MindReply

MindReply now includes a production-ready Campaign Concept Studio alongside the existing product surfaces. The studio lets marketing teams enter a short brief, target audience, product details, tone, and channels, then generates a concise campaign concept, five copy variants, a launch checklist, image prompts, and generated images.

Campaign Studio

Route: /campaign-studio

The studio uses server-side OpenAI calls only. The client sends campaign inputs to a Next.js API route, and the server uses the OpenAI Responses API for structured text generation and the image API for creative direction assets.

What is included

• Next.js frontend with app routes and health checks
• Express backend with /health, admin chat, connectors, and analytics endpoints
• Campaign concept studio UI with loading, error, and empty states
• Structured campaign generation with copy variants, checklist, and image prompts
• Image generation for campaign direction visuals
• Stripe, Anthropic, Google, Clerk, Sentry, and Neon/Postgres integration points
• Docker, Docker Compose, and Kubernetes deployment assets
• GitHub Actions automation for CI, security, and deployment

Environment variables

Use .env.example as the source of truth. At minimum, set:

• OPENAI_API_KEY
• OPENAI_TEXT_MODEL (default: gpt-5.5)
• OPENAI_IMAGE_MODEL (default: gpt-image-2)
• existing app runtime variables for auth, database, analytics, and deployment

Install

corepack enable
pnpm install

Run

pnpm dev

Open the campaign studio at:

/campaign-studio

Validate

bash automation/scripts/validate.sh
bash automation/scripts/security.sh

Deploy

The repository uses main as the canonical production branch. Production validation runs through GitHub Actions, and the deploy step is intentionally separated from client-side code.

Tuning points

Adjust these later in app/api/campaign-studio/generate/route.ts:

• text model selection
• prompt instructions
• structured output schema
• image model selection
• image size and count

Validation plan

• Empty input is rejected before the API call
• Structured JSON is validated on the server
• Image generation failures return partial results instead of breaking the whole response
• The client displays loading, error, and empty states

Deployment notes

• Keep OPENAI_API_KEY server-side only
• The client never calls OpenAI directly
• Use the campaign studio route as the only backend boundary for generation
• Confirm /api/health and /health before release

Production layout

This repo is best treated as a single source of truth with these lanes:

• main for production-ready code
• feature branches for changes
• GitHub Actions for validation and deployment
• Vercel for the Next.js frontend
• container/Kubernetes assets for environments that need them

Notes

The repository already contains multiple deployment scripts and migration notes. The goal of this cleanup is to keep the canonical path simple: documented setup, validated CI, and one clear production flow.