
+
+function renderSection(nearby: NearbySchool[]) {
+  return render(
+    <NearbySchoolsSection
+      urn={100001}
+      schoolName="Meadowbrook Primary School"
+      phase="Primary"
+      thisMetricValue={72}
+      nearby={nearby}
+    />,
+  );
+}
+
+describe('render gates', () => {
+  it.each([
+    ['undefined', undefined],
+    ['null', null],
+    ['empty', []],
+    ['a single school', [school()]],
+  ])('renders nothing for %s', (_label, value) => {
+    expect(shouldRenderNearby(value as NearbySchool[] | null | undefined)).toBe(false);
+  });
+
+  it('renders for two or more schools', () => {
+    expect(shouldRenderNearby([school(), school({ urn: 100003 })])).toBe(true);
+  });
+
+  it('returns null rather than an empty shell below the minimum', () => {
+    const { container } = renderSection([school()]);
+    expect(container).toBeEmptyDOMElement();
+  });
+});
+
+describe('what the section claims', () => {
+  it('never claims a similar intake, because it does not rank on one', () => {
+    renderSection([school(), school({ urn: 100003, shared: [] })]);
+    expect(screen.queryByText(/similar intake/i)).not.toBeInTheDocument();
+  });
+
+  it('is headed "Other schools nearby", not "similar"', () => {
+    renderSection([school(), school({ urn: 100003 })]);
+    expect(screen.getByRole('heading', { name: 'Other schools nearby' })).toBeInTheDocument();
+  });
+
+  it('shows chips for what is shared', () => {
+    renderSection([school({ shared: ['Mixed', 'Roman Catholic'] }), school({ urn: 100003 })]);
+    expect(screen.getAllByText('Roman Catholic').length).toBe(1);
+  });
+
+  it('shows no chips at all when nothing is shared, rather than inventing one', () => {
+    const { container } = render(
+      <NearbySchoolsSection
+        urn={100001}
+        schoolName="Meadowbrook Primary School"
+        phase="Primary"
+        thisMetricValue={72}
+        nearby={[school({ shared: [] }), school({ urn: 100003, shared: [] })]}
+      />,
+    );
+    // The card still carries its distance, name, type and figure — just no
+    // claim of likeness.
+    expect(container.querySelectorAll('li ul').length).toBe(0);
+    expect(screen.getAllByText(/miles away/).length).toBe(2);
+  });
+});
+
+describe('what the lede calls the set', () => {
+  it.each([
+    ['Primary', 'primary schools'],
+    ['Middle deemed primary', 'primary schools'],
+    ['Secondary', 'secondary schools'],
+    ['Middle deemed secondary', 'secondary schools'],
+    ['All-through', 'all-through schools'],
+    // GIAS phase 6. Its candidates span the whole secondary group, so no
+    // single noun fits and it takes the honest general one.
+    ['16 plus', 'schools and colleges'],
+    ['', 'schools'],
+    [null, 'schools'],
+  ])('calls a %s school\'s neighbours "%s"', (phase, expected) => {
+    expect(nearbyNoun(phase)).toBe(expected);
+  });
+
+  it('never calls a sixth form college\'s neighbours primary schools', () => {
+    render(
+      <NearbySchoolsSection
+        urn={100001}
+        schoolName="Barnet Sixth Form College"
+        phase="16 plus"
+        thisMetricValue={null}
+        nearby={[school(), school({ urn: 100003 })]}
+      />,
+    );
+    expect(screen.getByText(/Other schools and colleges near Barnet Sixth Form College/)).toBeInTheDocument();
+    expect(screen.queryByText(/primary schools/)).not.toBeInTheDocument();
+  });
+});
+
+describe('cards', () => {
+  it('links each school to its canonical slug', () => {
+    renderSection([school(), school({ urn: 100003, school_name: 'Oakfield Primary School' })]);
+    const link = screen.getByRole('link', { name: /Willow Lane Primary School/ });
+    expect(link).toHaveAttribute('href', '/school/100002-willow-lane-primary-school');
+  });
+
+  it('shows the distance and the shared characteristics', () => {
+    renderSection([school(), school({ urn: 100003 })]);
+    expect(screen.getAllByText('0.6 miles away').length).toBeGreaterThan(0);
+    expect(screen.getAllByText('Mixed').length).toBeGreaterThan(0);
+  });
+
+  it('renders a missing figure as "Not published", never as a number', () => {
+    renderSection([school({ metric_value: null }), school({ urn: 100003 })]);
+    expect(screen.getByText('Not published')).toBeInTheDocument();
+  });
+
+  it('anchors each figure against this school', () => {
+    renderSection([school(), school({ urn: 100003 })]);
+    expect(screen.getAllByText('72% at this school').length).toBe(2);
+  });
+
+  it('offers the compare bar once, for this school', () => {
+    renderSection([school(), school({ urn: 100003 })]);
+    expect(screen.getByTestId('compare-bar')).toHaveTextContent('bar for 100001');
+  });
+
+  it('keeps every card in the DOM, including the ones scrolled out of view', () => {
+    const six = Array.from({ length: 6 }, (_, n) =>
+      school({ urn: 100002 + n, school_name: `Peer ${n} School` }),
+    );
+    renderSection(six);
+    expect(screen.getAllByRole('link', { name: /Peer \d School/ })).toHaveLength(6);
+  });
+
+  it('offers no arrows when three cards fit the row', () => {
+    renderSection([school(), school({ urn: 100003 }), school({ urn: 100004 })]);
+    expect(screen.queryByRole('button', { name: /More schools/ })).not.toBeInTheDocument();
+  });
+
+  it('offers arrows once there is a fourth school', () => {
+    renderSection(Array.from({ length: 4 }, (_, n) => school({ urn: 100002 + n })));
+    expect(screen.getByRole('button', { name: /More schools/ })).toBeInTheDocument();
+    expect(screen.getByRole('button', { name: /Previous schools/ })).toBeInTheDocument();
+  });
+
+  it('says distances are straight-line, and offers no method panel', () => {
+    const { container } = renderSection([school(), school({ urn: 100003 })]);
+    expect(screen.getByText(/straight-line from this school/i)).toBeInTheDocument();
+    expect(container.querySelector('details')).toBeNull();
+  });
+});

nextjs-app/__tests__/lib/schoolSections.test.ts

@@ -172,3 +172,39 @@ describe('buildSecondaryNavItems', () => {
     expect(ids).not.toContain('history');
   });
 });
+
+describe('the nearby-schools nav item', () => {
+  const navInput = {
+    ofsted: null, admissions: null, admissionDistance: null,
+    hasLocation: true, yearlyDataLength: 1,
+  };
+
+  it('appears on both templates when the section renders', () => {
+    const primary = computeSchoolFlags(primaryFixture);
+    const secondary = computeSecondaryFlags(secondaryFixture);
+    const input = { ...navInput, hasNearbySchools: true };
+
+    expect(buildNavItems(primary, input).map((i) => i.id)).toContain('nearby');
+    expect(buildSecondaryNavItems(secondary, input).map((i) => i.id)).toContain('nearby');
+  });
+
+  it('is absent when the section does not render', () => {
+    const primary = computeSchoolFlags(primaryFixture);
+    const secondary = computeSecondaryFlags(secondaryFixture);
+    const input = { ...navInput, hasNearbySchools: false };
+
+    expect(buildNavItems(primary, input).map((i) => i.id)).not.toContain('nearby');
+    expect(buildSecondaryNavItems(secondary, input).map((i) => i.id)).not.toContain('nearby');
+  });
+
+  it('is absent when nothing says either way', () => {
+    const primary = computeSchoolFlags(primaryFixture);
+    expect(buildNavItems(primary, navInput).map((i) => i.id)).not.toContain('nearby');
+  });
+
+  it('comes last, because the section renders last', () => {
+    const primary = computeSchoolFlags(primaryFixture);
+    const ids = buildNavItems(primary, { ...navInput, hasNearbySchools: true }).map((i) => i.id);
+    expect(ids[ids.length - 1]).toBe('nearby');
+  });
+});

nextjs-app/__tests__/payload/routes.test.ts

@@ -38,8 +38,7 @@ describe('payload mount points', () => {
   });
 
   it('isolates CMS tables in their own postgres schema', () => {
-    // Blog content must sit outside `public`, where the app tables, Airflow's
-    // metadata and scripts/migrate_csv_to_db.py --drop all live.
+    // Blog content must stay separate from school marts and Airflow metadata.
     expect(CONFIG).toMatch(/schemaName:\s*['"]payload['"]/);
   });
 });

nextjs-app/app/(frontend)/error.tsx

@@ -0,0 +1,12 @@
+'use client';
+
+export default function ErrorPage({ reset }: { error: Error & { digest?: string }; reset: () => void }) {
+  return (
+    <main style={{ maxWidth: '48rem', margin: '4rem auto', padding: '1.5rem' }}>
+      <h1>We couldn’t load this page</h1>
+      <p>School information is temporarily unavailable. Please try again.</p>
+      <button type="button" onClick={reset}>Try again</button>
+      <p><a href="/">Return to school search</a></p>
+    </main>
+  );
+}

nextjs-app/app/(frontend)/page.tsx

@@ -85,73 +85,50 @@ export default async function HomePage({ searchParams }: HomePageProps) {
     params.has_sixth_form
   );
 
-  // Fetch data on server with error handling
-  try {
-    const [filtersData, dataInfo] = await Promise.all([fetchFilters(), fetchDataInfo().catch(() => null)]);
+  // Failures propagate to the retryable error boundary.
+  const [filtersData, dataInfo] = await Promise.all([fetchFilters(), fetchDataInfo().catch(() => null)]);
 
