docs(flags): Unleash does not create flags by itself

The runbook said a flag 'appears in the Unleash UI after the backend has
evaluated it once'. That is wrong. SDKs read definitions from the server
and never register anything, and metrics for an unknown flag are
discarded — so a declared flag is evaluated on every request, stays
False forever, and never shows up until someone creates it by hand.

Found the way these things usually are: staging had been running the
flag code for a while and the UI was still empty.

Also names the environment trap while here — each stack's token is
scoped to one environment, so toggling the other does nothing visible
and looks like the flag is broken.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015mWQnpye9F299NVRCCSRvj

backend/app.py

@@ -35,6 +35,7 @@ from .data_loader import (
     search_schools_typesense,
 )
 from .data_loader import get_data_info as get_db_info
+from . import flags
 from .places import build_place_registry
 from .schemas import METRIC_DEFINITIONS, RANKING_COLUMNS, SCHOOL_COLUMNS
 from .utils import clean_for_json, convert_to_native
@@ -459,6 +460,7 @@ def validate_postcode(postcode: Optional[str]) -> Optional[str]:
 async def lifespan(app: FastAPI):
     """Application lifespan - startup and shutdown events."""
     global _sitemaps
+    flags.init()
     print("Loading school data from marts...")
     df = load_school_data()
     if df.empty:
@@ -806,7 +808,13 @@ async def get_school_details(request: Request, urn: int):
         "census": supplementary.get("census"),
         "admissions": supplementary.get("admissions"),
         "admissions_history": supplementary.get("admissions_history") or [],
-        "admission_distance": supplementary.get("admission_distance"),
+        # Behind a flag, and withheld at the source rather than rendered-but-
+        # hidden: this endpoint is public and unauthenticated, so a field left
+        # in the payload is a published field. The key is absent, not null —
+        # null would state that this school has no cut-off, which is a
+        # different claim from "we are not publishing cut-offs".
+        **({"admission_distance": supplementary.get("admission_distance")}
+           if flags.is_enabled("admission_distance") else {}),
         "sen_detail": supplementary.get("sen_detail"),
         "phonics": supplementary.get("phonics"),
         "deprivation": supplementary.get("deprivation"),
@@ -1263,6 +1271,18 @@ async def get_place(request: Request, kind: str, slug: str,
     }
 
 
+@app.get("/api/flags")
+@limiter.limit(f"{settings.rate_limit_per_minute}/minute")
+async def get_feature_flags(request: Request):
+    """Every declared flag and its current value.
+
+    Internal only. The Next proxy denies this path, because the response names
+    every unreleased feature the codebase knows about — which is exactly what
+    shipping dark is meant to keep quiet.
+    """
+    return flags.all_flags()
+
+
 @app.get("/api/data-info")
 @limiter.limit(f"{settings.rate_limit_per_minute}/minute")
 async def get_data_info(request: Request):

backend/config.py

@@ -42,6 +42,16 @@ class Settings(BaseSettings):
     typesense_url: str = "http://localhost:8108"
     typesense_api_key: str = ""
 
+    # Feature flags (Unleash). An empty unleash_url disables flags entirely and
+    # every flag evaluates False — the correct behaviour for local development
+    # and CI, and the reason no test needs a running Unleash.
+    unleash_url: str = ""
+    unleash_api_token: str = ""
+    unleash_app_name: str = "schoolcompare-backend"
+    # On a named volume, so a restart during an Unleash outage keeps
+    # last-known state instead of reverting a released feature to dark.
+    unleash_cache_directory: str = "/app/.unleash"
+
     # Analytics
     ga_measurement_id: Optional[str] = "G-J0PCVT14NY"  # Google Analytics 4 Measurement ID
 

backend/flags.py

@@ -0,0 +1,111 @@
+"""Feature flags: what can be switched, and what is switched right now.
+
+Ship-dark, not a kill switch. Flags let work merge and deploy without becoming
+visible; they are expected to flip about monthly, by a person, deliberately.
+Nothing here does percentage rollouts or user targeting — the site has no user
+identity to target.
+
+Unleash holds the state. It does not hold the list. REGISTRY below is that
+list, and it exists for three reasons: the SDK evaluates an unknown flag to
+False, so without a registry that is an *undeclared* False, indistinguishable
+from a typo; /api/flags needs a key set to return when Unleash is unreachable;
+and a flag in the UI but not in the registry is orphaned and should be visibly
+so rather than quietly authoritative.
+
+Every flag defaults to False. There is no per-flag default, because a flag that
+defaults on is a kill switch, and this is not one.
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass
+from datetime import date
+
+from .config import settings
+
+logger = logging.getLogger(__name__)
+
+# A flag is temporary scaffolding. See test_a_flag_older_than_the_limit.
+MAX_FLAG_AGE_DAYS = 90
+
+
+@dataclass(frozen=True)
+class Flag:
+    # One string: the registry key, the Unleash flag name, and the JSON key in
+    # /api/flags. snake_case, matching the API's existing convention. No case
+    # transformation anywhere, so there is no mapping layer to get wrong.
+    name: str
+    description: str          # one line: what turning this on reveals
+    added: date               # for the staleness tripwire
+
+
+REGISTRY: dict[str, Flag] = {
+    f.name: f for f in (
+        Flag(
+            name="admission_distance",
+            description=(
+                "The last-distance-offered figure on the Admissions tile and "
+                "the 'How far away are you?' section on school pages."
+            ),
+            added=date(2026, 8, 23),
+        ),
+    )
+}
+
+
+_client = None
+
+
+def init() -> None:
+    """Start the Unleash client, or log why flags are all off.
+
+    Called once from the app lifespan. Never raises: a flag system that can
+    stop the API from booting is worse than one that is switched off.
+    """
+    global _client
+    if not settings.unleash_url or not settings.unleash_api_token:
+        logger.warning(
+            "Unleash is not configured (UNLEASH_URL / UNLEASH_API_TOKEN); "
+            "every feature flag evaluates to False.")
+        return
+
+    try:
+        from UnleashClient import UnleashClient
+
+        _client = UnleashClient(
+            url=settings.unleash_url,
+            app_name=settings.unleash_app_name,
+            custom_headers={"Authorization": settings.unleash_api_token},
+            cache_directory=settings.unleash_cache_directory,
+            refresh_interval=15,
+        )
+        _client.initialize_client()
+        logger.info("Unleash client initialised against %s", settings.unleash_url)
+    except Exception:
+        # Fail closed and keep serving. The SDK also evaluates everything False
+        # until its first successful sync, so this is the same direction.
+        _client = None
+        logger.exception("Unleash client failed to start; flags are all False.")
+
+
+def is_enabled(name: str) -> bool:
+    """Whether `name` is on. False for anything unknown, unreachable or broken."""
+    if name not in REGISTRY:
+        logger.error(
+            "undeclared feature flag %r was evaluated; returning False. "
+            "Add it to backend/flags.py REGISTRY or fix the name.", name)
+        return False
+    if _client is None:
+        return False
+    try:
+        return bool(_client.is_enabled(
+            name, fallback_function=lambda feature_name, context: False))
+    except Exception:
+        logger.exception("flag %r failed to evaluate; returning False", name)
+        return False
+
+
+def all_flags() -> dict[str, bool]:
+    """Every declared flag and its current value. Serves /api/flags."""
+    return {name: is_enabled(name) for name in REGISTRY}

