fix(deploy): run migrations before server boot to prevent stale schema errors

Sidekiq was starting before migrations ran, causing RuntimeError on the
group_type enum. Moved db:chatwoot_prepare from post_start into the rails
entrypoint and made sidekiq depend on the rails healthcheck.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

docker-compose.coolify.yaml

@@ -46,7 +46,7 @@ services:
       - command:
         - sh
         - '-c'
-        - 'bundle exec rails db:chatwoot_prepare && bundle exec rails branding:update && if [ -n "$${BRAND_ASSETS_URL}" ]; then deployment/extract_brand_assets.sh "$${BRAND_ASSETS_URL}"; fi'
+        - 'bundle exec rails branding:update && if [ -n "$${BRAND_ASSETS_URL}" ]; then deployment/extract_brand_assets.sh "$${BRAND_ASSETS_URL}"; fi'
     healthcheck:
       test:
         - CMD-SHELL
@@ -61,8 +61,12 @@ services:
     volumes:
       - 'storage:/app/storage'
     depends_on:
-      - postgres
-      - redis
+      rails:
+        condition: service_healthy
+      postgres:
+        condition: service_started
+      redis:
+        condition: service_started
     environment:
       - NODE_ENV=production
       - RAILS_ENV=production

docker/entrypoints/rails.sh

@@ -30,5 +30,11 @@ do
   sleep 2;
 done
 
+# Run pending migrations before starting the server so that workers
+# (sidekiq) that depend on the rails healthcheck never see a stale schema.
+echo "Running db:chatwoot_prepare..."
+bundle exec rake db:chatwoot_prepare
+echo "Database preparation complete."
+
 # Execute the main process of the container
 exec "$@"