-    // Only fetch schools if there are search parameters
-    let schoolsData;
-    if (hasSearchParams) {
-      schoolsData = await fetchSchools({
-        search: params.search,
-        local_authority: params.local_authority,
-        school_type: params.school_type,
-        phase: params.phase,
-        postcode: params.postcode,
-        radius,
-        page,
-        page_size: 50,
-        gender: params.gender,
-        admissions_policy: params.admissions_policy,
-        has_sixth_form: params.has_sixth_form,
-      });
-    } else {
-      // Empty state by default
-      schoolsData = { schools: [], page: 1, page_size: 50, total: 0, total_pages: 0 };
-    }
-
-    const resolvedFilters = filtersData || { local_authorities: [], school_types: [], years: [], phases: [], genders: [], admissions_policies: [] };
-    // `unique_schools`, not `total_schools` — the latter is not a field this
-    // endpoint returns, and reading it silently yielded null on every request.
-    const total = dataInfo?.unique_schools ?? null;
-    const years = dataInfo?.years_available ?? [];
-    return (
-      <HomeView
-        autosuggest={autosuggest}
-        initialSchools={schoolsData}
-        filters={resolvedFilters}
-        totalSchools={total}
-        howItWorks={hasSearchParams ? null : <HowItWorksSection />}
-        editorial={hasSearchParams ? null : (
-          <EditorialSection
-            totalSchools={total}
-            localAuthorityCount={resolvedFilters.local_authorities.length}
-            earliestYearLabel={years.length ? formatAcademicYear(years[0]) : null}
-            latestYearLabel={years.length ? formatAcademicYear(years[years.length - 1]) : null}
-          />
-        )}
-      />
-    );
-  } catch (error) {
-    console.error('Error fetching data for home page:', error);
-
-    const emptyFilters = { local_authorities: [], school_types: [], years: [], phases: [], genders: [], admissions_policies: [] };
-    return (
-      <HomeView
-        autosuggest={autosuggest}
-        initialSchools={{ schools: [], page: 1, page_size: 50, total: 0, total_pages: 0 }}
-        filters={emptyFilters}
-        totalSchools={null}
-        howItWorks={hasSearchParams ? null : <HowItWorksSection />}
-        editorial={hasSearchParams ? null : (
-          <EditorialSection
-            totalSchools={null}
-            localAuthorityCount={0}
-            earliestYearLabel={null}
-            latestYearLabel={null}
-          />
-        )}
-      />
-    );
+  // Only fetch schools if there are search parameters
+  let schoolsData;
+  if (hasSearchParams) {
+    schoolsData = await fetchSchools({
+      search: params.search,
+      local_authority: params.local_authority,
+      school_type: params.school_type,
+      phase: params.phase,
+      postcode: params.postcode,
+      radius,
+      page,
+      page_size: 50,
+      gender: params.gender,
+      admissions_policy: params.admissions_policy,
+      has_sixth_form: params.has_sixth_form,
+    });
+  } else {
+    // Empty state by default
+    schoolsData = { schools: [], page: 1, page_size: 50, total: 0, total_pages: 0 };
   }
+
+  const resolvedFilters = filtersData || { local_authorities: [], school_types: [], years: [], phases: [], genders: [], admissions_policies: [] };
+  // `unique_schools`, not `total_schools` — the latter is not a field this
+  // endpoint returns, and reading it silently yielded null on every request.
+  const total = dataInfo?.unique_schools ?? null;
+  const years = dataInfo?.years_available ?? [];
+  return (
+    <HomeView
+      autosuggest={autosuggest}
+      initialSchools={schoolsData}
+      filters={resolvedFilters}
+      totalSchools={total}
+      howItWorks={hasSearchParams ? null : <HowItWorksSection />}
+      editorial={hasSearchParams ? null : (
+        <EditorialSection
+          totalSchools={total}
+          localAuthorityCount={resolvedFilters.local_authorities.length}
+          earliestYearLabel={years.length ? formatAcademicYear(years[0]) : null}
+          latestYearLabel={years.length ? formatAcademicYear(years[years.length - 1]) : null}
+        />
+      )}
+    />
+  );
 }

nextjs-app/app/(frontend)/release.json/route.ts

@@ -0,0 +1,17 @@
+import { readFile } from 'node:fs/promises';
+import path from 'node:path';
+
+export const dynamic = 'force-dynamic';
+export const runtime = 'nodejs';
+
+export async function GET() {
+  try {
+    const frontend = JSON.parse(await readFile(path.join(process.cwd(), 'build-info.json'), 'utf8'));
+    const base = process.env.FASTAPI_URL || 'http://localhost:8000/api';
+    const res = await fetch(`${base}/release`, { cache: 'no-store', signal: AbortSignal.timeout(5000) });
+    if (!res.ok) throw new Error('Backend identity unavailable');
+    return Response.json({ frontend, backend: await res.json() }, { headers: { 'Cache-Control': 'no-store', 'X-Robots-Tag': 'noindex' } });
+  } catch {
+    return Response.json({ detail: 'Release identity unavailable' }, { status: 503, headers: { 'Cache-Control': 'no-store', 'X-Robots-Tag': 'noindex' } });
+  }
+}

nextjs-app/app/(frontend)/school/[slug]/page.tsx

@@ -4,10 +4,11 @@
  * URL format: /school/138267-school-name-here
  */
 
-import { fetchSchoolDetails, fetchSchools, fetchNationalAverages } from '@/lib/api';
+import { APIFetchError, fetchSchoolDetails, fetchSchools, fetchNationalAverages } from '@/lib/api';
 import { notFound, redirect } from 'next/navigation';
 import { SchoolDetailShell } from '@/components/school/SchoolDetailShell';
 import { NearbyPlaces } from '@/components/school/NearbyPlaces';
+import { shouldRenderNearby } from '@/components/school/NearbySchoolsSection';
 import { schoolBreadcrumbJsonLd, type SchoolPlace } from '@/lib/jsonld';
 import { PrimarySchoolSections } from '@/components/school/PrimarySchoolSections';
 import { SecondarySchoolSections } from '@/components/school/SecondarySchoolSections';
@@ -146,8 +147,8 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
       fetchNationalAverages().catch(() => null),
     ]);
   } catch (error) {
-    console.error(`Failed to fetch school ${urn}:`, error);
-    notFound();
+    if (error instanceof APIFetchError && error.status === 404) notFound();
+    throw error;
   }
 
   const { school_info, yearly_data, absence_data, ofsted, census, admissions, admissions_history, admission_distance, deprivation, finance, destinations } = data;
@@ -155,6 +156,8 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
   // nothing rather than throwing, which is how this shipped without a
   // lockstep deploy of the two images.
   const places: SchoolPlace[] = data.places ?? [];
+  // Absent on an older API build, exactly like `places` above.
+  const nearbySchools = data.nearby_schools ?? [];
 
   // Redirect bare URN to canonical slug URL
   const canonicalSlug = schoolUrl(urn, school_info.school_name).replace('/school/', '');
@@ -186,6 +189,7 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
     admissions: admissions ?? null,
     admissionDistance: admission_distance ?? null,
     hasLocation: school_info.latitude != null && school_info.longitude != null,
+    hasNearbySchools: shouldRenderNearby(nearbySchools),
     yearlyDataLength: yearly_data.length,
   };
   const primaryNavItems = buildNavItems(primaryFlags, navInput);
@@ -262,6 +266,7 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
             finance={finance ?? null}
             nationalAvg={nationalAvg}
             destinations={destinations ?? null}
+            nearbySchools={nearbySchools}
             flags={secondaryFlags}
           />
         </SchoolDetailShell>
@@ -284,6 +289,7 @@ export default async function SchoolPage({ params }: SchoolPageProps) {
             deprivation={deprivation ?? null}
             finance={finance ?? null}
             nationalAvg={nationalAvg}
+            nearbySchools={nearbySchools}
             flags={primaryFlags}
           />
         </SchoolDetailShell>

nextjs-app/components/HomeView.tsx

@@ -213,6 +213,16 @@ export function HomeView({ initialSchools, filters, totalSchools, howItWorks, ed
   const [isLoadingMap, setIsLoadingMap] = useState(false);
   const prevSearchParamsRef = useRef(searchParams.toString());
   const mapParamsRef = useRef<string>('');
+  const loadMoreController = useRef<AbortController | null>(null);
+  // Identity changes even for A → B → A, so an old A response stays stale.
+  const searchScope = useRef({ key: searchParams.toString() });
+  if (searchScope.current.key !== searchParams.toString()) {
+    searchScope.current = { key: searchParams.toString() };
+  }
+  useEffect(() => {
+    setIsLoadingMore(false);
+    return () => { loadMoreController.current?.abort(); };
+  }, [searchParams]);
   const [geoState, setGeoState] = useState<'idle' | 'requesting' | 'error'>('idle');
   const [geoError, setGeoError] = useState<string | null>(null);
   /*
@@ -274,17 +284,27 @@ export function HomeView({ initialSchools, filters, totalSchools, howItWorks, ed
     if (resultsView !== 'map' || !isLocationSearch) return;
     const paramsKey = searchParams.toString();
     if (paramsKey === mapParamsRef.current) return;
-    mapParamsRef.current = paramsKey;
+    const controller = new AbortController();
+    const scope = searchScope.current;
+    const current = () => !controller.signal.aborted && searchScope.current === scope;
     setIsLoadingMap(true);
     const params: Record<string, any> = {};
     searchParams.forEach((value, key) => { params[key] = value; });
     params.page = 1;
     params.page_size = 500;
-    fetchSchools(params, { cache: 'no-store' })
-      .then(r => setMapSchools(r.schools))
-      .catch(() => setMapSchools(initialSchools.schools))
-      .finally(() => setIsLoadingMap(false));
-  }, [resultsView, searchParams]);
+    fetchSchools(params, { cache: 'no-store', signal: controller.signal })
+      .then(r => {
+        if (!current()) return;
+        mapParamsRef.current = paramsKey;
+        setMapSchools(r.schools);
+      })
+      .catch(() => {
+        if (current()) setMapSchools(initialSchools.schools);
+        // No cache marker on failure: opening the map again retries.
+      })
+      .finally(() => { if (current()) setIsLoadingMap(false); });
+    return () => controller.abort();
+  }, [resultsView, searchParams, initialSchools.schools]);
 
   // Fetch LA averages when secondary or mixed schools are visible
   useEffect(() => {
@@ -305,19 +325,25 @@ export function HomeView({ initialSchools, filters, totalSchools, howItWorks, ed
     if (isLoadingMore || !hasMore) return;
     track('results_load_more', { next_page: currentPage + 1 });
     setIsLoadingMore(true);
+    const scope = searchScope.current;
+    const controller = new AbortController();
+    loadMoreController.current?.abort();
+    loadMoreController.current = controller;
+    const current = () => !controller.signal.aborted && searchScope.current === scope;
     try {
       const params: Record<string, any> = {};
       searchParams.forEach((value, key) => { params[key] = value; });
       params.page = currentPage + 1;
       params.page_size = initialSchools.page_size;
-      const response = await fetchSchools(params, { cache: 'no-store' });
+      const response = await fetchSchools(params, { cache: 'no-store', signal: controller.signal });
+      if (!current()) return;
       setAllSchools(prev => [...prev, ...response.schools]);
       setCurrentPage(response.page);
       setHasMore(response.page < response.total_pages);
     } catch {
       // silently ignore
     } finally {
-      setIsLoadingMore(false);
+      if (current()) setIsLoadingMore(false);
     }
   };
 

nextjs-app/components/school/AddToCompareButton.tsx

@@ -0,0 +1,45 @@
+'use client';
+
+/**
+ * The per-card basket toggle.
+ *
+ * The links around it are server-rendered, so the section works with JS off;
+ * this adds the basket interaction on top rather than being what makes the
+ * section function.
+ */
+
+import { useComparisonContext } from '@/context/ComparisonContext';
+import type { School, NearbySchool } from '@/lib/types';
+import styles from './NearbySchools.module.css';
+
+export function AddToCompareButton({ school }: { school: NearbySchool }) {
+  const { addSchool, removeSchool, selectedSchools } = useComparisonContext();
+  const selected = selectedSchools.some((s) => s.urn === school.urn);
+
+  const toggle = () => {
+    if (selected) {
+      removeSchool(school.urn);
+      return;
+    }
+    // The basket only needs identity and display fields; the compare page
+    // fetches everything it renders by URN.
+    addSchool({
+      urn: school.urn,
+      school_name: school.school_name,
+      school_type: school.school_type,
+      age_range: school.age_range,
+    } as School);
+  };
+
+  return (
+    <button
+      type="button"
+      className={styles.add}
+      onClick={toggle}
+      aria-pressed={selected}
+    >
+      {selected ? '✓ Added to compare' : '+ Add to compare'}
+      <span className={styles.srOnly}> — {school.school_name}</span>
+    </button>
+  );
+}