backend/tests/test_flags.py

@@ -0,0 +1,163 @@
+"""Tests for the feature flag layer (spec 2026-08-23).
+
+None of these need a running Unleash. That is the point: an unset UNLEASH_URL
+means every flag is False, which is what local development and CI get.
+"""
+
+from datetime import date, timedelta
+
+from backend import flags
+
+
+def test_every_declared_flag_is_keyed_by_its_own_name():
+    # One string is the registry key, the Unleash flag name and the JSON key.
+    # A mismatch here would mean the UI toggles a flag the code never reads.
+    for key, flag in flags.REGISTRY.items():
+        assert key == flag.name
+
+
+def test_flag_names_are_snake_case():
+    # Matches the API's existing convention (admission_distance,
+    # rwm_expected_pct) so no case transformation exists to get wrong.
+    for name in flags.REGISTRY:
+        assert name == name.lower()
+        assert "-" not in name and " " not in name
+
+
+def test_an_unconfigured_client_evaluates_every_flag_false(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    for name in flags.REGISTRY:
+        assert flags.is_enabled(name) is False
+
+
+def test_an_undeclared_flag_is_false_rather_than_an_error(monkeypatch):
+    # A typo'd flag name must not raise in a request path. It is logged as an
+    # error, because an undeclared flag is always a bug.
+    monkeypatch.setattr(flags, "_client", None)
+    assert flags.is_enabled("no_such_flag") is False
+
+
+def test_an_exploding_client_is_false_rather_than_a_500(monkeypatch):
+    class Boom:
+        def is_enabled(self, *a, **kw):
+            raise RuntimeError("unleash is on fire")
+
+    monkeypatch.setattr(flags, "_client", Boom())
+    name = next(iter(flags.REGISTRY))
+    assert flags.is_enabled(name) is False
+
+
+def test_all_flags_reports_every_declared_flag(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    assert set(flags.all_flags()) == set(flags.REGISTRY)
+    assert all(v is False for v in flags.all_flags().values())
+
+
+def test_a_flag_older_than_the_limit_fails_this_test():
+    """A tripwire, not an assertion about correctness.
+
+    Flags are temporary scaffolding and the failure mode of every flag system
+    is accumulation. This fails on the day a flag turns 90, on whatever PR
+    happens to be open — which is the point: someone has to decide.
+
+    To fix: delete the flag and the branches that read it, or, if it genuinely
+    still needs to exist, move its `added` date and say why in the commit.
+    """
+    stale = [
+        f.name for f in flags.REGISTRY.values()
+        if date.today() - f.added > timedelta(days=flags.MAX_FLAG_AGE_DAYS)
+    ]
+    assert not stale, (
+        f"Flags older than {flags.MAX_FLAG_AGE_DAYS} days: {stale}. "
+        "Remove the flag and the code branches it guards, or move its `added` "
+        "date deliberately."
+    )
+
+
+def _client():
+    from fastapi.testclient import TestClient
+    from backend import app as app_module
+    return TestClient(app_module.app, raise_server_exceptions=False)
+
+
+def test_the_flags_endpoint_lists_every_declared_flag(monkeypatch):
+    monkeypatch.setattr(flags, "_client", None)
+    body = _client().get("/api/flags").json()
+    assert set(body) == set(flags.REGISTRY)
+
+
+def test_the_flags_endpoint_answers_false_when_unleash_is_unreachable(monkeypatch):
+    # The endpoint must still answer. A frontend that cannot read flags renders
+    # everything dark, which is right; one that gets a 500 renders nothing.
+    monkeypatch.setattr(flags, "_client", None)
+    res = _client().get("/api/flags")
+    assert res.status_code == 200
+    assert all(v is False for v in res.json().values())
+
+
+def _school_payload(monkeypatch, *, flag_on: bool):
+    """Fetch one school's payload with the distance flag forced on or off.
+
+    The DataFrame shape is copied from test_school_details.py rather than
+    minimised: the endpoint reads a wide set of GIAS columns, and a trimmed
+    frame fails for reasons that have nothing to do with flags.
+    """
+    import numpy as np
+    import pandas as pd
+    from fastapi.testclient import TestClient
+    from backend import app as app_module
+
+    df = pd.DataFrame([{
+        "urn": 150275,
+        "school_name": "West London Performing Arts Academy",
+        "phase": "Secondary",
+        "school_type": "Special post 16 institution",
+        "trust_name": None,
+        "religious_denomination": "Does not apply",
+        "gender": None,
+        "age_range": "16-25",
+        "admissions_policy": None,
+        "capacity": np.nan,
+        "gias_total_pupils": np.nan,
+        "headteacher_name": None,
+        "website": None,
+        "ofsted_grade": np.nan,
+        "local_authority": "Ealing",
+        "address": "268 Northfield Avenue, London, W5 4UB",
+        "postcode": "W5 4UB",
+        "latitude": 51.4986,
+        "longitude": -0.3148,
+        "year": np.nan,
+        "total_pupils": np.nan,
+        "eligible_pupils": np.nan,
+        "rwm_expected_pct": np.nan,
+    }])
+
+    monkeypatch.setattr(app_module, "load_school_data", lambda: df)
+    # Two arguments: get_supplementary_data(db, urn). See backend/app.py.
+    monkeypatch.setattr(
+        app_module, "get_supplementary_data",
+        lambda db, urn: {"admission_distance": {"distance_m": 772.49,
+                                                "year": 2024}})
+    monkeypatch.setattr(flags, "is_enabled", lambda name: flag_on)
+
+    client = TestClient(app_module.app, raise_server_exceptions=False)
+    res = client.get("/api/schools/150275")
+    assert res.status_code == 200, res.text
+    return res.json()
+
+
+def test_the_distance_field_is_absent_when_the_flag_is_off(monkeypatch):
+    """Absent, not null, and withheld at the source.
+
+    /api/schools/ is public and unauthenticated. Leaving a withheld field in
+    the payload while declining to render it hands the record to anyone who
+    opens the network tab — the reasoning already recorded in c9a1892.
+    """
+    body = _school_payload(monkeypatch, flag_on=False)
+    assert "admission_distance" not in body
+
+
+def test_the_distance_field_is_present_when_the_flag_is_on(monkeypatch):
+    body = _school_payload(monkeypatch, flag_on=True)
+    assert body["admission_distance"]["distance_m"] == 772.49

