fix(schema): cascade-delete VideoConfig/LintResult/BulkJobItem/VideoPlaylist with their Video

Channel import crashed on the hard-delete purge path (videos confirmed
gone from YouTube for 30+ days): prisma.video.deleteMany() hit a FK
violation on VideoConfig_videoId_fkey. All four required relations
pointing at Video had no onDelete behavior set, defaulting to RESTRICT
- any of the four would have hit the same crash, not just VideoConfig.
First real channel import against a real, aged dataset is what
surfaced this; nothing had exercised the 30-day hard-delete path before.

.gitignore

@@ -3,6 +3,7 @@ node_modules/
 
 # Build output
 backend/dist/
+backend/tsconfig.tsbuildinfo
 frontend/.next/
 frontend/tsconfig.tsbuildinfo
 

backend/.dockerignore

@@ -1,5 +1,6 @@
 node_modules
 dist
+tsconfig.tsbuildinfo
 .env
 .env.development
 .env.production

backend/prisma/migrations/20260811151354_cascade_delete_video_relations/migration.sql

@@ -0,0 +1,23 @@
+-- DropForeignKey
+ALTER TABLE "BulkJobItem" DROP CONSTRAINT "BulkJobItem_videoId_fkey";
+
+-- DropForeignKey
+ALTER TABLE "LintResult" DROP CONSTRAINT "LintResult_videoId_fkey";
+
+-- DropForeignKey
+ALTER TABLE "VideoConfig" DROP CONSTRAINT "VideoConfig_videoId_fkey";
+
+-- DropForeignKey
+ALTER TABLE "VideoPlaylist" DROP CONSTRAINT "VideoPlaylist_videoId_fkey";
+
+-- AddForeignKey
+ALTER TABLE "VideoConfig" ADD CONSTRAINT "VideoConfig_videoId_fkey" FOREIGN KEY ("videoId") REFERENCES "Video"("id") ON DELETE CASCADE ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE "LintResult" ADD CONSTRAINT "LintResult_videoId_fkey" FOREIGN KEY ("videoId") REFERENCES "Video"("id") ON DELETE CASCADE ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE "BulkJobItem" ADD CONSTRAINT "BulkJobItem_videoId_fkey" FOREIGN KEY ("videoId") REFERENCES "Video"("id") ON DELETE CASCADE ON UPDATE CASCADE;
+
+-- AddForeignKey
+ALTER TABLE "VideoPlaylist" ADD CONSTRAINT "VideoPlaylist_videoId_fkey" FOREIGN KEY ("videoId") REFERENCES "Video"("id") ON DELETE CASCADE ON UPDATE CASCADE;

backend/prisma/schema.prisma

@@ -174,7 +174,7 @@ enum LintStatus {
 model VideoConfig {
   id              String    @id @default(cuid())
   videoId         String    @unique
-  video           Video     @relation(fields: [videoId], references: [id])
+  video           Video     @relation(fields: [videoId], references: [id], onDelete: Cascade)
   templateId      String?
   blockOrder      Json
   blockOverrides  Json
@@ -312,7 +312,7 @@ model SavedView {
 model LintResult {
   id            String       @id @default(cuid())
   videoId       String
-  video         Video        @relation(fields: [videoId], references: [id])
+  video         Video        @relation(fields: [videoId], references: [id], onDelete: Cascade)
   ruleCode      String
   severity      LintSeverity
   targetField   String?
@@ -369,7 +369,7 @@ model BulkJobItem {
   bulkJobId      String
   bulkJob        BulkJob  @relation(fields: [bulkJobId], references: [id])
   videoId        String
-  video          Video    @relation(fields: [videoId], references: [id])
+  video          Video    @relation(fields: [videoId], references: [id], onDelete: Cascade)
   beforeSnapshot Json?
   afterSnapshot  Json?
   status         String   @default("pending")
@@ -415,7 +415,7 @@ model VideoPlaylist {
   videoId    String
   playlistId String
   position   Int?
-  video      Video    @relation(fields: [videoId], references: [id])
+  video      Video    @relation(fields: [videoId], references: [id], onDelete: Cascade)
   playlist   Playlist @relation(fields: [playlistId], references: [id])
 
   @@id([videoId, playlistId])

backend/src/modules/auth/auth.service.ts

@@ -67,6 +67,7 @@ export class AuthService {
               youtubeChannelId,
               name: channelName,
               uploadsPlaylistId,
+              supplementalVideoIds: [],
               youtubeAccessToken: this.encrypt(accessToken),
               youtubeRefreshToken: refreshToken ? this.encrypt(refreshToken) : undefined,
               youtubeTokenExpiry: new Date(Date.now() + 3600 * 1000),

backend/tsconfig.json

@@ -9,6 +9,7 @@
     "target": "ES2021",
     "sourceMap": true,
     "outDir": "./dist",
+    "rootDir": "./src",
     "baseUrl": "./",
     "incremental": true,
     "skipLibCheck": true,
@@ -17,5 +18,6 @@
     "strictBindCallApply": false,
     "forceConsistentCasingInFileNames": false,
     "noFallthroughCasesInSwitch": false
-  }
+  },
+  "exclude": ["node_modules", "dist", "prisma"]
 }

documentation/YouTube Studio Flow/05 - Development/05 - Deployment and Operations.md