nextjs-app/components/school/NearbySchools.module.css

@@ -0,0 +1,64 @@
+.heading { font-family: var(--font-display); font-size: 1.4rem; letter-spacing: -0.4px; margin: 0; }
+.lede { margin: 0.5rem 0 1.25rem; color: var(--text-secondary); max-width: 64ch; }
+.caption { margin: 1rem 0 0; font-size: 0.72rem; color: var(--text-muted); }
+
+.top { display: flex; align-items: flex-start; justify-content: space-between; gap: 1rem; }
+.arrows { display: flex; gap: 0.5rem; flex: none; }
+.arrow { width: 44px; height: 44px; display: grid; place-items: center; cursor: pointer; border: 1px solid var(--border-strong); border-radius: 999px; background: var(--bg-card); color: var(--brand); }
+.arrow:hover:not(:disabled) { border-color: var(--brand); background: var(--brand-bg); }
+.arrow:disabled { opacity: 0.35; cursor: default; }
+.arrow svg { width: 17px; height: 17px; }
+
+/* A scroller, not a paginated view: every card is in the DOM and the arrows
+   only move the viewport across them. The 2px padding keeps focus rings from
+   being clipped — and is why the arrows' edge test needs a tolerance. */
+.scroller { display: grid; grid-auto-flow: column; grid-auto-columns: calc((100% - 1.8rem) / 3); gap: 0.9rem; overflow-x: auto; scroll-snap-type: x mandatory; padding: 2px; margin: -2px; list-style: none; scrollbar-width: none; -ms-overflow-style: none; }
+.scroller::-webkit-scrollbar { display: none; }
+@media (max-width: 820px) { .scroller { grid-auto-columns: calc((100% - 0.9rem) / 2); } }
+/* Touch widths (MOBILE.md): the arrows would take 96px from a 328px card and
+   crush the lede into four lines, for a control swiping already provides. They
+   go, and the documented right-edge fade carries the affordance — lifting at
+   the end of the travel, where there is nothing more to hint at. */
+@media (max-width: 640px) {
+  .top { display: block; }
+  .arrows { display: none; }
+  .scroller { grid-auto-columns: 86%; mask-image: linear-gradient(to right, #000 calc(100% - 28px), transparent); }
+  .scroller[data-at-end="true"] { mask-image: none; }
+}
+
+.school { position: relative; display: flex; flex-direction: column; scroll-snap-align: start; border: 1px solid var(--border); border-radius: 8px; padding: 1rem; background: var(--bg-card); }
+.school:hover { border-color: var(--border-strong); }
+
+.distance { margin: 0 0 0.6rem; font-size: 0.75rem; color: var(--text-muted); }
+.name { font-family: var(--font-display); font-size: 1rem; line-height: 1.35; margin: 0 0 0.35rem; }
+.name a { color: var(--text-primary); text-decoration: none; }
+/* The whole card is the link target; the button sits above it on z-index. */
+.name a::after { content: ""; position: absolute; inset: 0; border-radius: 8px; }
+.name a:hover { color: var(--brand); text-decoration: underline; }
+.meta { margin: 0 0 0.75rem; font-size: 0.78rem; color: var(--text-muted); }
+
+.shared { display: flex; flex-wrap: wrap; gap: 0.35rem; list-style: none; margin: 0 0 0.85rem; padding: 0; }
+.chip { font-size: 0.72rem; line-height: 1.4; padding: 0.25rem 0.5rem; border-radius: 999px; background: var(--brand-bg); color: var(--brand); border: 1px solid transparent; }
+
+.metric { margin-top: auto; padding-top: 0.8rem; border-top: 1px solid var(--border); }
+/* No valence colour here, deliberately: green and terracotta mean "against the
+   England average" everywhere else on the site, and colouring a neighbour
+   against this school would read as ranking the neighbours. */
+.value { font-family: var(--font-display); font-size: 1.6rem; font-weight: 700; letter-spacing: -0.6px; margin: 0; color: var(--text-primary); }
+.valueAbsent { font-size: 0.95rem; font-weight: 600; margin: 0; color: var(--text-muted); }
+.metricLabel { margin: 0.25rem 0 0; font-size: 0.75rem; color: var(--text-secondary); }
+.metricRef { margin: 0.1rem 0 0; font-size: 0.75rem; color: var(--text-muted); }
+
+.add { position: relative; z-index: 1; margin-top: 0.85rem; width: 100%; min-height: 44px; font: inherit; font-size: 0.82rem; font-weight: 500; cursor: pointer; border-radius: 8px; border: 1px solid var(--border-strong); background: var(--bg-card); color: var(--brand); }
+.add:hover { border-color: var(--brand); background: var(--brand-bg); }
+.add[aria-pressed="true"] { border-color: var(--brand); background: var(--brand-bg); font-weight: 600; }
+
+.footer { display: flex; flex-wrap: wrap; align-items: center; justify-content: space-between; gap: 0.85rem; margin-top: 1.25rem; padding-top: 1.1rem; border-top: 1px solid var(--border); }
+.footer p { margin: 0; font-size: 0.78rem; color: var(--text-muted); }
+.footer strong { display: block; font-size: 0.88rem; font-weight: 600; color: var(--text-primary); }
+/* Coral is the one decisive action per screen, and in this section this is it. */
+.compare { min-height: 44px; padding: 0 1.25rem; border-radius: 8px; font-size: 0.88rem; font-weight: 600; background: var(--action); color: var(--action-on); border: 1px solid var(--action); text-decoration: none; display: inline-flex; align-items: center; }
+.compare:hover { background: var(--action-strong); border-color: var(--action-strong); }
+.compare[aria-disabled="true"] { opacity: 0.45; pointer-events: none; }
+
+.srOnly { position: absolute; width: 1px; height: 1px; padding: 0; margin: -1px; overflow: hidden; clip: rect(0 0 0 0); white-space: nowrap; border: 0; }

nextjs-app/components/school/NearbySchoolsCarousel.tsx

@@ -0,0 +1,127 @@
+'use client';
+
+/**
+ * The scroller and its arrows.
+ *
+ * `children` are the server-rendered cards and `header` the server-rendered
+ * heading and lede: both stay server components, passed through, so this file
+ * owns a DOM ref and nothing else. That is what keeps all six links in the
+ * initial HTML — a carousel that mounted cards on click would put four of the
+ * six beyond a crawler and beyond a reader with no JavaScript.
+ *
+ * With JavaScript off this degrades to a horizontally scrollable row, which is
+ * still usable by touch and trackpad.
+ */
+
+import { useCallback, useEffect, useRef, useState, type ReactNode } from 'react';
+import styles from './NearbySchools.module.css';
+
+/** Three cards fit the row, so fewer than four has nowhere to scroll to.
+ *  Below 640px the arrows are not rendered at all — see the stylesheet. */
+const VISIBLE = 3;
+
+/**
+ * Why a tolerance rather than `=== 0`.
+ *
+ * The scroller carries 2px of padding so focus rings are not clipped, and
+ * scroll-snap treats that padding as the first card's snap position — a row at
+ * rest reports scrollLeft 2, not 0. Sub-pixel rounding moves it again at other
+ * zoom levels. An exact test leaves the back arrow live on first paint,
+ * pointing nowhere.
+ */
+const EDGE = 8;
+
+export function NearbySchoolsCarousel({
+  count,
+  labelledBy,
+  header,
+  children,
+}: {
+  count: number;
+  labelledBy: string;
+  header: ReactNode;
+  children: ReactNode;
+}) {
+  const scroller = useRef<HTMLUListElement>(null);
+  const [atStart, setAtStart] = useState(true);
+  const [atEnd, setAtEnd] = useState(false);
+  const scrollable = count > VISIBLE;
+
+  const sync = useCallback(() => {
+    const node = scroller.current;
+    if (!node) return;
+    const max = node.scrollWidth - node.clientWidth;
+    setAtStart(node.scrollLeft <= EDGE);
+    setAtEnd(node.scrollLeft >= max - EDGE);
+  }, []);
+  // `atEnd` is not only the forward arrow's disabled state: below 640px, where
+  // no arrow is rendered, it is the only thing driving the scroll-fade.
+
+  // Also on mount: the first measurement can only happen once there is layout.
+  useEffect(sync, [sync]);
+
+  const page = (direction: 1 | -1) => {
+    const node = scroller.current;
+    if (!node) return;
+    // A page is what the reader can see, so the viewport is the step.
+    node.scrollBy({ left: direction * node.clientWidth, behavior: 'smooth' });
+  };
+
+  return (
+    <>
+      <div className={styles.top}>
+        {header}
+        {scrollable && (
+          <div className={styles.arrows}>
+            <button
+              type="button"
+              className={styles.arrow}
+              onClick={() => page(-1)}
+              disabled={atStart}
+              aria-label="Previous schools"
+            >
+              <Chevron direction="prev" />
+            </button>
+            <button
+              type="button"
+              className={styles.arrow}
+              onClick={() => page(1)}
+              disabled={atEnd}
+              aria-label="More schools"
+            >
+              <Chevron direction="next" />
+            </button>
+          </div>
+        )}
+      </div>
+
+      <ul
+        ref={scroller}
+        className={styles.scroller}
+        onScroll={sync}
+        data-at-end={atEnd}
+        {...(scrollable
+          ? { tabIndex: 0, role: 'group', 'aria-labelledby': labelledBy }
+          : {})}
+      >
+        {children}
+      </ul>
+    </>
+  );
+}
+
+function Chevron({ direction }: { direction: 'prev' | 'next' }) {
+  return (
+    <svg
+      viewBox="0 0 24 24"
+      fill="none"
+      stroke="currentColor"
+      strokeWidth="2"
+      strokeLinecap="round"
+      strokeLinejoin="round"
+      aria-hidden="true"
+    >
+      <path d={direction === 'prev' ? 'M15 18l-6-6 6-6' : 'M9 18l6-6-6-6'} />
+    </svg>
+  );
+}