docker-compose.portainer.staging.yml

@@ -16,6 +16,8 @@
 #   ADMIN_API_KEY          — Backend admin API key
 #   TYPESENSE_API_KEY      — Typesense admin API key
 #   TYPESENSE_SEARCH_KEY   — Typesense search-only key (exposed to frontend)
+#   UNLEASH_URL            — http://<unleash-ip>:4242/api (empty = all flags off)
+#   UNLEASH_API_TOKEN      — Unleash *client* token, environment: development
 #   AIRFLOW_ADMIN_USER     — Airflow admin username (password auto-generated, see api-server logs)
 #   STAGING_DB_IP          — macvlan IP for staging Postgres (default 10.0.1.190)
 #   STAGING_FRONTEND_IP    — macvlan IP for staging frontend (default 10.0.1.151)
@@ -55,6 +57,12 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
+    volumes:
+      - unleash_cache:/app/.unleash
     depends_on:
       sc_database:
         condition: service_healthy
@@ -212,3 +220,4 @@ volumes:
   postgres_data:
   typesense_data:
   airflow_logs:
+  unleash_cache:

docker-compose.portainer.unleash.yml

@@ -0,0 +1,73 @@
+# Portainer Stack Definition for School Compare — UNLEASH (feature flags)
+#
+# Deploy as a *separate* Portainer stack ("schoolcompare-unleash"), alongside
+# the production and staging stacks. It deliberately belongs to neither: a
+# staging redeploy must not be able to disturb production's flag state, and a
+# production redeploy must not disturb staging's.
+#
+# One instance serves both environments. Open-source Unleash ships with
+# `development` and `production` environments and environment-scoped client
+# tokens, so the same flag holds independent state in each — which is what
+# lets a feature be on in staging, where the E2E journeys exercise it, while
+# production stays dark.
+#
+# Portainer environment variables (set in Portainer UI -> Stack -> Environment):
+#   UNLEASH_DB_PASSWORD    — PostgreSQL password for the Unleash database
+#   UNLEASH_ADMIN_PASSWORD — initial admin password for the Unleash UI
+#   UNLEASH_IP             — macvlan IP for the Unleash server (default 10.0.1.152)
+
+services:
+
+  # ── PostgreSQL (Unleash's own; nothing else uses it) ──────────────────
+  unleash_db:
+    container_name: sc_unleash_postgres
+    image: postgres:16-alpine
+    environment:
+      POSTGRES_USER: unleash
+      POSTGRES_PASSWORD: ${UNLEASH_DB_PASSWORD}
+      POSTGRES_DB: unleash
+    volumes:
+      - unleash_postgres_data:/var/lib/postgresql/data
+    networks:
+      - unleash
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U unleash"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+      start_period: 10s
+    restart: unless-stopped
+
+  # ── Unleash server (UI + client API on 4242) ──────────────────────────
+  unleash:
+    container_name: sc_unleash
+    image: unleashorg/unleash-server:6
+    environment:
+      DATABASE_URL: postgres://unleash:${UNLEASH_DB_PASSWORD}@unleash_db:5432/unleash
+      DATABASE_SSL: "false"
+      INIT_ADMIN_API_TOKENS: ""
+      UNLEASH_DEFAULT_ADMIN_PASSWORD: ${UNLEASH_ADMIN_PASSWORD}
+    depends_on:
+      unleash_db:
+        condition: service_healthy
+    networks:
+      unleash: {}
+      macvlan:
+        ipv4_address: ${UNLEASH_IP:-10.0.1.152}
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:4242/health || exit 1"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 30s
+    restart: unless-stopped
+
+networks:
+  unleash:
+    driver: bridge
+  macvlan:
+    external:
+      name: macvlan
+
+volumes:
+  unleash_postgres_data:

docker-compose.portainer.yml

@@ -7,6 +7,8 @@
 #   ADMIN_API_KEY          — Backend admin API key
 #   TYPESENSE_API_KEY      — Typesense admin API key
 #   TYPESENSE_SEARCH_KEY   — Typesense search-only key (exposed to frontend)
+#   UNLEASH_URL            — http://<unleash-ip>:4242/api (empty = all flags off)
+#   UNLEASH_API_TOKEN      — Unleash *client* token, environment: production
 #   AIRFLOW_ADMIN_USER     — Airflow admin username (password auto-generated, see api-server logs)
 
 services:
@@ -44,6 +46,12 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
+    volumes:
+      - unleash_cache:/app/.unleash
     depends_on:
       sc_database:
         condition: service_healthy
@@ -201,3 +209,4 @@ volumes:
   postgres_data:
   typesense_data:
   airflow_logs:
+  unleash_cache:

docker-compose.yml

@@ -36,6 +36,10 @@ services:
       ADMIN_API_KEY: ${ADMIN_API_KEY:-changeme}
       TYPESENSE_URL: http://typesense:8108
       TYPESENSE_API_KEY: ${TYPESENSE_API_KEY:-changeme}
+      # Unset means every feature flag is False — the correct dark state for an
+      # environment with no Unleash, not a failure.
+      UNLEASH_URL: ${UNLEASH_URL:-}
+      UNLEASH_API_TOKEN: ${UNLEASH_API_TOKEN:-}
     volumes:
       - ./data:/app/data:ro
     depends_on:

docs/DEPLOY.md

@@ -150,3 +150,73 @@ token Gitea Actions provides automatically (`secrets.GITEA_TOKEN` — no setup
 needed), and fails the check only when a finding is rated
 **severe** (would break prod, leak data, or corrupt data). Minor findings are
 informational and never block a merge.
+
+## Feature flags (Unleash)
+
+Flag state lives in a self-hosted Unleash instance, deployed as its own
+Portainer stack from `docker-compose.portainer.unleash.yml`. It is separate
+from the application stacks on purpose — redeploying staging must not be able
+to disturb production's flags.
+
+The flags themselves are declared in `backend/flags.py`. Unleash holds the
+state; the registry holds the list. A flag in the UI that is not in the
+registry is orphaned and nothing reads it.
+
+### First-time setup
+
+1. Deploy the stack in Portainer. Set `UNLEASH_DB_PASSWORD`,
+   `UNLEASH_ADMIN_PASSWORD` and (optionally) `UNLEASH_IP`.
+2. Log in to the UI at `http://<UNLEASH_IP>:4242` as `admin`.
+3. Create one **client** API token per environment:
+   - `schoolcompare-staging`, environment **development**
+   - `schoolcompare-prod`, environment **production**
+
+   Client tokens, not admin tokens — the backend only reads.
+4. Put each token in the matching Portainer stack's `UNLEASH_API_TOKEN`
+   variable, and set `UNLEASH_URL` to `http://<UNLEASH_IP>:4242/api`.
+5. Redeploy the application stacks.
+
+### Adding a flag to Unleash
+
+**Unleash does not create flags by itself.** The SDK reads definitions from the
+server and never registers anything, and metrics for a flag the server has
+never heard of are discarded. So a flag declared in `backend/flags.py` will be
+evaluated on every request, stay `False` forever, and never appear in the UI
+until someone creates it there by hand.
+
+For each flag in the registry, create one in Unleash with:
+
+- **Name** — character for character what `backend/flags.py` declares.
+  snake_case, no hyphens or spaces. A typo produces a flag that looks correct
+  in the UI and is read by nothing.
+- **Type** — Release. No strategies, constraints or variants: these are plain
+  on/off switches, by design.
+
+### Turning a feature on
+
+Toggle the flag in the environment matching the stack you mean: **development**
+for staging, **production** for prod. The token in each stack is scoped to one
+environment, so toggling the other one has no visible effect.
+
+The SDK refreshes every 15 seconds, so the API reflects the change almost at
+once; the pages follow on their own schedule, below.
+
+A flip reaches school pages within about five minutes and place pages within
+the hour. Next's ISR does the propagating — it revalidates a route at the
+*lowest* `revalidate` among that route's fetches, which is 300s for
+`/school/[slug]` and 3600s for the place pages. There is no webhook, and
+adding one would only be worth it if flips ever needed to be instant.
+
+### When Unleash is unreachable
+
+Every flag evaluates to `False` and the site serves as though nothing were
+switched on. That is deliberate — an unfinished feature staying hidden is the
+safe direction — but it means a *released* feature disappears if a backend
+container cold-starts with an empty cache while Unleash is down. The SDK's
+disk cache is on a named volume so restarts keep last-known state, and flags
+are removed from the code within 90 days (enforced by a test), which bounds
+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.

