Docker Compose refused to adopt the existing portfolio_net network because
it lacked the expected com.docker.compose.network label (created outside
Compose). Mark it as external (matching the dev setup) and pre-create it
in the deployment workflow to ensure it always exists before compose up.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Fix ActivityFeed: Remove dynamic import that was causing it to disappear in production
* Fix ActivityFeed hydration error: Move localStorage read to useEffect to prevent server/client mismatch
* Update Node.js version to 25 in Gitea workflows
- Fix EBADENGINE error for camera-controls@3.1.2 which requires Node.js >=22
- Update production-deploy.yml, dev-deploy.yml, and ci-cd-with-gitea-vars.yml.disabled
- Node.js v25 matches local development environment
* Update Dockerfile to use Node.js 25
- Update base image from node:20 to node:25
- Matches Gitea workflow configuration and camera-controls@3.1.2 requirements
* Fix production deployment: Start database dependencies
- Remove --no-deps flag which prevented postgres and redis from starting
- Remove --build flag as image is already built in previous step
- This fixes 'Can't reach database server at postgres:5432' error
* Fix postgres health check in production
- Remove init-db.sql volume mount (not available in CI/CD environment)
- Init script not needed as Prisma handles schema migrations
- Postgres will initialize empty database automatically
* Fix cache permission error in Docker container
- Create cache directories AFTER copying standalone files
- Create both fetch-cache and images subdirectories
- Set proper ownership for nextjs user
- Fixes EACCES permission denied errors for prerender cache
* Fix German jogging fallback text
* Use Directus content in production
* fix: Security vulnerability - block malicious file requests
* fix: Switch projects to Directus, add security fixes and example projects
- Use Docker health check status instead of host-based curl
- Test from inside container instead of from host
- Better error messages and debugging
- More reliable health check that doesn't depend on port mapping
- Use docker compose ps to get correct container ID (avoids staging container)
- Verify container is from production compose file before health check
- Accept deployment if Docker health check reports healthy (even if HTTP test fails)
- Better error messages and debugging output
- Fix container ID selection to avoid matching staging containers
- Add N8N_WEBHOOK_URL, N8N_SECRET_TOKEN, N8N_API_KEY to docker-compose.production.yml
- Export environment variables in workflow before docker-compose up
- Improve error logging in chat API for better debugging
- Add better error handling in ChatWidget component
- Create setup guide for n8n chat configuration
- Prioritize Docker health status (most reliable)
- Test HTTP endpoint from inside container using docker exec
- Fallback to host-based HTTP test if available
- Better debugging output showing both internal and external tests
- Final verification uses Docker health status as authoritative
- Install curl in runner stage for health checks
- Increase health check timeout to 90 attempts (3 minutes)
- Improve health check logic to prioritize HTTP endpoint
- Add better debugging output during health check wait
- Show container status and logs during health check failures
- Remove branch-specific variable names (not needed)
- Each workflow uses its own default based on branch
- Users only need to set general variables, not branch-specific ones
- Redesign staging banner as floating box in bottom-right corner
- Better UX: doesn't block content, dismissible, modern design
- Support NEXT_PUBLIC_BASE_URL_PRODUCTION and NEXT_PUBLIC_BASE_URL_DEV
- Support LOG_LEVEL_PRODUCTION and LOG_LEVEL_DEV
- Fallback to general variables if branch-specific not set
- Add comprehensive GITEA_VARIABLES_SETUP.md guide
- Allows independent configuration for production and dev branches
- Created separate workflows for production and dev deployments
- Production branch → dk0.dev (port 3000)
- Dev branch → dev.dk0.dev (port 3002)
- Zero-downtime deployment pattern (start new, wait for health, remove old)
- Complete isolation between environments (separate containers, databases, networks)
- Cleaned up unused code and files:
- Removed unused GhostEditor and ResizableGhostEditor components
- Removed old/unused workflows and markdown files
- Fixed docker-compose references
- Upgraded dependencies to latest compatible versions
- Fixed TypeScript errors in editor page
- Updated staging to use dev.dk0.dev domain