nextjs-app/components/school/NearbySchoolsCompareBar.tsx

@@ -0,0 +1,59 @@
+'use client';
+
+/**
+ * The selection count and the one decisive action in the section.
+ *
+ * The CTA is a real link, not a handler: /compare already parses `urns` from
+ * the query string, so the hand-off needs no new compare plumbing. It counts
+ * this school plus whatever the reader ticked, because comparing a shortlist
+ * without the school they are looking at is not what they asked for.
+ */
+
+import Link from 'next/link';
+import { useComparisonContext } from '@/context/ComparisonContext';
+import type { NearbySchool } from '@/lib/types';
+import styles from './NearbySchools.module.css';
+
+export function NearbySchoolsCompareBar({
+  thisUrn,
+  candidates,
+}: {
+  thisUrn: number;
+  candidates: NearbySchool[];
+}) {
+  const { selectedSchools } = useComparisonContext();
+
+  // Only the schools this section offers, in the order the cards show them —
+  // the basket may hold schools picked up elsewhere on the site, and this bar
+  // speaks for this section.
+  const offered = candidates
+    .map((c) => c.urn)
+    .filter((urn) => selectedSchools.some((s) => s.urn === urn));
+
+  const count = offered.length;
+  const href = `/compare?urns=${[thisUrn, ...offered].join(',')}`;
+
+  return (
+    <div className={styles.footer}>
+      <p aria-live="polite">
+        <strong>
+          {count
+            ? `${count} school${count === 1 ? '' : 's'} selected`
+            : 'Compare side by side'}
+        </strong>
+        {count
+          ? 'This school is included automatically.'
+          : 'Add a school to compare it with this one.'}
+      </p>
+      {count ? (
+        <Link className={styles.compare} href={href}>
+          Compare {count + 1} schools →
+        </Link>
+      ) : (
+        <span className={styles.compare} aria-disabled="true">
+          Compare →
+        </span>
+      )}
+    </div>
+  );
+}

nextjs-app/components/school/NearbySchoolsSection.tsx

@@ -0,0 +1,152 @@
+/**
+ * NearbySchoolsSection — the nearest eligible schools, closest first. Server
+ * component; only the carousel, the compare bar and the add-to-compare button
+ * are client-side.
+ *
+ * "Other schools nearby", not "similar" ones: the order is distance and only
+ * distance. The hard filters upstream still guarantee the set is comparable —
+ * same phase, same selectivity, mainstream never beside special — but nothing
+ * here ranks by how alike two schools are, so the heading does not say it does.
+ *
+ * The chips report what a school shares, and may be absent entirely. That is
+ * information for the reader to weigh, not a verdict this section has already
+ * reached on their behalf.
+ *
+ * There is deliberately no "how these are chosen" panel: the method is already
+ * visible in the lede, the chips and the distances. The single caption line is
+ * not a method note — it is the one thing a card cannot self-correct.
+ */
+
+import Link from 'next/link';
+import type { NearbySchool } from '@/lib/types';
+import { schoolUrl } from '@/lib/utils';
+import { AddToCompareButton } from './AddToCompareButton';
+import { NearbySchoolsCarousel } from './NearbySchoolsCarousel';
+import { NearbySchoolsCompareBar } from './NearbySchoolsCompareBar';
+import { Section } from './sectionShared';
+import styles from './NearbySchools.module.css';
+
+const MINIMUM = 2;
+
+export function shouldRenderNearby(nearby?: NearbySchool[] | null): boolean {
+  return (nearby?.length ?? 0) >= MINIMUM;
+}
+
+/**
+ * What the lede calls the set of schools it is showing.
+ *
+ * Derived from the school's own GIAS phase rather than the template it renders
+ * with, because those disagree for "16 plus" (GIAS phase 6): a sixth-form
+ * college renders the primary template — computeSchoolFlags tests for the
+ * substring "secondary" — while the backend correctly matches it against the
+ * secondary group. Taking the noun from the template would print "Other primary
+ * schools near <sixth form college>" above a row of secondaries.
+ *
+ * A 16-plus school's candidates span the whole secondary group, so no single
+ * noun fits and it gets the honest general one.
+ */
+export function nearbyNoun(phase: string | null | undefined): string {
+  const text = (phase ?? '').trim().toLowerCase();
+  if (text === 'all-through') return 'all-through schools';
+  if (text === '16 plus') return 'schools and colleges';
+  if (text.includes('secondary')) return 'secondary schools';
+  if (text.includes('primary')) return 'primary schools';
+  return 'schools';
+}
+
+function metricLabel(key: string): string {
+  return key === 'attainment_8_score' ? 'Attainment 8' : 'Reading, writing & maths';
+}
+
+function formatMetric(value: number | null, key: string): string {
+  if (value == null) return 'Not published';
+  return key === 'attainment_8_score' ? value.toFixed(1) : `${Math.round(value)}%`;
+}
+
+export function NearbySchoolsSection({
+  urn,
+  schoolName,
+  phase,
+  thisMetricValue,
+  nearby,
+}: {
+  urn: number;
+  schoolName: string;
+  /** The school's own GIAS phase, not the template it renders with. */
+  phase: string | null | undefined;
+  thisMetricValue: number | null;
+  nearby?: NearbySchool[] | null;
+}) {
+  if (!shouldRenderNearby(nearby)) return null;
+  const schools = nearby as NearbySchool[];
+
+  // One card matched on phase alone, so the section may not claim the set
+  // shares an intake with this school.
+  const metricKey = schools[0].metric_key;
+  const noun = nearbyNoun(phase);
+
+  return (
+    <Section id="nearby">
+      <NearbySchoolsCarousel
+        count={schools.length}
+        labelledBy="nearby-schools-heading"
+        header={
+          <div>
+            <h2 id="nearby-schools-heading" className={styles.heading}>
+              Other schools nearby
+            </h2>
+            <p className={styles.lede}>{`Other ${noun} near ${schoolName}.`}</p>
+          </div>
+        }
+      >
+        {schools.map((school) => (
+          <li key={school.urn} className={styles.school}>
+            <p className={styles.distance}>{school.distance_miles} miles away</p>
+            <h3 className={styles.name}>
+              <Link href={schoolUrl(school.urn, school.school_name)}>
+                {school.school_name}
+              </Link>
+            </h3>
+            <p className={styles.meta}>
+              {[school.school_type, school.age_range && `Ages ${school.age_range}`]
+                .filter(Boolean)
+                .join(' · ')}
+            </p>
+            {school.shared.length > 0 && (
+              <ul className={styles.shared}>
+                {school.shared.map((label) => (
+                  <li key={label} className={styles.chip}>{label}</li>
+                ))}
+              </ul>
+            )}
+            <div className={styles.metric}>
+              <p
+                className={
+                  school.metric_value == null ? styles.valueAbsent : styles.value
+                }
+              >
+                {formatMetric(school.metric_value, school.metric_key)}
+              </p>
+              <p className={styles.metricLabel}>{metricLabel(school.metric_key)}</p>
+              {thisMetricValue != null && (
+                <p className={styles.metricRef}>
+                  {formatMetric(thisMetricValue, metricKey)} at this school
+                </p>
+              )}
+            </div>
+            <AddToCompareButton school={school} />
+          </li>
+        ))}
+      </NearbySchoolsCarousel>
+
+      <NearbySchoolsCompareBar thisUrn={urn} candidates={schools} />
+
+      {/* The one caveat the cards cannot make on their own: a reader who takes
+          "0.6 miles away" for the walk has been misled, and nothing else here
+          corrects that. */}
+      <p className={styles.caption}>
+        Distances are straight-line from this school, not road distance.
+      </p>
+    </Section>
+  );
+}

nextjs-app/components/school/PrimarySchoolSections.tsx

@@ -14,6 +14,7 @@
 import type {
   School, SchoolResult, AbsenceData, OfstedInspection, SchoolCensus,
   SchoolAdmissions, SchoolAdmissionDistance, SchoolDeprivation, SchoolFinance, NationalAverages,
+  NearbySchool,
 } from '@/lib/types';
 import { ofstedLegacyAreas } from '@/lib/utils';
 import type { SchoolFlags } from '@/lib/schoolSections';
@@ -26,6 +27,7 @@ import { HistorySection } from './HistorySection';
 import { SchoolLifeSection } from './SchoolLifeSection';
 import { LocalAreaSection } from './LocalAreaSection';
 import { FinancesSection } from './FinancesSection';
+import { NearbySchoolsSection } from './NearbySchoolsSection';
 
 export interface PrimarySchoolSectionsProps {
   schoolInfo: School;
@@ -39,13 +41,15 @@ export interface PrimarySchoolSectionsProps {
   deprivation: SchoolDeprivation | null;
   finance: SchoolFinance | null;
   nationalAvg: NationalAverages | null;
+  /** Nearby schools of a comparable intake. Absent on an older API build. */
+  nearbySchools?: NearbySchool[];
   flags: SchoolFlags;
 }
 
 export function PrimarySchoolSections({
   schoolInfo, yearlyData, absenceData, ofsted, census,
   admissions, admissionsHistory, admissionDistance,
-  deprivation, finance, nationalAvg, flags,
+  deprivation, finance, nationalAvg, nearbySchools, flags,
 }: PrimarySchoolSectionsProps) {
   const primaryAvg = nationalAvg?.primary ?? {};
   const secondaryAvg = nationalAvg?.secondary ?? {};
@@ -146,6 +150,15 @@ export function PrimarySchoolSections({
       )}
 
       {flags.hasFinance && finance && <FinancesSection finance={finance} />}
+
+      {/* Last: it is where the reader goes next, not part of this school. */}
+      <NearbySchoolsSection
+        urn={schoolInfo.urn}
+        schoolName={schoolInfo.school_name}
+        phase={schoolInfo.phase}
+        thisMetricValue={flags.latestResults?.rwm_expected_pct ?? null}
+        nearby={nearbySchools}
+      />
     </>
   );
 }

nextjs-app/components/school/SchoolDetailShell.module.css

@@ -283,6 +283,21 @@
 }
 
 
+/* `position: sticky` with a z-index makes .sectionNav a stacking context, so
+   the jump sheet's own z-index only orders it INSIDE that context. Against the
+   fixed bottom tab bar (Navigation.module.css, z-index 1000) what counts is
+   .sectionNav's 10 — which is why the sheet's last item was painted over, and
+   untappable, once the list grew long enough to reach the bar.
+
+   Lifted only while the sheet is open, and only to 1100: above the bar, below
+   the comparison toast (2000), the fullscreen map (5000) and the info popover
+   (9999). This rule carries the z-index and nothing else; every other
+   declaration belongs to .sectionNav in both states. */
+.sectionNavSheetOpen {
+    z-index: 1100;
+}
+
+
 .sectionNavBack {
     flex: none;
     display: inline-flex;

nextjs-app/components/school/SchoolDetailShell.tsx

@@ -344,7 +344,10 @@ export function SchoolDetailShell({
       </header>
 
       {/* Sticky Section Navigation — docks under the global header */}
-      <nav className={styles.sectionNav} aria-label="Page sections">
+      <nav
+        className={`${styles.sectionNav}${sectionsOpen ? ` ${styles.sectionNavSheetOpen}` : ''}`}
+        aria-label="Page sections"
+      >
         <button onClick={scrollToTop} className={styles.sectionNavBack} aria-label="Back to top">
           <span aria-hidden="true">↑</span>
           <span className={styles.sectionNavBackLabel}>Top</span>

