# Platform Digital — Complete Technical Reference

> Ship your agentic platform in weeks, not years. 150k+ lines of production-ready TypeScript for building AI-powered development platforms like Lovable, Replit, Bolt, and v0.

## Overview

Platform Digital is a production-ready boilerplate for building agentic development platforms. It provides the complete infrastructure needed to launch a platform where users describe what they want, and AI builds it — with live preview, real-time collaboration, and one-click deployment. Building this from scratch takes 12-18 months and $300k+ in engineering. Platform Digital delivers it as a one-time purchase with full source code ownership.

## Codebase Statistics

| Metric | Count |
|--------|-------|
| Total lines of TypeScript | 150,000+ |
| API endpoints | 90 |
| React components | 151 |
| Database models | 24 |
| Zustand state stores | 18 |
| Custom hooks | 34 |
| AI modes | 5 |
| Test files | 20 |
| Cron jobs | 3 |

## Tech Stack

| Category | Technology |
|----------|-----------|
| Frontend Framework | Next.js (App Router), React 19, TypeScript |
| Database | PostgreSQL (Neon recommended with Prisma Accelerate) |
| ORM | Prisma with adapters (Neon, PostgreSQL) |
| Authentication | Clerk (OAuth, SSO, 2FA) |
| AI Providers | Claude Agent SDK, OpenAI, Cerebras |
| Real-time Sync | Liveblocks (WebSocket-based) |
| Sandbox | WebContainer API |
| Payments | Stripe with webhooks |
| Rate Limiting | Upstash Redis |
| Monitoring | Sentry with source maps |
| Styling | Tailwind CSS, Framer Motion |
| State Management | Zustand (18 stores) |
| Testing | Vitest, React Testing Library, MSW |
| Deployment | Vercel (with cron jobs, Edge functions) |

## Core Systems

### AI Generation Pipeline

5 AI modes for different use cases:
- **Default**: Standard agentic code generation with full context
- **Quick Edit**: Fast, targeted modifications to existing code
- **Explore**: Research and analysis mode for understanding codebases
- **Plan**: Structured planning before code generation
- **Ultrathink**: Deep reasoning for complex architectural decisions

The pipeline follows a Research → Plan → Generate → Validate → Review → Fix workflow with streaming responses and checkpoint-based state recovery.

### Multi-Model AI Orchestration

Supports three AI providers simultaneously:
- **Claude Agent SDK**: Primary agent for code generation with tool use
- **OpenAI**: Code review system and fallback generation
- **Cerebras (Llama 3.3-70B)**: Ultra-fast consensus validation

Models can be swapped, combined, or extended. The architecture supports adding new providers with minimal code changes.

### Checkpoint-Based Credit System

Solves the fairness problem of charging for AI generations:
1. Credits are reserved upfront based on estimated token usage
2. GenerationCheckpoint records state every N files (default 5)
3. On success: actual cost is confirmed, balance is deducted
4. On failure: reserved credits are automatically refunded
5. Status fields prevent race conditions in concurrent requests

Includes:
- CreditBalance tracking (current balance, lifetime totals)
- CreditTransaction ledger (PURCHASE, USAGE, REFUND, BONUS, ADJUSTMENT)
- GenerationSession tracking (reserved, confirmed, final credits)
- UsageRecord for detailed per-operation cost logging
- Credit reconciliation cron job to detect and fix drift
- Circuit breaker: per-hour cost limits ($10/hour default)

### CRDT File Synchronization

Production-grade multi-client file editing:
- Vector clocks on ProductFile entries for ordering
- FileOperation tracks all edits with sequence numbers
- Tombstone pattern for soft deletes (preserves history)
- Supports multi-client sync without conflicts
- Real-time bridging between WebContainer and Liveblocks

### Browser-Based Sandbox (WebContainer)

Full browser-based development environment:
- Live preview with complete file system access
- DOM tagging: semantic tagging of UI elements for AI interaction
- Browser automation: injected script for AI to click/type (HMR-aware)
- Console capture: logs errors for AI debugging
- Screenshot generation via Puppeteer
- Template snapshot (~124MB pre-built environment, cached per browser)

### DOM Selection System

Users click any element in the live preview, and the AI understands exactly what they mean:
- Semantic DOM tagging for element identification
- Visual highlighting of selected elements
- Context extraction for AI prompts
- No need to describe locations in text

### 7-Phase Specification Engine

Structured planning before any code is written:
1. Vision definition
2. Audience analysis
3. User journey mapping
4. UX/UI layout design
5. Data modeling
6. Tech foundation decisions
7. Constraints and quality standards

Includes auto-task generation that converts PRD build scope into actionable project tasks, and mockup generation via Replicate.

