m-master
Context-aware marketing platform.
Core Rules
Deployment workflow
production branch: all routine development and deployment happen here
GitHub Actions deployment
- pushes to
production deploy to the Vercel production target and rebind m-master.vercel.app
- Vercel Git auto-deploy should stay disabled to avoid duplicate deployments
Required repository secret
VERCEL_TOKEN: token that can deploy and manage aliases for the watchers0930s-projects/m-master Vercel project
Vercel strategy
- GitHub Actions is the deployment entry point
- local CLI direct deploys should be avoided except for emergency maintenance
m-master.vercel.app is the fixed production domain
Promotion checklist
- Work only on
production
- Run local verification: at minimum
npm run build
- Push
production
- Wait for the GitHub Actions production deploy
- Verify
m-master.vercel.app
If production promotion is not clean
- Do not force-push
production
- If a direct
production push is blocked by branch protection or branch divergence, create a temporary alignment branch from the latest origin/production
- Apply the verified content onto that branch
- Open a clean PR from the alignment branch into
production
- If the production deploy workflow does not auto-start after merge, run
deploy.yml with workflow_dispatch on production
Database
- Separate Neon database for
m-master
- Prisma uses:
DATABASE_URL for application queriesDIRECT_URL for schema operations
Image generation
OPENAI_API_KEY: enables GPT image generation for the image studioOPENAI_TEXT_MODEL: optional override for GPT draft generation, defaults to gpt-5.4-miniOPENAI_IMAGE_MODEL: optional override, defaults to gpt-image-1-mini- cost-optimized default:
low quality and 1 variant per request
- if the API key is missing or the OpenAI request fails, the app falls back to the built-in SVG generator
- channel draft generation also uses the same
OPENAI_API_KEY; if the request fails, it falls back to the built-in rule-based draft generator