nextjs-app/components/school/SecondarySchoolSections.tsx

@@ -14,7 +14,7 @@
 import type {
   School, SchoolResult, AbsenceData, OfstedInspection, SchoolCensus,
   SchoolAdmissions, SchoolAdmissionDistance, SchoolDeprivation, SchoolFinance, NationalAverages,
-  SchoolDestinations,
+  SchoolDestinations, NearbySchool,
 } from '@/lib/types';
 import { ofstedLegacyAreas } from '@/lib/utils';
 import type { SecondaryFlags } from '@/lib/schoolSections';
@@ -27,6 +27,7 @@ import { DistanceSection } from './DistanceSection';
 import { SecondaryHistorySection } from './SecondaryHistorySection';
 import { WellbeingSection } from './WellbeingSection';
 import { FinancesSection } from './FinancesSection';
+import { NearbySchoolsSection } from './NearbySchoolsSection';
 import styles from './schoolSections.module.css';
 
 export interface SecondarySchoolSectionsProps {
@@ -47,13 +48,15 @@ export interface SecondarySchoolSectionsProps {
   finance: SchoolFinance | null;
   nationalAvg: NationalAverages | null;
   destinations: SchoolDestinations | null;
+  /** Nearby schools of a comparable intake. Absent on an older API build. */
+  nearbySchools?: NearbySchool[];
   flags: SecondaryFlags;
 }
 
 export function SecondarySchoolSections({
   schoolInfo, yearlyData, ofsted, census,
   admissions, admissionsHistory, admissionDistance,
-  deprivation, finance, nationalAvg, destinations, flags,
+  deprivation, finance, nationalAvg, destinations, nearbySchools, flags,
 }: SecondarySchoolSectionsProps) {
   const secondaryAvg = nationalAvg?.secondary ?? {};
 
@@ -141,6 +144,15 @@ export function SecondarySchoolSections({
       {flags.hasFinance && finance && (
         <FinancesSection finance={finance} showPremises />
       )}
+
+      {/* Last: it is where the reader goes next, not part of this school. */}
+      <NearbySchoolsSection
+        urn={schoolInfo.urn}
+        schoolName={schoolInfo.school_name}
+        phase={schoolInfo.phase}
+        thisMetricValue={flags.latestResults?.attainment_8_score ?? null}
+        nearby={nearbySchools}
+      />
     </div>
   );
 }

nextjs-app/jest.config.cjs

@@ -9,6 +9,7 @@ const createJestConfig = nextJest({
 const customJestConfig = {
   setupFilesAfterEnv: ['<rootDir>/jest.setup.js'],
   testEnvironment: 'jest-environment-jsdom',
+  modulePathIgnorePatterns: ['<rootDir>/.next/'],
   moduleNameMapper: {
     '^@/(.*)$': '<rootDir>/$1',
   },

nextjs-app/lib/api.ts

@@ -311,26 +311,6 @@ export async function fetchDataInfo(
   return handleResponse<DataInfoResponse>(response);
 }
 
-// ============================================================================
-// Client-Side Fetcher (for SWR)
-// ============================================================================
-
-/**
- * Generic fetcher function for use with SWR
- * @example
- * ```tsx
- * const { data, error } = useSWR('/api/schools', fetcher);
- * ```
- */
-export async function fetcher<T>(url: string): Promise<T> {
-  // If it's already a full URL, use it directly
-  // Otherwise, prepend the API_BASE_URL
-  const fullUrl = url.startsWith('http') ? url : `${API_BASE_URL}${url.startsWith('/') ? url : `/${url}`}`;
-
-  const response = await fetch(fullUrl);
-  return handleResponse<T>(response);
-}
-
 // ============================================================================
 // Geocoding API
 // ============================================================================
@@ -396,10 +376,3 @@ export function calculateDistance(
   const c = 2 * Math.atan2(Math.sqrt(a), Math.sqrt(1 - a));
   return R * c;
 }
-
-/**
- * Convert kilometers to miles
- */
-export function kmToMiles(km: number): number {
-  return km * 0.621371;
-}

nextjs-app/lib/places.ts

@@ -1,3 +1,5 @@
+import { APIFetchError } from './api';
+
 /**
  * Client for the places API.
  *
@@ -69,7 +71,7 @@ const API = process.env.FASTAPI_URL || process.env.NEXT_PUBLIC_API_URL
 
 export async function fetchPlaces(): Promise<PlaceSummary[]> {
   const res = await fetch(`${API}/places`, { next: { revalidate: 604800 } });
-  if (!res.ok) return [];
+  if (!res.ok) throw new APIFetchError("Unable to load places", res.status);
   return (await res.json()).places ?? [];
 }
 
@@ -79,6 +81,7 @@ export async function fetchPlace(
   const q = phase ? `?phase=${encodeURIComponent(phase)}` : '';
   const res = await fetch(`${API}/places/${kind}/${slug}${q}`,
     { next: { revalidate: 604800 } });
-  if (!res.ok) return null;
+  if (res.status === 404) return null;
+  if (!res.ok) throw new APIFetchError("Unable to load place", res.status);
   return res.json();
 }

nextjs-app/lib/schoolSections.ts

@@ -128,6 +128,10 @@ export interface NavItemsInput {
    *  measure a postcode, so the nav must gate on them too or it will link to an
    *  anchor that was never rendered. */
   hasLocation?: boolean;
+  /** Whether the nearby-schools section will render. Optional for the same
+   *  reason hasLocation is: the nav must never link to an anchor that was not
+   *  rendered, and absent has to mean "no section". */
+  hasNearbySchools?: boolean;
   yearlyDataLength: number;
 }
 
@@ -142,7 +146,10 @@ export interface NavItemsInput {
  */
 export function buildNavItems(
   flags: SchoolFlags,
-  { ofsted, admissions, admissionDistance, hasLocation, yearlyDataLength }: NavItemsInput,
+  {
+    ofsted, admissions, admissionDistance, hasLocation,
+    hasNearbySchools, yearlyDataLength,
+  }: NavItemsInput,
 ): NavItem[] {
   const navItems: NavItem[] = [];
   if (ofsted) navItems.push({ id: 'ofsted', label: 'Ofsted' });
@@ -161,6 +168,8 @@ export function buildNavItems(
   if (flags.hasSchoolLife) navItems.push({ id: 'school-life', label: 'School Life' });
   if (flags.hasDeprivation) navItems.push({ id: 'local-area', label: 'Local Area' });
   if (flags.hasFinance) navItems.push({ id: 'finances', label: 'Finances' });
+  // Last, because the section renders last — the scroll-spy reads this order.
+  if (hasNearbySchools) navItems.push({ id: 'nearby', label: 'Nearby schools' });
   return navItems;
 }
 
@@ -239,7 +248,10 @@ export function computeSecondaryFlags({
  */
 export function buildSecondaryNavItems(
   flags: SecondaryFlags,
-  { ofsted, admissions, admissionDistance, hasLocation, yearlyDataLength }: NavItemsInput,
+  {
+    ofsted, admissions, admissionDistance, hasLocation,
+    hasNearbySchools, yearlyDataLength,
+  }: NavItemsInput,
 ): NavItem[] {
   const navItems: NavItem[] = [];
   if (ofsted) navItems.push({ id: 'ofsted', label: 'Ofsted' });
@@ -257,5 +269,7 @@ export function buildSecondaryNavItems(
   if (yearlyDataLength > 1) navItems.push({ id: 'history', label: 'History' });
   if (flags.hasWellbeing) navItems.push({ id: 'wellbeing', label: 'Wellbeing' });
   if (flags.hasFinance) navItems.push({ id: 'finances', label: 'Finances' });
+  // Last, because the section renders last — the scroll-spy reads this order.
+  if (hasNearbySchools) navItems.push({ id: 'nearby', label: 'Nearby schools' });
   return navItems;
 }

nextjs-app/lib/types.ts

@@ -346,6 +346,26 @@ export interface SchoolsResponse {
   };
 }
 
+/**
+ * A nearby school, from the nearest-first set the detail page shows.
+ *
+ * `shared` is what this school genuinely has in common with the one being
+ * viewed, and may be empty. It is reported, never ranked on: an earlier
+ * version ordered by it and buried the school down the road under faith
+ * matches three times further away.
+ */
+export interface NearbySchool {
+  urn: number;
+  school_name: string;
+  distance_miles: number;
+  school_type: string | null;
+  age_range: string | null;
+  shared: string[];
+  metric_value: number | null;
+  metric_key: string;
+  metric_year: number | null;
+}
+
 export interface SchoolDetailsResponse {
   school_info: School;
   /**
@@ -357,6 +377,14 @@ export interface SchoolDetailsResponse {
    * authority both fall below the publish threshold has nowhere to link.
    */
   places?: SchoolPlace[];
+  /**
+   * Up to six nearby eligible schools, nearest first.
+   *
+   * Optional for the same reason as `places`: a frontend deployed ahead of the
+   * API that serves this must render without it. Absent and empty mean the
+   * same thing here — no section.
+   */
+  nearby_schools?: NearbySchool[];
   yearly_data: SchoolResult[];
   absence_data: AbsenceData | null;
   // Supplementary data (null until Kestra populates)

nextjs-app/payload.config.ts

@@ -24,9 +24,8 @@ export default buildConfig({
   typescript: { outputFile: path.resolve(dirname, 'payload-types.ts') },
   db: postgresAdapter({
     pool: { connectionString: process.env.DATABASE_URL },
-    // Its own schema, so no pipeline operation on `public` can reach blog
-    // content. scripts/migrate_csv_to_db.py --drop lives in that blast radius,
-    // as does Airflow's metadata. The schema itself is created by the initial
+    // Its own schema separates blog content from pipeline-managed school
+    // tables and Airflow metadata. The schema is created by the initial
     // migration: schemaName says where tables go, it does not create anything.
     //
     // prodMigrations runs pending migrations during server init. Without it a

pipeline/Dockerfile

@@ -40,4 +40,9 @@ ENV AIRFLOW_HOME=/opt/airflow
 ENV AIRFLOW__CORE__DAGS_FOLDER=/opt/pipeline/dags
 ENV PYTHONPATH=/opt/pipeline
 
+ARG BUILD_SHA=development
+ARG BUILD_ID=development
+LABEL io.schoolcompare.build-id=$BUILD_ID
+LABEL io.schoolcompare.commit=$BUILD_SHA
+
 CMD ["airflow", "api-server"]

pipeline/scripts/sync_typesense.py

@@ -11,6 +11,10 @@ Usage:
 from __future__ import annotations
 
 import argparse
+import json
+import logging
+import re
+import uuid
 import os
 import sys
 import time
@@ -112,63 +116,78 @@ def build_document(row: dict) -> dict:
     return doc
 