docs/superpowers/specs/2026-08-23-feature-flags-design.md

@@ -0,0 +1,294 @@
+# Feature Flags — Design
+
+**Date:** 2026-08-23
+**Status:** approved for planning
+**First consumer:** the last-distance-offered feature (`admission_distance`)
+
+## Goal
+
+Let work merge to `main` and deploy to production without becoming visible,
+so that releasing a feature stops being the same event as deploying it.
+
+The site has no way to do this today. A feature is either on `main` and live,
+or it is on a branch. That forces long-lived branches for anything not ready,
+and it makes every promotion to production an all-or-nothing decision about
+everything queued behind it.
+
+This is a **ship-dark** capability, not a kill switch. Flags are expected to
+flip on the order of once a month, by a person, deliberately. Nothing here is
+designed for flipping something off in seconds under pressure, and nothing
+here does percentage rollouts, user targeting or A/B tests — the site has no
+user identity to target.
+
+## Decision: Unleash
+
+Flag state is held in a self-hosted [Unleash](https://www.getunleash.io/)
+instance (Apache-2.0), not in the repository.
+
+A lighter option was considered and rejected by the project owner: a typed
+registry in each runtime with environment-variable overrides set in the
+Portainer stack files, which would have needed no new container and kept flag
+state in git. The argument for Unleash is that it provides a UI and an audit
+log without a deploy, and that flags are expected to become an ongoing
+operational tool rather than an occasional one.
+
+Two consequences follow from choosing a service, and this design exists mostly
+to handle them:
+
+1. **Flag state lives outside the repository.** `main` is no longer the whole
+   truth about what is switched on. The registry in §2 exists to bound that.
+2. **A flag can change without a deploy**, so nothing else clears the caches
+   that a deploy would have cleared. §4 establishes how long a flip takes to
+   become visible, and why that is short enough to need no extra mechanism.
+
+Also considered: Flagsmith (heavier — Django, Postgres and Redis), GrowthBook
+(requires MongoDB), and Flipt v2 (the closest conceptual fit, git-native, but
+now under the Fair Core Licence — source-available, not OSI open source).
+
+## 1. Topology
+
+A third Portainer stack, `docker-compose.portainer.unleash.yml`, holding
+`unleashorg/unleash-server` and its own PostgreSQL 16. It is on the macvlan so
+both application stacks can reach it, and it belongs to neither of them — a
+staging redeploy must not be able to disturb production's flag state, and vice
+versa.
+
+One instance serves both environments. Open-source Unleash ships with
+`development` and `production` environments and environment-scoped client
+tokens, so the same flag holds independent state in each: staging's FastAPI
+carries a `development` token, production's carries a `production` one.
+
+That property is what makes ship-dark testable. A feature can be **on in
+staging and off in production** for as long as it takes, which means the `e2e/`
+journeys exercise it against staging while production stays unchanged.
+
+## 2. The registry
+
+Unleash supplies flag *state* and the toggle UI. It does not supply the list of
+flags. `backend/flags.py` declares every flag the code knows about:
+
+```python
+@dataclass(frozen=True)
+class Flag:
+    name: str            # identical in the registry, in Unleash, and in JSON
+    description: str     # one line: what turning this on reveals
+    added: date          # for the staleness test in §8
+```
+
+**Every flag defaults to `False`.** There is no per-flag default field, because
+a flag that defaults on is not a ship-dark flag — it is a kill switch, and this
+design does not offer one. A single unconditional default also means the
+fallback path has no branching to get wrong.
+
+Three reasons the registry is not optional:
+
+- The Unleash SDK evaluates an unknown flag to `False`. Without a registry that
+  is an *undeclared* false — indistinguishable from a typo in a flag name.
+- `/api/flags` needs a key set to return when Unleash is unreachable. It cannot
+  enumerate flags it has never heard of.
+- A flag present in the Unleash UI but absent from the registry is orphaned,
+  and should be visibly so rather than quietly authoritative.
+
+**Naming.** One string, used unchanged as the registry key, the Unleash flag
+name, and the JSON key in `/api/flags`. It is snake_case, matching the API's
+existing convention (`admission_distance`, `rwm_expected_pct`) and the mirrored
+types in `nextjs-app/lib/types.ts`. No case transformation anywhere, so there
+is no mapping layer to get wrong.
+
+## 3. Read paths
+
+### Backend
+
+`backend/flags.py` wraps `UnleashClient` behind `is_enabled(name: str) -> bool`.
+
+Fail-closed is the default rather than something added: the Python SDK
+evaluates every flag to `False` until it has synchronised with the server. An
+unfinished feature therefore stays hidden when Unleash is unreachable, which is
+the correct direction for ship-dark.
+
+The SDK's fcache directory is mounted on a named volume so a container restart
+during an Unleash outage keeps last-known state rather than reverting a
+released feature to dark. The registry default remains `False`, so the worst
+case is a feature disappearing, never one appearing.
+
+### Frontend
+
+`nextjs-app/lib/flags.ts` exposes `getFlags(): Promise<Flags>`, a single
+server-side fetch of `/api/flags` returning a typed record. Server components
+only — no flag value reaches the browser bundle, and `package.json` gains no
+Unleash dependency. The Unleash client library stays entirely inside the
+service that already owns every other piece of data the frontend renders.
+
+The cost, named plainly: a purely front-end flag must still be declared in a
+Python file. It is a flat data edit rather than programming, and the return is
+one list, so nobody has to ask which service knows about a given flag.
+
+### `/api/flags` must not be publicly reachable
+
+`nextjs-app/app/api/[...path]/route.ts` proxies **everything** under `/api/` to
+FastAPI. Left alone, `https://www.schoolcompare.co.uk/api/flags` would return
+`{"admission_distance": false, ...}` — publishing the name and state of every
+unreleased feature, which defeats the purpose of shipping dark.
+
+The proxy therefore gains a denylist, and `flags` is on it: a request for a
+denied path returns 404 rather than being forwarded. Next's own `getFlags()` is
+unaffected because it calls `FASTAPI_URL` directly across the Docker network
+and never transits the public proxy.
+
+This is a general hole rather than a flags-specific one — the proxy will
+forward any future internal endpoint too — so the denylist is written as a
+named constant with a comment saying what belongs on it.
+
+## 4. Propagation
+
+**Time-based revalidation is sufficient. There is no webhook.**
+
+An earlier draft of this section specified two Unleash webhooks and a
