aec2027

Campaign landing page for "L'Avenir en Commun" (Jean-Luc Mélenchon, presidential election 2027 / La France Insoumise). Canonical domain: aec2027.fr.

Repo layout

The repo is an npm-workspaces + Turborepo monorepo (apps/*):

• apps/web/ — the Next.js app (App Router, TypeScript, Vitest). This is the site.
• apps/backend/ — an empty placeholder for a possible future in-repo backend.

See CLAUDE.md for the full design-system conventions and docs/plans/ for in-progress planning artifacts.

Running locally without Docker

npm install          # installs all workspaces
npm run dev           # turbo run dev -> apps/web's `next dev`, http://localhost:3000

A Makefile wraps the common tasks — run make to list the targets.

Running with Docker

docker compose up app                         # http://localhost:3000 (+ the audio origin)
docker compose --profile test run --rm test   # apps/web's typecheck + test suite

Without compose:

docker build .                     # production image (runner target)
docker build --target test .       # test image

There are no build args: every page is force-dynamic, so nothing about the audio origin or the open-dataec API is baked into the image. Configure both through environment: at run time (see below) — changing the audio host doesn't need a rebuild.

docker compose up app also starts audio, a dev-only nginx:alpine origin on http://localhost:8081 that serves apps/web/public/audio/ over HTTP. It exists so the app can be exercised against real, multi-megabyte narration without committing it to git or rebuilding to change it — the directory is bind-mounted read-only, its contents are gitignored, and .dockerignore keeps them out of the image. Production points AUDIO_CDN_BASE_URL at object storage instead. See apps/web/public/audio/README.md.

Testing

npm run typecheck    # via Turborepo
npm run test         # via Turborepo — same commands the Docker `test` target runs

Environment variables

All three are runtime values (docker-compose.yml's environment:), never build args. apps/web/.env.example documents them for non-Docker runs.

• OPEN_DATAEC_API_URL (optional) — base URL of the open-dataec API. Fetched server-side only, so a Docker-internal hostname like http://open-dataec:3000 is fine. Falls back to http://localhost:3000. See apps/web/lib/dataec.ts.
• AUDIO_CDN_BASE_URL (optional) — base URL of the narration audio origin. Unlike the above, this one is emitted into the HTML as an <audio> src and fetched by the browser, so it must be reachable from the visitor's browser — never http://audio:80. Unset means Next.js serves apps/web/public/audio/<slug>.mp3 itself. Compose defaults it to the dev audio service, http://localhost:8081/audio. See apps/web/lib/audio.ts.
• AUDIO_AVAILABLE_SLUGS (optional) — comma-separated chapter slugs that actually have a recording. Empty by default: no narration exists yet (2026-07-26), so no listen button renders anywhere rather than offering one that 404s. Malformed entries are ignored. See apps/web/lib/audioAvailability.ts.

Whatever origin serves the audio must honour HTTP Range requests — the player's slider and ±15 s buttons seek mid-file. Next.js's public/ handler, nginx, and object storage all do.