 
+def publish_collection(client, rows: list[dict]) -> str:
+    """Validate a new collection before moving the alias; keep rollback data.
+
+    Caller holds the database advisory lock across reading and publication so
+    overlapping school-data DAGs cannot publish or prune each other's work.
+    Failed drafts are left for the next successful publication to prune: a lost
+    alias-update response must never cause deletion of a potentially live index.
+    """
+    if not rows or len({r["urn"] for r in rows}) != len(rows):
+        raise ValueError("Search source must contain nonempty, unique school URNs")
+    name = f"schools_{int(time.time())}_{uuid.uuid4().hex[:12]}"
+    try:
+        previous = client.aliases["schools"].retrieve()["collection_name"]
+    except typesense.exceptions.ObjectNotFound:
+        previous = None
+    client.collections.create({**COLLECTION_SCHEMA, "name": name})
+    for i in range(0, len(rows), 500):
+        batch = [build_document(r) for r in rows[i:i + 500]]
+        results = client.collections[name].documents.import_(batch, {"action": "upsert"})
+        if isinstance(results, str):
+            results = [json.loads(line) for line in results.splitlines() if line.strip()]
+        if len(results) != len(batch) or any(r.get("success") is not True for r in results):
+            raise ValueError("Search import failed; live alias unchanged")
+    if client.collections[name].retrieve()["num_documents"] != len(rows):
+        raise ValueError("Search document count mismatch; live alias unchanged")
+    client.aliases.upsert("schools", {"collection_name": name})
+
+    # Retention is best-effort and must not make successful publication fail.
+    try:
+        keep = {name, previous}
+        keep.update(a["collection_name"] for a in client.aliases.retrieve()["aliases"])
+        for collection in client.collections.retrieve():
+            old = collection["name"]
+            if old not in keep and re.fullmatch(r"schools_\d+(?:_[0-9a-f]+)?", old):
+                client.collections[old].delete()
+    except Exception:
+        logging.getLogger(__name__).exception("Search published, but old collection cleanup failed")
+    return name
+
+
 def sync(typesense_url: str, api_key: str):
+    from urllib.parse import urlparse
+    url = urlparse(typesense_url)
     client = typesense.Client({
-        "nodes": [{"host": typesense_url.split("//")[-1].split(":")[0],
-                    "port": typesense_url.split(":")[-1],
-                    "protocol": "http"}],
+        "nodes": [{"host": url.hostname, "port": str(url.port or 8108),
+                   "protocol": url.scheme}],
         "api_key": api_key,
         "connection_timeout_seconds": 10,
     })
-
-    # Create timestamped collection for zero-downtime swap
-    ts = int(time.time())
-    collection_name = f"schools_{ts}"
-
-    print(f"Creating collection: {collection_name}")
-    schema = {**COLLECTION_SCHEMA, "name": collection_name}
-    client.collections.create(schema)
-
-    # Fetch data from marts — join fact_performance if it exists
     conn = get_db_connection()
-    with conn.cursor(cursor_factory=psycopg2.extras.RealDictCursor) as cur:
-        # Check whether the merged fact table exists
-        cur.execute("""
-            SELECT table_name FROM information_schema.tables
-            WHERE table_schema = 'marts' AND table_name = 'fact_performance'
-        """)
-        has_fact_performance = cur.fetchone() is not None
-
-        query = QUERY_BASE
-        if has_fact_performance:
-            query = query.replace(
-                "l.longitude as lng",
-                "l.longitude as lng,\n        p.rwm_expected_pct,\n        p.progress_8_score",
-            )
-            query += QUERY_PERFORMANCE_JOIN
-
-        cur.execute(query)
-        rows = cur.fetchall()
-    conn.close()
-
-    print(f"Indexing {len(rows)} schools...")
-
-    # Batch import
-    batch_size = 500
-    for i in range(0, len(rows), batch_size):
-        batch = [build_document(r) for r in rows[i : i + batch_size]]
-        client.collections[collection_name].documents.import_(batch, {"action": "upsert"})
-        print(f"  Indexed {min(i + batch_size, len(rows))}/{len(rows)}")
-
-    # Swap alias
-    print("Swapping alias 'schools' → new collection")
     try:
-        client.aliases.upsert("schools", {"collection_name": collection_name})
-    except Exception:
-        # If alias doesn't exist yet, create it
-        client.aliases.upsert("schools", {"collection_name": collection_name})
-
-    print("Done.")
+        with conn.cursor(cursor_factory=psycopg2.extras.RealDictCursor) as cur:
+            # Session-scoped lock is released even on errors when conn closes.
+            cur.execute("SELECT pg_advisory_lock(731042019)")
+            cur.execute("""
+                SELECT table_name FROM information_schema.tables
+                WHERE table_schema = 'marts' AND table_name = 'fact_performance'
+            """)
+            has_fact_performance = cur.fetchone() is not None
+            query = QUERY_BASE
+            if has_fact_performance:
+                query = query.replace(
+                    "l.longitude as lng",
+                    "l.longitude as lng, p.rwm_expected_pct, p.progress_8_score",
+                )
+                query += QUERY_PERFORMANCE_JOIN
+            cur.execute(query)
+            rows = cur.fetchall()
+        name = publish_collection(client, rows)
+        print(f"Published {len(rows)} schools in {name}")
+    finally:
+        conn.close()
 
 
 def main():

pipeline/tests/test_sync_typesense.py

@@ -0,0 +1,92 @@
+import importlib.util
+from pathlib import Path
+from unittest.mock import MagicMock
+import pytest
+
+
+@pytest.fixture
+def sync_module(monkeypatch):
+    folder = Path(__file__).resolve().parents[1] / 'scripts'
+    monkeypatch.syspath_prepend(str(folder))
+    spec = importlib.util.spec_from_file_location('sync_typesense', folder / 'sync_typesense.py')
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+@pytest.fixture
+def client():
+    c = MagicMock()
+    c.aliases.__getitem__.return_value.retrieve.return_value = {'collection_name': 'schools_2'}
+    c.aliases.retrieve.return_value = {'aliases': [{'collection_name': 'schools_99'}]}
+    c.collections.__getitem__.return_value.documents.import_.return_value = [{'success': True}]
+    c.collections.__getitem__.return_value.retrieve.return_value = {'num_documents': 1}
+    c.collections.retrieve.return_value = [{'name': n} for n in ['schools_1', 'schools_2', 'schools_99', 'unrelated']]
+    return c
+
+
+@pytest.fixture
+def rows():
+    return [{'urn': 100001, 'school_name': 'Example', 'phase_code': 2,
+             'school_type_code': 1, 'local_authority': 'Testshire', 'postcode': 'TS1 1AA',
+             'total_pupils': 250}]
+
+
+@pytest.mark.parametrize('results', [[{'success': False}], [], [{'success': True}, {'success': True}]])
+def test_partial_import_never_moves_alias(sync_module, client, rows, results):
+    client.collections.__getitem__.return_value.documents.import_.return_value = results
+    with pytest.raises(ValueError):
+        sync_module.publish_collection(client, rows)
+    client.aliases.upsert.assert_not_called()
+    client.collections.__getitem__.return_value.delete.assert_not_called()
+
+
+def test_count_mismatch_does_not_publish(sync_module, client, rows):
+    client.collections.__getitem__.return_value.retrieve.return_value = {'num_documents': 0}
+    with pytest.raises(ValueError):
+        sync_module.publish_collection(client, rows)
+    client.aliases.upsert.assert_not_called()
+
+
+@pytest.mark.parametrize('empty', [True, False])
+def test_invalid_source_never_creates_collection(sync_module, client, rows, empty):
+    with pytest.raises(ValueError):
+        sync_module.publish_collection(client, [] if empty else rows + rows)
+    client.collections.create.assert_not_called()
+
+
+def test_success_retains_previous_and_other_live_aliases(sync_module, client, rows):
+    # Distinct mock per collection allows checking exactly which one was deleted.
+    collections = {}
+    def get(name):
+        if name not in collections:
+            c = MagicMock()
+            c.documents.import_.return_value = '{"success":true}\n'
+            c.retrieve.return_value = {'num_documents': 1}
+            collections[name] = c
+        return collections[name]
+    client.collections.__getitem__.side_effect = get
+    name = sync_module.publish_collection(client, rows)
+    client.aliases.upsert.assert_called_once_with('schools', {'collection_name': name})
+    assert set(collections) == {name, 'schools_1'}
+    collections['schools_1'].delete.assert_called_once()
+
+
+def test_uncertain_alias_update_does_not_delete_candidate(sync_module, client, rows):
+    client.aliases.upsert.side_effect = RuntimeError('response lost')
+    with pytest.raises(RuntimeError):
+        sync_module.publish_collection(client, rows)
+    client.collections.__getitem__.return_value.delete.assert_not_called()
+
+
+def test_sync_closes_database_when_publication_fails(sync_module, monkeypatch):
+    conn = MagicMock()
+    monkeypatch.setattr(sync_module, 'get_db_connection', lambda: conn)
+    monkeypatch.setattr(sync_module.typesense, 'Client', lambda _: MagicMock())
+    def fail(*args): raise ValueError('import rejected')
+    monkeypatch.setattr(sync_module, 'publish_collection', fail)
+    with pytest.raises(ValueError):
+        sync_module.sync('http://localhost:8108', 'dummy')
+    conn.close.assert_called_once()
+    statements = [call.args[0] for call in conn.cursor.return_value.__enter__.return_value.execute.call_args_list]
+    assert statements[0] == 'SELECT pg_advisory_lock(731042019)'