+`revalidateTag('flags')` purge, on the premise that pages cache for seven days.
+That premise was wrong, and checking it removed the most complex part of the
+design.
+
+Next uses the **lowest** `revalidate` among a route's fetches to set the
+revalidation frequency of the whole route — the segment-level
+`export const revalidate` does not override a lower value inside it. Measured
+against this codebase:
+
+| Page family | Segment | Lowest fetch | Effective |
+|---|---|---|---|
+| `/school/[slug]` | 604800 | `fetchSchoolDetails` at 300 | **5 minutes** |
+| `/schools/*` | 604800 | `fetchNationalAverages` at 3600 | **1 hour** |
+
+The Unleash SDK polls every 15 seconds, so a flip reaches school pages within
+about five minutes and place pages within the hour, unaided. Flags flip
+monthly, by hand, deliberately. That is fast enough.
+
+What this removes: two webhook integrations, a `/api/revalidate-flags` route, a
+shared-secret-in-a-query-string scheme, an idempotency requirement against
+duplicate and out-of-order delivery, and a rule that every fetch in
+`nextjs-app/lib/` carry a cache tag. None of it has to be built, maintained, or
+kept correct as new fetches are added.
+
+**If instant flips are ever wanted**, the webhook is the way to add them, and it
+is purely additive — nothing in this design has to change first.
+
+### Two constraints this leaves behind
+
+**Never flag content on a `force-static` page.** `app/admissions/page.tsx`
+declares `export const dynamic = 'force-static'`, so it is baked at build time
+and never revalidates. A flag gating anything on such a page would not take
+effect until the next deploy, silently. If a flag ever needs to reach one, that
+page must first move to ISR.
+
+**A route-family flag still needs the sitemap rebuilt.** The sitemap is held in
+memory and rebuilt only at startup or via `POST /api/admin/regenerate-sitemap`.
+No flag in scope touches the sitemap (§6), so this is deferred with the route
+case rather than solved now — but a route flag must not ship without it, or the
+sitemap will advertise URLs that `notFound()`.
+
+## 5. What "off" means, per surface
+
+| Surface | Off |
+|---|---|
+| Route | `notFound()`, **and** absent from the sitemap, **and** absent from nav |
+| UI element | Not rendered; surrounding page byte-identical to today |
+| API field | Key **absent**, not `null` |
+| API endpoint | 404, not 403 |
+
+The three parts of the route rule move together or not at all. Submitting URLs
+to Google that return 404 is the bug fixed in PR #124, and a flag is a new way
+to reintroduce it.
+
+An API field is withheld **at the source**, never rendered-but-hidden. The
+precedent is already set in this codebase by commit `c9a1892`: `/api/schools/`
+is public and unauthenticated, so leaving a withheld field in the payload hands
+the record to anyone who opens the network tab.
+
+## 6. First consumer: `admission_distance`
+
+The last-distance-offered feature is merged to `main` and live on staging.
+Production has never received it: `/api/schools/100010` on production carries
+no `admission_distance` key, and no Distance section renders.
+
+It needs **exactly one gate** — `backend/app.py:809`, where the field is
+attached to the school payload:
+
+```python
+"admission_distance": (
+    supplementary.get("admission_distance")
+    if flags.is_enabled("admission_distance") else None
+),
+```
+
+The frontend follows with no change. `DistanceSection` already returns `null`
+when `admission_distance?.distance_m == null`, and `PrimarySchoolSections`
+already conditions the admissions block on `(admissions || admissionDistance)`.
+The off-state is the commonest state on the site — only 57 local authorities
+publish cut-off distances at all — so it is well covered by construction.
+
+The flag does not touch the sitemap: school pages exist either way.
+
+Intended lifecycle: default off, so production receives the code dark on the
+next promotion; on in the `development` environment so staging keeps testing
+it; flipped on in `production` when the owner chooses.
+
+**This flag exercises two of the three surfaces** in §5 — API field and UI
+element. No route case ships with it. The route rule is specified but unproven
+until a route-shaped flag exists, and should be treated as such.
+
+## 7. Testing
+
+**Backend unit.** The registry is well-formed; an unknown flag evaluates
+`False`; `/api/flags` returns every declared flag with its default when the
+SDK is unreachable; `admission_distance` is absent from the school payload when
+the flag is off and present when on.
+
+**Frontend unit.** `getFlags()` returns declared defaults when `/api/flags`
+fails, rather than throwing and taking the page with it.
+
+**E2E.** Journeys read `/api/flags` and gate flag-dependent assertions on it,
+matching the `test.skip` shape the suite already uses.
+
+One trap to avoid, worth stating because the existing distance journeys walk
+straight into it: they already skip when no school has a published figure, so
+with the flag off they would skip silently and the suite would go green. The
+gate must be explicit — **if `/api/flags` reports `admission_distance` on, then
+a school with a cut-off must be found**, converting a silent skip into a real
+assertion.
+
+## 8. Lifecycle
+
+A flag is temporary scaffolding, and the failure mode of every flag system is
+accumulation.
+
+The registry records the date each flag was added, and a backend test fails any
+flag older than **90 days**. Removing a flag means deleting the registry entry,
+the branches that read it, and the flag in the Unleash UI.
+
+Unleash SDK usage metrics stay enabled, so the UI shows which flags are still
+being evaluated — the evidence needed to retire one safely.
+
+## 9. Risks
+
+**Production gains a homelab dependency.** If Unleash is unreachable when a
+production container cold-starts with an empty cache, every flag evaluates
+`False` and any feature currently switched on disappears. The fcache volume
+covers restarts; the 90-day lifecycle rule bounds how long any feature is
+exposed to this. It is a real regression risk and the reason flags must be
+retired rather than left on indefinitely.
+
+**Flag state is not in git.** `main` no longer tells you what production is
+showing. The registry lists what *can* be flagged; only the Unleash UI says
+what *is*. This is inherent to the choice of a service.
+
+**A large promotion backlog exists.** Production is running the
+pre-SEO-programme build — no place pages, and a sitemap still declaring the
+apex host. The first promotion after this work ships that entire backlog. The
+flag isolates the distance feature from it and nothing else.
+
+## Out of scope
+
+- Percentage rollouts, user targeting, A/B testing, and Unleash strategies
+  beyond simple on/off. Flags are booleans.
+- Pipeline and dbt flags. Airflow and dbt are not flag consumers.
+- Client-side flag evaluation. Flags are server-side only.
+- Automatic flag removal. The staleness test reports; a person deletes.