@@ -17,7 +17,7 @@ The production setup is a single-host Docker Compose deployment using **Traefik*
 | `postgres` | postgres:16-alpine | Database — bound to `127.0.0.1:5432` (not public) |
 | `redis` | redis:7-alpine | BullMQ queues — `noeviction` policy, AOF persistence |
 
-**SSL / routing**: Traefik handles TLS termination with automatic Let's Encrypt certificates. Both API (`/api` prefix) and frontend run on the same domain — Traefik routes by path prefix. The `traefik-network` external network must exist before deploy.
+**SSL / routing**: Traefik handles TLS termination with automatic Let's Encrypt certificates. Both API (`/api` prefix) and frontend run on the same domain — Traefik routes by path prefix. The `proxy` external network (the host's existing Traefik network) must exist before deploy.
 
 **Build**: Multi-stage Dockerfile (`node:22-alpine`). Builder compiles TypeScript; runner installs prod-only deps. The Prisma CLI is copied from builder stage so the `migrate` service can run schema migrations.
 

infrastructure/README.md

@@ -24,8 +24,8 @@ Registry (`git.devils.zone/devil/youtube-studio-flow-{backend,frontend}`), gebau
 6. Container starten: `docker compose up -d`
 
 Für ein Update auf eine neue Version: `IMAGE_TAG` in `.env` anpassen (oder `latest`
-belassen), dann Schritte 4–6 wiederholen. Das externe Docker-Netzwerk `traefik-network`
+belassen), dann Schritte 4–6 wiederholen. Das externe Docker-Netzwerk `proxy`
-muss vorher existieren (`docker network create traefik-network`), falls Traefik das nicht
+muss vorher existieren (`docker network create proxy`), falls Traefik das nicht
 bereits selbst anlegt.
 
 ## Daten-Persistenz

infrastructure/docker-compose.yml

@@ -3,6 +3,7 @@ services:
   # Runs database migrations once before the API and worker start.
   migrate:
     image: git.devils.zone/devil/youtube-studio-flow-backend:${IMAGE_TAG:-latest}
+    pull_policy: always
     command: npx prisma migrate deploy
     restart: "no"
     environment:
@@ -15,6 +16,7 @@ services:
 
   api:
     image: git.devils.zone/devil/youtube-studio-flow-backend:${IMAGE_TAG:-latest}
+    pull_policy: always
     restart: unless-stopped
     environment:
       NODE_ENV: production
@@ -44,16 +46,17 @@ services:
         condition: service_completed_successfully
     healthcheck:
       test: ["CMD-SHELL", "wget -qO- http://localhost:3001/api/v1/health || exit 1"]
-      interval: 15s
+      interval: 10s
       timeout: 5s
-      retries: 3
+      retries: 10
-      start_period: 30s
+      start_period: 60s
     networks:
       - app-network
-      - traefik-network
+      - proxy
 
   worker:
     image: git.devils.zone/devil/youtube-studio-flow-backend:${IMAGE_TAG:-latest}
+    pull_policy: always
     restart: unless-stopped
     command: node dist/worker.js
     environment:
@@ -82,6 +85,7 @@ services:
     # Dockerfile default, /api/v1) — the relative path works because Traefik serves both
     # frontend and API on the same domain.
     image: git.devils.zone/devil/youtube-studio-flow-frontend:${IMAGE_TAG:-latest}
+    pull_policy: always
     restart: unless-stopped
     labels:
       - "traefik.enable=true"
@@ -91,11 +95,14 @@ services:
       - "traefik.http.routers.studioflow-frontend.priority=1"
       - "traefik.http.services.studioflow-frontend.loadbalancer.server.port=3000"
     depends_on:
+      # service_started, not service_healthy: a slow/unhealthy api shouldn't block the
+      # whole `compose up` (and Portainer tearing down what it created) - the frontend
+      # itself doesn't need api to be ready at container-start time.
       api:
-        condition: service_healthy
+        condition: service_started
     networks:
       - app-network
-      - traefik-network
+      - proxy
 
   postgres:
     image: postgres:16-alpine
@@ -142,5 +149,5 @@ volumes:
 networks:
   app-network:
     driver: bridge
-  traefik-network:
+  proxy:
     external: true