scripts/ci/release.py

@@ -0,0 +1,174 @@
+"""Release identity checks and promotion of the exact digests that passed E2E.
+
+Uses only the standard library and Docker Buildx. No registry mutation happens
+until every image in the set has been resolved and its build labels validated.
+"""
+import argparse
+import json
+import os
+from pathlib import Path
+import re
+import subprocess
+import time
+from urllib.error import HTTPError, URLError
+from urllib.request import Request, urlopen
+
+COMPONENTS = ('BACKEND', 'FRONTEND', 'PIPELINE')
+
+
+def docker(*args):
+    return subprocess.check_output(['docker', 'buildx', 'imagetools', *args], text=True).strip()
+
+
+def check_sha(sha):
+    if not re.fullmatch(r'[0-9a-f]{40}', sha):
+        raise ValueError('Expected a full commit SHA')
+    return sha
+
+
+def check_digest(digest):
+    if not re.fullmatch(r'sha256:[0-9a-f]{64}', digest):
+        raise ValueError('Expected an immutable image digest')
+    return digest
+
+
+def image_identity(ref):
+    image = json.loads(docker('inspect', ref, '--format', '{{json .Image}}'))
+    configs = [image] if 'config' in image else list(image.values())
+    identities = set()
+    for config in configs:
+        labels = config['config']['Labels']
+        identities.add((labels['io.schoolcompare.commit'], labels['io.schoolcompare.build-id']))
+    if len(identities) != 1:
+        raise ValueError('Image platforms disagree about their release identity')
+    return next(iter(identities))
+
+
+def resolve_images(sha, verified=False, expected_build_id=None):
+    check_sha(sha)
+    refs = []
+    build_ids = set()
+    for component in COMPONENTS:
+        image = f"{os.environ['REGISTRY']}/{os.environ[component + '_IMAGE_NAME']}"
+        if verified:
+            manifest = json.loads(docker('inspect', f'{image}:verified-{sha}', '--format', '{{json .Manifest}}'))
+            digest = check_digest(manifest['digest'])
+        else:
+            digest = check_digest(os.environ[component + '_DIGEST'])
+        ref = f'{image}@{digest}'
+        actual_sha, build_id = image_identity(ref)
+        if actual_sha != sha or not re.fullmatch(r'[0-9a-f]{32}', build_id):
+            raise ValueError(f'Unrecognised release identity for {component}')
+        if expected_build_id is not None and build_id != expected_build_id:
+            raise ValueError(f'Build identity mismatch for {component}')
+        build_ids.add(build_id)
+        refs.append((image, ref))
+    if len(build_ids) != 1:
+        raise ValueError('Refusing a mixed image set')
+    return {'sha': sha, 'build_id': build_ids.pop(), 'images': refs}
+
+
+def verify(sha, build_id):
+    release = resolve_images(sha, expected_build_id=build_id)
+    for image, ref in release['images']:
+        docker('create', '-t', f'{image}:verified-{sha}', ref)
+    return release
+
+
+def promote(sha):
+    release = resolve_images(sha, verified=True)
+    # Resolve all targets first; never discover a missing candidate halfway through.
+    for image, _ in release['images']:
+        try:
+            docker('create', '-t', f'{image}:prod-previous', f'{image}:prod')
+        except subprocess.CalledProcessError:
+            print(f'No rollback pointer saved for {image}', flush=True)
+    for image, ref in release['images']:
+        docker('create', '-t', f'{image}:prod', ref)
+    return release
+
+
+def matches(payload, sha, build_id):
+    return isinstance(payload, dict) and all(
+        payload.get(component) == {'sha': sha, 'build_id': build_id}
+        for component in ('frontend', 'backend'))
+
+
+def describe_identity(payload):
+    """Log only release fields, never arbitrary response bodies or secret URLs."""
+    if not isinstance(payload, dict):
+        return 'Invalid release response: expected a JSON object'
+    identities = []
+    for component in ('frontend', 'backend'):
+        identity = payload.get(component)
+        if not isinstance(identity, dict):
+            identities.append(f'{component}=missing or invalid')
+            continue
+        values = []
+        for field, length in (('sha', 40), ('build_id', 32)):
+            value = identity.get(field)
+            valid = isinstance(value, str) and (
+                value == 'development' or re.fullmatch(r'[0-9a-f]{' + str(length) + '}', value))
+            values.append(f'{field}={value if valid else "missing or invalid"}')
+        identities.append(f'{component}: {", ".join(values)}')
+    return 'Release mismatch: ' + '; '.join(identities)
+
+
+def wait(base_url, sha, build_id, timeout):
+    check_sha(sha)
+    if not re.fullmatch(r'[0-9a-f]{32}', build_id):
+        raise ValueError('Missing expected build identity')
+    deadline = time.monotonic() + timeout
+    last_observation = 'No response received'
+    print(f'Waiting for deployed release {sha} / {build_id}', flush=True)
+    while time.monotonic() < deadline:
+        try:
+            req = Request(f'{base_url.rstrip("/")}/release.json?check={time.time_ns()}',
+                          headers={'Cache-Control': 'no-cache',
+                                   'User-Agent': 'SchoolCompare-Release-Check/1.0',
+                                   'Accept': 'application/json'})
+            with urlopen(req, timeout=min(10, max(.1, deadline - time.monotonic()))) as response:
+                payload = json.load(response)
+            if matches(payload, sha, build_id):
+                print(f'Verified deployed release {sha} / {build_id}')
+                return
+            observation = describe_identity(payload)
+        except HTTPError as exc:
+            observation = f'Release endpoint returned HTTP {exc.code}'
+            exc.close()
+        except URLError as exc:
+            observation = f'Release endpoint connection failed ({type(exc.reason).__name__})'
+        except OSError as exc:
+            observation = f'Release endpoint request failed ({type(exc).__name__})'
+        except ValueError:
+            observation = 'Release endpoint returned invalid JSON or request configuration'
+        if observation != last_observation:
+            print(observation, flush=True)
+        last_observation = observation
+        time.sleep(min(5, max(0, deadline - time.monotonic())))
+    raise RuntimeError('Deployment did not report the expected frontend/backend release '
+                       f'{sha} / {build_id}. Last observation: {last_observation}')
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('action', choices=['wait', 'verify', 'promote'])
+    parser.add_argument('--timeout', type=float, default=300)
+    parser.add_argument('--release', type=Path)
+    parser.add_argument('--output', type=Path)
+    args = parser.parse_args()
+    identity = json.loads(args.release.read_text()) if args.release else {
+        'sha': os.environ.get('EXPECTED_SHA', ''),
+        'build_id': os.environ.get('EXPECTED_BUILD_ID', ''),
+    }
+    if args.action == 'wait':
+        wait(os.environ['BASE_URL'], identity['sha'], identity['build_id'], args.timeout)
+        return
+    result = (verify(identity['sha'], identity['build_id']) if args.action == 'verify'
+              else promote(identity['sha']))
+    if args.output:
+        args.output.write_text(json.dumps(result))
+
+
+if __name__ == '__main__':
+    main()

scripts/ci/tests/test_release.py

@@ -0,0 +1,130 @@
+import json
+from io import BytesIO
+from urllib.error import HTTPError, URLError
+from unittest.mock import Mock
+import pytest
+from scripts.ci import release
+
+SHA = 'a' * 40
+BUILD = 'b' * 32
+DIGESTS = ['sha256:' + c * 64 for c in '123']
+
+
+@pytest.fixture
+def docker(monkeypatch):
+    monkeypatch.setenv('REGISTRY', 'registry.example')
+    refs = {}
+    for component, digest in zip(release.COMPONENTS, DIGESTS):
+        monkeypatch.setenv(component + '_IMAGE_NAME', component.lower())
+        monkeypatch.setenv(component + '_DIGEST', digest)
+        refs[f'registry.example/{component.lower()}'] = digest
+    def run(*args):
+        if args[0] == 'create': return ''
+        if args[-1] == '{{json .Manifest}}':
+            return json.dumps({'digest': refs[args[1].split(':')[0]]})
+        return json.dumps({'config': {'Labels': {'io.schoolcompare.commit': SHA,
+                                                'io.schoolcompare.build-id': BUILD}}})
+    mock = Mock(side_effect=run)
+    monkeypatch.setattr(release, 'docker', mock)
+    return mock
+
+
+def test_wrong_deployed_build_is_rejected_even_at_same_commit():
+    assert not release.matches({'frontend': {'sha': SHA, 'build_id': BUILD},
+                                'backend': {'sha': SHA, 'build_id': 'c' * 32}}, SHA, BUILD)
+    assert release.matches({c: {'sha': SHA, 'build_id': BUILD} for c in ('frontend', 'backend')}, SHA, BUILD)
+
+
+def test_verification_tags_the_captured_digests(docker):
+    release.verify(SHA, BUILD)
+    creates = [c.args for c in docker.call_args_list if c.args[0] == 'create']
+    assert len(creates) == 3
+    for call, digest in zip(creates, DIGESTS):
+        assert call[-1].endswith('@' + digest)
+        assert call[2].endswith(':verified-' + SHA)
+
+
+def test_promotion_resolves_all_verified_images_before_mutation(docker):
+    result = release.promote(SHA)
+    assert result['build_id'] == BUILD
+    calls = [c.args for c in docker.call_args_list]
+    first_write = next(i for i, c in enumerate(calls) if c[0] == 'create')
+    assert first_write == 6  # each of three candidates needs manifest + config
+    assert all(c[-1].endswith('@' + d) for c, d in zip(calls[-3:], DIGESTS))
+
+
+def test_mixed_builds_fail_before_any_tag_is_changed(docker, monkeypatch):
+    identities = iter([(SHA, BUILD), (SHA, 'c' * 32), (SHA, BUILD)])
+    monkeypatch.setattr(release, 'image_identity', lambda _: next(identities))
+    with pytest.raises(ValueError, match='mixed'):
+        release.promote(SHA)
+    assert not any(c.args[0] == 'create' for c in docker.call_args_list)
+
+
+def test_missing_candidate_fails_before_any_tag_is_changed(docker):
+    docker.side_effect = RuntimeError('missing verified tag')
+    with pytest.raises(RuntimeError): release.promote(SHA)
+    assert not any(c.args[0] == 'create' for c in docker.call_args_list)
+
+
+@pytest.fixture
+def poll(monkeypatch):
+    now = [0.0]
+    monkeypatch.setattr(release.time, 'monotonic', lambda: now[0])
+    monkeypatch.setattr(release.time, 'sleep', lambda seconds: now.__setitem__(0, now[0] + seconds))
+    opener = Mock()
+    monkeypatch.setattr(release, 'urlopen', opener)
+    return opener
+
+
+def response(payload):
+    return BytesIO(json.dumps(payload).encode())
+
+
+def test_wait_identifies_its_client_and_retries_until_both_services_match(poll, capsys):
+    poll.side_effect = [
+        HTTPError('https://secret.example', 503, 'unavailable', {}, None),
+        response({'frontend': {'sha': SHA, 'build_id': BUILD},
+                  'backend': {'sha': SHA, 'build_id': 'c' * 32}}),
+        response({component: {'sha': SHA, 'build_id': BUILD}
+                  for component in ('frontend', 'backend')}),
+    ]
+    release.wait('https://secret.example/', SHA, BUILD, 15)
+    assert poll.call_count == 3
+    request = poll.call_args.args[0]
+    assert request.get_header('User-agent') == 'SchoolCompare-Release-Check/1.0'
+    assert request.get_header('Cache-control') == 'no-cache'
+    assert request.get_header('Accept') == 'application/json'
+    assert '/release.json?check=' in request.full_url
+    output = capsys.readouterr().out
+    assert 'HTTP 503' in output
+    assert 'backend: sha=' + SHA + ', build_id=' + 'c' * 32 in output
+    assert 'Verified deployed release' in output
+    assert 'secret.example' not in output
+
+
+@pytest.mark.parametrize('failure, expected', [
+    (lambda: HTTPError('https://secret.example', 403, 'secret response', {}, None), 'HTTP 403'),
+    (lambda: URLError(OSError('secret address')), 'connection failed (OSError)'),
+    (lambda: TimeoutError('secret address'), 'request failed (TimeoutError)'),
+    (lambda: BytesIO(b'<html>secret response</html>'), 'invalid JSON'),
+    (lambda: response([]), 'expected a JSON object'),
+    (lambda: response({'frontend': {'sha': 'secret response'}}), 'missing or invalid'),
+])
+def test_wait_timeout_reports_last_failure_without_leaking_response_or_url(poll, capsys, failure, expected):
+    poll.side_effect = lambda *args, **kwargs: result_or_raise(failure())
+    with pytest.raises(RuntimeError) as error:
+        release.wait('https://secret.example', SHA, BUILD, 10)
+    assert expected in str(error.value)
+    assert SHA in str(error.value)
+    assert BUILD in str(error.value)
+    output = capsys.readouterr().out
+    assert sum(expected in line for line in output.splitlines()) == 1
+    assert 'secret' not in output + str(error.value)
+    assert poll.call_count == 2
+
+
+def result_or_raise(result):
+    if isinstance(result, Exception):
+        raise result
+    return result