e2e/tests/journeys.spec.ts

@@ -1226,6 +1226,27 @@ const CUTOFF_CANDIDATE_URNS = [
   101099, 100553, 102574, 100769,                 // mixed
 ];
 
+/**
+ * Whether the last-distance-offered feature is switched on here.
+ *
+ * Read from the data rather than from /api/flags, which the public proxy
+ * denies on purpose — the endpoint names unreleased features. The observable
+ * effect is the field's presence: the flag is off iff no candidate school
+ * carries an `admission_distance` key at all.
+ *
+ * The distinction that matters: `admission_distance: null` means this school
+ * has no published cut-off, and the key being ABSENT means cut-offs are not
+ * being published at all.
+ */
+async function distanceFeatureIsOn(page: Page): Promise<boolean> {
+  for (const urn of CUTOFF_CANDIDATE_URNS) {
+    const res = await page.request.get(`/api/schools/${urn}`);
+    if (!res.ok()) continue;
+    if ('admission_distance' in (await res.json())) return true;
+  }
+  return false;
+}
+
 async function schoolWithCutoff(page: Page) {
   for (const urn of CUTOFF_CANDIDATE_URNS) {
     const res = await page.request.get(`/api/schools/${urn}`);
@@ -1237,6 +1258,59 @@ async function schoolWithCutoff(page: Page) {
   return null;
 }
 
+test('when the distance feature is on, a school with a cut-off is findable', async ({ page }) => {
+  /*
+   * The gate that stops the other distance journeys passing vacuously.
+   *
+   * They all skip when schoolWithCutoff() finds nothing, which is right when
+   * the feature is off — but it means a feature that is *supposed* to be on
+   * and is silently broken shows up as a green run full of skips. This test
+   * fails in that case.
+   */
+  test.skip(!(await distanceFeatureIsOn(page)),
+    'the admission_distance flag is off in this environment');
+
+  expect(await schoolWithCutoff(page),
+    'the distance feature is on, but no candidate school has a cut-off — '
+    + 'the flag is on and the data or the query behind it is broken')
+    .not.toBeNull();
+});
+
+test('with the distance feature off, the section is absent rather than empty', async ({ page }) => {
+  // Shipping dark means the page renders as it did before the feature existed,
+  // not as a feature with its content removed.
+  test.skip(await distanceFeatureIsOn(page),
+    'the admission_distance flag is on in this environment');
+
+  // A school that exists, found rather than hardcoded — a 404 page would
+  // satisfy the absent-heading assertion without proving anything.
+  //
+  // A plain loop, not Array.find: find's predicate is synchronous, so an async
+  // one returns a Promise, every Promise is truthy, and it would always hand
+  // back the first URN whether or not that school exists.
+  let urn: number | null = null;
+  for (const candidate of CUTOFF_CANDIDATE_URNS) {
+    if ((await page.request.get(`/api/schools/${candidate}`)).ok()) {
+      urn = candidate;
+      break;
+    }
+  }
+  expect(urn, 'no candidate school resolves in this environment').not.toBeNull();
+
+  await page.goto(`/school/${urn}`);
+  await expect(page.locator('h1')).toBeVisible();
+
+  await expect(page.getByRole('heading', { name: /How far away are you\?/ }))
+    .toHaveCount(0);
+});
+
+test('/api/flags is not reachable from the public internet', async ({ page }) => {
+  // It names every unreleased feature and whether it is on. Next reads it
+  // server-side over the Docker network; the public proxy must deny it.
+  const res = await page.request.get('/api/flags');
+  expect(res.status()).toBe(404);
+});
+
 test('a published cut-off distance is shown with the year it belongs to', async ({ page }) => {
   const found = await schoolWithCutoff(page);
   test.skip(found === null, 'no school in the sample has a published cut-off distance yet');

nextjs-app/__tests__/api/proxyDenylist.test.ts

@@ -0,0 +1,31 @@
+/**
+ * The /api/* proxy is public. Anything it forwards is on the internet.
+ *
+ * @jest-environment node
+ */
+// The docblock above is load-bearing. jest.config.js sets jsdom globally, and
+// NextRequest/NextResponse need the Web Fetch API globals that only the node
+// environment provides — under jsdom this suite fails on import, not on an
+// assertion.
+import { NextRequest } from 'next/server';
+import { GET } from '@/app/api/[...path]/route';
+
+function request(path: string) {
+  return new NextRequest(`http://localhost:3000/api/${path}`);
+}
+
+describe('public API proxy', () => {
+  it('refuses to forward internal-only paths', async () => {
+    // /api/flags names every unreleased feature and its state. Forwarding it
+    // publishes the thing shipping dark exists to keep quiet.
+    const res = await GET(request('flags'), { params: Promise.resolve({ path: ['flags'] }) });
+    expect(res.status).toBe(404);
+  });
+
+  it('does not deny a path that merely starts with the same letters', async () => {
+    // A prefix match would take /api/flagship down with /api/flags.
+    const res = await GET(
+      request('flagship'), { params: Promise.resolve({ path: ['flagship'] }) });
+    expect(res.status).not.toBe(404);
+  });
+});