scripts/migrate-local-db-to-server.sh

@@ -0,0 +1,90 @@
+#!/usr/bin/env bash
+# One-time migration: dump the local dev Postgres and restore it into the
+# server's Postgres over SSH, replacing whatever is there.
+#
+# DESTRUCTIVE on the target: drops and recreates the remote database.
+#
+# Fill in the placeholders below (or export them as env vars before running),
+# then run from the project root: bash scripts/migrate-local-db-to-server.sh
+#
+# Prerequisites:
+#   - Local Postgres running: cd infrastructure && docker compose up -d postgres redis
+#   - SSH access to the server
+#   - The remote postgres container's name (find it with `docker ps` on the
+#     server, or check the container list in Portainer for your stack - it'll
+#     be something like <stack-name>-postgres-1)
+
+set -euo pipefail
+
+# Git Bash/MSYS auto-translates POSIX-looking path args (e.g. /tmp/foo) into
+# Windows paths before handing them to docker.exe - but these paths are meant
+# to be interpreted inside the Linux container, not on the Windows host.
+export MSYS_NO_PATHCONV=1
+
+# ── Fill these in ──────────────────────────────────────────────────────────
+SSH_USER="${SSH_USER:-dummyuser}"
+SSH_HOST="${SSH_HOST:-dummy.server.example}"
+SSH_PORT="${SSH_PORT:-22}"
+REMOTE_POSTGRES_CONTAINER="${REMOTE_POSTGRES_CONTAINER:-studioflow-postgres-1}"
+REMOTE_POSTGRES_USER="${REMOTE_POSTGRES_USER:-studioflow}"
+REMOTE_POSTGRES_DB="${REMOTE_POSTGRES_DB:-studioflow}"
+# ────────────────────────────────────────────────────────────────────────────
+
+LOCAL_DB_USER="studioflow"
+LOCAL_DB_NAME="studioflow"
+DUMP_FILE="studioflow_migration_$(date +%Y%m%d_%H%M%S).dump"
+
+echo "==> Finding local Postgres container"
+LOCAL_PG_CONTAINER=$(docker ps --format '{{.Names}}' | grep -i postgres | head -1)
+if [ -z "$LOCAL_PG_CONTAINER" ]; then
+  echo "No running local postgres container found. Start it first:"
+  echo "  cd infrastructure && docker compose up -d postgres redis"
+  exit 1
+fi
+echo "    Using: $LOCAL_PG_CONTAINER"
+
+echo "==> Dumping local database ($LOCAL_DB_NAME)"
+docker exec "$LOCAL_PG_CONTAINER" pg_dump -U "$LOCAL_DB_USER" -d "$LOCAL_DB_NAME" -F c -f "/tmp/$DUMP_FILE"
+docker cp "${LOCAL_PG_CONTAINER}:/tmp/$DUMP_FILE" "./$DUMP_FILE"
+docker exec "$LOCAL_PG_CONTAINER" rm "/tmp/$DUMP_FILE"
+echo "    Dump size: $(du -h "./$DUMP_FILE" | cut -f1)"
+
+echo ""
+echo "About to overwrite the database on ${SSH_HOST} (container: ${REMOTE_POSTGRES_CONTAINER})."
+read -p "Type YES to continue: " CONFIRM
+if [ "$CONFIRM" != "YES" ]; then
+  echo "Aborted. Local dump kept at ./$DUMP_FILE"
+  exit 1
+fi
+
+echo "==> Copying dump to server"
+scp -P "$SSH_PORT" "./$DUMP_FILE" "${SSH_USER}@${SSH_HOST}:/tmp/$DUMP_FILE"
+
+echo "==> Restoring on server"
+ssh -p "$SSH_PORT" "${SSH_USER}@${SSH_HOST}" bash -s <<EOF
+set -euo pipefail
+echo "  Copying dump into container..."
+docker cp "/tmp/$DUMP_FILE" "${REMOTE_POSTGRES_CONTAINER}:/tmp/$DUMP_FILE"
+
+echo "  Terminating existing connections..."
+docker exec "$REMOTE_POSTGRES_CONTAINER" psql -U "$REMOTE_POSTGRES_USER" -d postgres -c \
+  "SELECT pg_terminate_backend(pid) FROM pg_stat_activity WHERE datname = '$REMOTE_POSTGRES_DB';" >/dev/null
+
+echo "  Dropping and recreating database..."
+docker exec "$REMOTE_POSTGRES_CONTAINER" psql -U "$REMOTE_POSTGRES_USER" -d postgres -c \
+  "DROP DATABASE IF EXISTS $REMOTE_POSTGRES_DB;" >/dev/null
+docker exec "$REMOTE_POSTGRES_CONTAINER" psql -U "$REMOTE_POSTGRES_USER" -d postgres -c \
+  "CREATE DATABASE $REMOTE_POSTGRES_DB OWNER $REMOTE_POSTGRES_USER;" >/dev/null
+
+echo "  Restoring data..."
+docker exec "$REMOTE_POSTGRES_CONTAINER" pg_restore -U "$REMOTE_POSTGRES_USER" -d "$REMOTE_POSTGRES_DB" --no-owner --no-privileges "/tmp/$DUMP_FILE"
+
+docker exec "$REMOTE_POSTGRES_CONTAINER" rm "/tmp/$DUMP_FILE"
+rm "/tmp/$DUMP_FILE"
+echo "  Done."
+EOF
+
+echo ""
+echo "==> Migration complete."
+echo "    Restart api/worker in Portainer (or they'll self-heal on next DB query)."
+echo "    Local dump kept at ./$DUMP_FILE - delete it once you've confirmed the server looks right."