scripts/ci/tests/test_workflows.py

@@ -0,0 +1,34 @@
+"""Check the dependency graph that ties tested digests to deployable images."""
+from pathlib import Path
+import yaml
+
+ROOT = Path(__file__).resolve().parents[3]
+
+
+def test_staging_verifies_identity_before_and_after_journeys():
+    workflow = yaml.safe_load((ROOT / '.gitea/workflows/deploy.yml').read_text())
+    assert workflow['concurrency'] == {'group': 'staging-release', 'cancel-in-progress': False}
+    jobs = workflow['jobs']
+    for component in ('backend', 'frontend', 'pipeline'):
+        job = jobs['build-' + component]
+        assert 'prepare' in job['needs']
+        assert job['outputs']['digest'] == '${{ steps.build.outputs.digest }}'
+        build = next(step for step in job['steps'] if step.get('id') == 'build')
+        assert 'BUILD_ID=${{ needs.prepare.outputs.build_id }}' in build['with']['build-args']
+    steps = jobs['e2e-staging']['steps']
+    runs = [step.get('run', '') for step in steps]
+    test = runs.index('npx playwright test')
+    assert 'release.py wait' in runs[test - 1]
+    assert 'release.py wait' in runs[test + 1]
+    assert 'release.py verify' in runs[-1]
+    for component in ('backend', 'frontend', 'pipeline'):
+        assert 'build-' + component in jobs['e2e-staging']['needs']
+        assert component.upper() + '_DIGEST' in steps[-1]['env']
+
+
+def test_promotion_uses_verified_digest_resolver_and_build_identity_poll():
+    workflow = yaml.safe_load((ROOT / '.gitea/workflows/promote.yml').read_text())
+    steps = workflow['jobs']['promote-prod']['steps']
+    runs = [step.get('run', '') for step in steps]
+    assert 'python3 scripts/ci/release.py promote --output release.json' in runs
+    assert runs[-1] == 'python3 scripts/ci/release.py wait --release release.json'

scripts/download_data.py

@@ -1,5 +1,8 @@
 #!/usr/bin/env python3
 """
+Historical standalone CSV utility; not part of the managed Meltano/dbt pipeline.
+See docs/LEGACY_CODE.md before using it for current school data.
+
 Data Download Helper Script
 
 This script provides instructions and utilities for downloading

scripts/fetch_real_data.py

@@ -1,5 +1,8 @@
 #!/usr/bin/env python3
 """
+Historical standalone CSV utility; not part of the managed Meltano/dbt pipeline.
+See docs/LEGACY_CODE.md before using it for current school data.
+
 Fetch real school performance data from UK Government sources.
 
 This script downloads KS2 (Key Stage 2) primary school data from:

scripts/geocode_schools.py

@@ -1,184 +0,0 @@
-#!/usr/bin/env python3
-"""
-Geocode all school postcodes and update the database.
-
-This script should be run as a weekly cron job to ensure all schools
-have up-to-date latitude/longitude coordinates.
-
-Usage:
-    python scripts/geocode_schools.py [--force]
-
-Options:
-    --force     Re-geocode all postcodes, even if already geocoded
-
-Crontab example (run every Sunday at 2am):
-    0 2 * * 0 cd /path/to/school_compare && /path/to/venv/bin/python scripts/geocode_schools.py >> /var/log/geocode_schools.log 2>&1
-"""
-
-import argparse
-import sys
-from datetime import datetime
-from pathlib import Path
-from typing import Dict, Tuple
-
-import requests
-
-# Add parent directory to path for imports
-sys.path.insert(0, str(Path(__file__).parent.parent))
-
-from backend.database import SessionLocal
-from backend.models import School
-
-
-def geocode_postcodes_bulk(postcodes: list) -> Dict[str, Tuple[float, float]]:
-    """
-    Geocode postcodes in bulk using postcodes.io API.
-    Returns dict of postcode -> (latitude, longitude).
-    """
-    results = {}
-    valid_postcodes = [
-        p.strip().upper()
-        for p in postcodes
-        if p and isinstance(p, str) and len(p.strip()) >= 5
-    ]
-    valid_postcodes = list(set(valid_postcodes))
-
-    if not valid_postcodes:
-        return results
-
-    batch_size = 100
-    total_batches = (len(valid_postcodes) + batch_size - 1) // batch_size
-
-    for i, batch_start in enumerate(range(0, len(valid_postcodes), batch_size)):
-        batch = valid_postcodes[batch_start : batch_start + batch_size]
-        print(f"  Geocoding batch {i + 1}/{total_batches} ({len(batch)} postcodes)...")
-
-        try:
-            response = requests.post(
-                "https://api.postcodes.io/postcodes",
-                json={"postcodes": batch},
-                timeout=30,
-            )
-            if response.status_code == 200:
-                data = response.json()
-                for item in data.get("result", []):
-                    if item and item.get("result"):
-                        pc = item["query"].upper()
-                        lat = item["result"].get("latitude")
-                        lon = item["result"].get("longitude")
-                        if lat and lon:
-                            results[pc] = (lat, lon)
-            else:
-                print(f"    Warning: API returned status {response.status_code}")
-        except Exception as e:
-            print(f"    Warning: Geocoding batch failed: {e}")
-
-    return results
-
-
-def geocode_schools(force: bool = False) -> None:
-    """
-    Geocode all schools in the database.
-
-    Args:
-        force: If True, re-geocode all postcodes even if already geocoded
-    """
-    print(f"\n{'='*60}")
-    print(f"School Geocoding Job - {datetime.now().isoformat()}")
-    print(f"{'='*60}\n")
-
-    db = SessionLocal()
-
-    try:
-        # Get schools that need geocoding
-        if force:
-            schools = db.query(School).filter(School.postcode.isnot(None)).all()
-            print(f"Force mode: Processing all {len(schools)} schools with postcodes")
-        else:
-            schools = db.query(School).filter(
-                School.postcode.isnot(None),
-                (School.latitude.is_(None)) | (School.longitude.is_(None))
-            ).all()
-            print(f"Found {len(schools)} schools without coordinates")
-
-        if not schools:
-            print("No schools to geocode. Exiting.")
-            return
-
-        # Extract unique postcodes
-        postcodes = list(set(
-            s.postcode.strip().upper()
-            for s in schools
-            if s.postcode
-        ))
-        print(f"Unique postcodes to geocode: {len(postcodes)}")
-
-        # Geocode in bulk
-        print("\nGeocoding postcodes...")
-        geocoded = geocode_postcodes_bulk(postcodes)
-        print(f"Successfully geocoded: {len(geocoded)} postcodes")
-
-        # Update database
-        print("\nUpdating database...")
-        updated_count = 0
-        failed_count = 0
-
-        for school in schools:
-            if not school.postcode:
-                continue
-
-            pc_upper = school.postcode.strip().upper()
-            coords = geocoded.get(pc_upper)
-
-            if coords:
-                school.latitude = coords[0]
-                school.longitude = coords[1]
-                updated_count += 1
-            else:
-                failed_count += 1
-
-        db.commit()
-
-        print(f"\nResults:")
-        print(f"  - Updated: {updated_count} schools")
-        print(f"  - Failed (invalid/not found): {failed_count} postcodes")
-
-        # Summary stats
-        total_with_coords = db.query(School).filter(
-            School.latitude.isnot(None),
-            School.longitude.isnot(None)
-        ).count()
-        total_schools = db.query(School).count()
-
-        print(f"\nDatabase summary:")
-        print(f"  - Total schools: {total_schools}")
-        print(f"  - Schools with coordinates: {total_with_coords}")
-        print(f"  - Coverage: {100*total_with_coords/total_schools:.1f}%")
-
-    except Exception as e:
-        print(f"Error during geocoding: {e}")
-        db.rollback()
-        raise
-    finally:
-        db.close()
-        print(f"\n{'='*60}")
-        print(f"Geocoding job completed - {datetime.now().isoformat()}")
-        print(f"{'='*60}\n")
-
-
-def main():
-    parser = argparse.ArgumentParser(
-        description="Geocode school postcodes and update database"
-    )
-    parser.add_argument(
-        "--force",
-        action="store_true",
-        help="Re-geocode all postcodes, even if already geocoded"
-    )
-    args = parser.parse_args()
-
-    geocode_schools(force=args.force)
-
-
-if __name__ == "__main__":
-    main()

scripts/migrate_csv_to_db.py

@@ -1,68 +0,0 @@
-#!/usr/bin/env python3
-"""
-CLI script for manual database migration.
-
-Usage:
-    python scripts/migrate_csv_to_db.py [--drop] [--geocode]
-
-Options:
-    --drop      Drop existing tables before migration (full reimport)
-    --geocode   Geocode postcodes (requires network access)
-"""
-
-import sys
-from pathlib import Path
-
-# Add parent directory to path for imports
-sys.path.insert(0, str(Path(__file__).parent.parent))
-
-import argparse
-
-from backend.config import settings
-from backend.database import Base, engine, init_db, set_db_schema_version
-from backend.migration import load_csv_data, migrate_data, run_full_migration
-from backend.version import SCHEMA_VERSION
-
-
-def main():
-    parser = argparse.ArgumentParser(
-        description="Migrate CSV data to PostgreSQL database"
-    )
-    parser.add_argument(
-        "--drop", action="store_true", help="Drop existing tables before migration"
-    )
-    parser.add_argument("--geocode", action="store_true", help="Geocode postcodes")
-    args = parser.parse_args()
-
-    print("=" * 60)
-    print("School Data Migration: CSV -> PostgreSQL")
-    print("=" * 60)
-    print(f"\nDatabase: {settings.database_url.split('@')[-1]}")
-    print(f"Data directory: {settings.data_dir}")
-    print(f"Target schema version: {SCHEMA_VERSION}")
-
-    if args.drop:
-        print("\nRunning full migration (drop + reimport)...")
-        success = run_full_migration(geocode=args.geocode)
-    else:
-        print("\nCreating tables (preserving existing data)...")
-        init_db()
-        print("\nLoading CSV data...")
-        df = load_csv_data(settings.data_dir)
-        if df.empty:
-            print("No data found to migrate!")
-            return 1
-        migrate_data(df, geocode=args.geocode)
-        success = True
-
-    if success:
-        # Ensure schema_version table exists
-        init_db()
-        set_db_schema_version(SCHEMA_VERSION)
-        print(f"\nSchema version set to {SCHEMA_VERSION}")
-
-    return 0 if success else 1
-
-
-if __name__ == "__main__":
-    sys.exit(main())