Woman's Fight AI

AI-powered social media content automation for entrepreneurs, women-led businesses, online and local shops, coaches and creators.

Business understanding → Business Brain → content strategy → 30-day plan → generation → platform adaptation → your approval → scheduling → publishing → analytics.

---

Run it locally

npm install
cp .env.example .env.local      # optional — the app runs without any keys
npm run dev

Then open http://localhost:3000.

Create an account at /signup. The first account in a fresh install becomes the admin, so /admin is reachable straight away.

Useful scripts

Command	What it does
npm run dev	Development server on port 3000
npm run build	Production build
npm run start	Serve the production build
npm run lint	ESLint
npm run typecheck	tsc --noEmit

---

Development mode

The app is fully explorable with no external services configured, and it says so rather than pretending:

Missing	What happens
Supabase	Auth and data use a local development store (.dev-data/db.json, salted password hashes). A banner says so.
CLAUDE_API_KEY	A clearly labelled local placeholder assistant answers instead of Claude. Every reply is marked as development output.
Image / video provider	Media generation is shown as a future feature and refuses rather than returning a fake asset.
Social platform keys	Every platform shows Coming soon. No connection is ever faked.
Stripe	Plans and limits are live and enforced; checkout says payments are not connected.

Adding the real keys switches the same code paths over — no component changes.

---

Connecting the real services

Supabase

1. Create a project at supabase.com.
2. Run supabase/schema.sql in the SQL editor. It creates every table, index, enum, trigger and row level security policy, plus the auth.users → profile + workspace trigger.
3. Put the project URL, anon key and service role key into .env.local.

Workspace isolation is enforced by Postgres RLS, not by application code.

Claude

Set CLAUDE_API_KEY (and optionally CLAUDE_MODEL). All AI calls go through lib/ai/ai-service.ts, which validates every structured response against a Zod schema before anything is saved.

The publishing worker

Scheduling does not depend on a browser staying open. Point a cron job at:

POST /api/scheduler/run
Authorization: Bearer $SCHEDULER_WORKER_SECRET

every few minutes. Due jobs retry with backoff, and a post is marked published only when a platform API returns a post id.

---

Architecture

app/
├── (marketing)/   landing, features, pricing, faq, about, contact
├── (auth)/        login, signup, forgot/reset password, verify email
├── onboarding/    conversational Business Brain setup
├── dashboard/     dashboard, create, planner, calendar, library, social, analytics, settings
├── admin/         overview, users, subscriptions, content, usage, publishing, logs
└── api/           ai/*, scheduler/run, social/callback/*

components/   ui, marketing, auth, dashboard, ai, content, calendar, social, admin
lib/
├── ai/          AIService, BusinessBrainService, ContentStrategyService,
│                ContentGenerationService, PlatformAdaptationService, language engine
├── content/     approval, editing, scheduling rules
├── data/        DataStore interface + Supabase and development implementations
├── media/       ImageProvider / VideoProvider interfaces
├── social/      SocialProvider interface + one class per platform
├── scheduler/   SchedulerService, PublishingService
├── billing/ analytics/ notifications/ usage/ auth/ supabase/ config/ utils/
types/        the domain model, mirroring the SQL schema
supabase/     schema.sql

Principles the code actually follows

• Approval before automation. Scheduling refuses unapproved content in the service layer, not just in the UI. Editing approved content sends it back for review.
• Never claim what has not happened. No "published" without an API confirmation, no "connected" without a real connection, no invented analytics.
• One Business Brain. The user explains their business once; every generation reads from it.
• Language follows the user. Bangla, Banglish and English are detected per message — there is no language selector in the chat.
• Data belongs to a workspace. user → workspace → business → content, so multiple brands, teams and agencies fit without a migration.
• Secrets stay server-side. API keys are never imported into client components, and logs are scrubbed of anything credential-shaped before being written.

---

Roadmap

Phase 2 — image generation, video generation, Facebook / Instagram / YouTube / TikTok / X / LinkedIn publishing, real scheduler workers, subscription billing.

Phase 3 — advanced analytics, AI performance insights, multiple businesses, team members, agency workspaces, brand kits, content repurposing, best-time recommendations.

---

AI does the work. You approve.