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

@@ -0,0 +1,34 @@
+import { getFlags } from '@/lib/flags';
+
+// jsdom provides no global fetch, so there is nothing for jest.spyOn to attach
+// to — assign it and restore the original afterwards. This is the first test
+// here to mock fetch; later ones should follow this shape.
+const realFetch = global.fetch;
+
+function mockFetch(impl: () => Promise<unknown>) {
+  global.fetch = jest.fn(impl) as unknown as typeof fetch;
+}
+
+describe('getFlags', () => {
+  afterEach(() => { global.fetch = realFetch; });
+
+  it('returns the flags the API reports', async () => {
+    mockFetch(async () => ({
+      ok: true,
+      json: async () => ({ admission_distance: true }),
+    }));
+    await expect(getFlags()).resolves.toEqual({ admission_distance: true });
+  });
+
+  it('returns no flags rather than throwing when the API is down', async () => {
+    // A page that cannot read flags must render everything dark, not 500.
+    // Fail-closed is the same direction as the backend's default.
+    mockFetch(async () => { throw new Error('ECONNREFUSED'); });
+    await expect(getFlags()).resolves.toEqual({});
+  });
+
+  it('returns no flags rather than throwing on a non-200', async () => {
+    mockFetch(async () => ({ ok: false, status: 503 }));
+    await expect(getFlags()).resolves.toEqual({});
+  });
+});

nextjs-app/app/api/[...path]/route.ts

@@ -26,8 +26,27 @@ function backendBase(): string {
 const STRIPPED_RESPONSE_HEADERS = ['content-encoding', 'content-length', 'transfer-encoding', 'connection'];
 const METHODS_WITH_BODY = new Set(['POST', 'PUT', 'PATCH', 'DELETE']);
 
+/*
+ * API paths this public proxy must not forward.
+ *
+ * Matched on the first segment, exactly — a prefix match would take
+ * /api/flagship down with /api/flags.
+ *
+ * `flags` is here because GET /api/flags names every unreleased feature the
+ * codebase knows about, along with whether it is on. Publishing that defeats
+ * the point of shipping dark. Next reads it server-side via FASTAPI_URL, on
+ * the Docker network, which never transits this route.
+ *
+ * Anything else internal-only belongs here too.
+ */
+const INTERNAL_ONLY_SEGMENTS = new Set(['flags']);
+
 async function handler(req: NextRequest, ctx: { params: Promise<{ path: string[] }> }) {
   const { path } = await ctx.params;
+  if (INTERNAL_ONLY_SEGMENTS.has(path[0])) {
+    return NextResponse.json({ detail: 'Not Found' }, { status: 404 });
+  }
+
   const target = `${backendBase()}/${path.join('/')}${req.nextUrl.search}`;
 
   const headers = new Headers(req.headers);

nextjs-app/jest.setup.js

@@ -12,6 +12,12 @@ jest.mock('next/navigation', () => ({
   useSearchParams: () => new URLSearchParams(),
 }));
 
+// Everything below this line is browser furniture, and this file runs for
+// every suite — including the ones that declare `@jest-environment node` to
+// test route handlers, where NextRequest needs Fetch API globals jsdom does
+// not provide. There is no `window` there, so guard rather than assume one.
+if (typeof window !== 'undefined') {
+
 // Mock window.matchMedia
 Object.defineProperty(window, 'matchMedia', {
   writable: true,
@@ -52,3 +58,5 @@ const localStorageMock = {
   clear: jest.fn(),
 };
 global.localStorage = localStorageMock;
+
+} // end: browser-only globals

nextjs-app/lib/flags.ts

@@ -0,0 +1,40 @@
+/**
+ * Reading feature flags.
+ *
+ * Server-side only. No flag value reaches the browser bundle, and there is no
+ * Unleash dependency in package.json — the SDK lives in FastAPI, which already
+ * owns every other piece of data this app renders.
+ *
+ * Flags are declared in backend/flags.py. A purely front-end flag still has to
+ * be declared there; it is a flat data edit, and the return is that one list
+ * answers "what flags exist" for the whole system.
+ */
+
+export type Flags = Record<string, boolean>;
+
+/*
+ * Reading flags pins the calling route to this ISR floor: Next uses the LOWEST
+ * revalidate among a route's fetches to set the whole route's revalidation
+ * frequency. 300s matches what /school/[slug] already sits at, so a page that
+ * reads flags is no more dynamic than a school page already is.
+ *
+ * It is also what makes a flip propagate without a webhook: five minutes on
+ * school pages, an hour on place pages, against flags that flip monthly.
+ */
+export const FLAGS_REVALIDATE = 300;
+
+const API = process.env.FASTAPI_URL || process.env.NEXT_PUBLIC_API_URL
+  || 'http://localhost:8000/api';
+
+/** Every flag and its value. Never throws: an unreadable flag is a dark one. */
+export async function getFlags(): Promise<Flags> {
+  try {
+    const res = await fetch(`${API}/flags`, {
+      next: { revalidate: FLAGS_REVALIDATE },
+    });
+    if (!res.ok) return {};
+    return await res.json();
+  } catch {
+    return {};
+  }
+}

nextjs-app/lib/types.ts

@@ -361,7 +361,12 @@ export interface SchoolDetailsResponse {
    * held back as a paid feature and are not part of this public payload — see
    * data_loader._admission_distance.
    */
-  admission_distance: SchoolAdmissionDistance | null;
+  /**
+   * Absent — not null — when the admission_distance flag is off. Null means
+   * "this school has no published cut-off"; absent means "cut-offs are not
+   * being published at all". They are different claims and the type says so.
+   */
+  admission_distance?: SchoolAdmissionDistance | null;
   deprivation: SchoolDeprivation | null;
   finance: SchoolFinance | null;
 }

requirements.txt

@@ -12,4 +12,4 @@ slowapi==0.1.9
 secure==0.3.0
 typesense==0.21.0
 numpy==1.26.4
-
+UnleashClient==6.0.1