### Distributed Rate Limiting

Production-grade abuse prevention via Upstash Redis:
- Per-user daily generation limits (1000 default)
- Per-project hourly generation limits (100 default)
- Cost-based circuit breaker ($10/hour default)
- Project creation throttling (10/hour)
- Leaky bucket algorithm for smooth rate control

### Encrypted Secrets Vault

Enterprise-grade API key management:
- Per-project encrypted storage (AES)
- Encryption key from environment variable
- Never committed to GitHub sync
- Never exposed to client bundles
- User-provided OAuth tokens stored securely

## API Endpoints (90 Total)

### Products (18 endpoints)
Create, read, update, delete products. Deploy to Vercel. Take screenshots via Puppeteer. Manage encrypted secrets. Sync with GitHub (two-way). Provision Supabase backends. Connect WebContainer. Create version snapshots. Stream generation status via SSE. Execute commands.

### Plans & Specification (11 endpoints)
Create, read, update, delete plans. Convert plans to products. Generate 7 specification phases. Advance phase transitions. Extract session memory for context.

### Projects & Tasks (12 endpoints)
Create and manage projects. CRUD for tasks with parent-child hierarchy. Regenerate failed tasks. Drag-and-drop task reordering. Activity feed tracking. Customer brief management. Link and complete plans.

### Conversations (6 endpoints)
Create and list conversations. Access and update conversation state. Create and edit messages with role-based access (user, assistant, system).

### Payments & Credits (8 endpoints)
Check credit balance. Calculate pricing. View usage analytics. Process Stripe webhooks. Admin usage dashboard. Admin credit grants.

### User & Auth (10 endpoints)
User preference management. Clerk webhook sync. Guest-to-user conversion. Anonymous session management. Session warmup. Health checks. CSP violation logging.

### Admin & System (10 endpoints)
Analytics dashboard. Data cleanup. Guest session expiry (cron). Stale generation cleanup (cron). Credit reconciliation (cron). System metrics. WebContainer snapshot serving.

### Collaboration & Sharing (7 endpoints)
Liveblocks authentication tokens. Public share links. In-app support chat. Customer analytics. AI suggestions.

## Database Schema (24 Models)

### User Management
- **User**: Clerk integration, GitHub/Supabase OAuth, onboarding state, preferences, admin flags
- **GuestSession**: Anonymous visitors with IP/fingerprint tracking, behavioral flags, 30-day expiry

### Products
- **Product**: Name, deployment status, file storage, secrets vault, GitHub/Supabase sync, custom knowledge
- **ProductFile**: Virtual file system with tombstones for soft deletes, vector clocks for CRDT
- **ProductSnapshot**: Versioned snapshots with GitHub commit metadata
- **ProductReview**: Code review results with pass/fail scores and fix instructions

### Projects
- **Project**: Plan-to-Product lifecycle wrapper with embedded specification data
- **Plan**: Standalone PRD with 7-phase specification and session memory
- **ProjectTask**: Auto-generated from Plans, parent-child hierarchy, status workflow
- **ProjectActivity**: Activity feed tracking all project events with actor metadata
- **ProjectBrief**: Per-project customer context (reason, objectives, learnings)

### Conversations
- **Conversation**: Multi-turn chat per product with status tracking
- **Message**: Role-based messages (user/assistant/system) with attachments
- **SupportSession**: In-app support chat with batched credit tracking
- **SupportMessage**: Individual support messages with code filtering

### Credits & Billing
- **CreditBalance**: Current balance, lifetime totals
- **CreditTransaction**: Individual transactions (PURCHASE, USAGE, REFUND, BONUS, ADJUSTMENT)
- **GenerationSession**: Checkpoint-based tracking (reserved, confirmed, final credits)
- **GenerationCheckpoint**: Per-checkpoint credit snapshots during generation
- **UsageRecord**: Detailed per-operation cost logging (tokens, provider, duration)

### Infrastructure
- **Customer**: Reusable customer profile (name, industry, vision, challenges, insights)
- **Folder**: User-created folders for organizing products
- **FileOperation**: CRDT vector clocks for distributed file sync
- **SystemAsset**: Binary storage for WebContainer snapshots

## Real-Time Collaboration

Powered by Liveblocks:
- Multi-user presence and cursor awareness
- Agent context sharing with summarized state
- File sync bridging between WebContainer and collaboration layer
- Conflict-free data synchronization

## Deployment Pipeline

- **Vercel one-click deploy**: Deploy user projects to *.yourapp.example subdomains
- **GitHub two-way sync**: Auto-push generated code, pull external changes
- **Supabase integration**: OAuth-based backend provisioning
- **Deployment status tracking**: NONE → PENDING → BUILDING → DEPLOYED → FAILED

## Guest Mode & Onboarding

- Anonymous trial access with temporary guest users
- Guest-to-user conversion with credit migration
- Multi-step guided onboarding flow
- Behavioral tracking for spam detection (session count, IP, fingerprint)

## Code Review System

- Multi-dimensional review: Security, Performance, Accessibility, UX, DX
- Auto-fix generation and application
- Configurable review iterations (default 3)
- Non-blocking async review flow

## Security

- Clerk authentication with SSO and 2FA
- Content Security Policy with violation reporting
- Encrypted secrets vault (AES, per-project)
- Distributed rate limiting (Upstash Redis)
- Strict CSP headers with WebContainer WASM exceptions
- Secrets never leaked to GitHub sync or client bundles

## Monitoring & Observability

- Sentry error tracking (client, server, edge)
- Source map uploads for readable stack traces
- Structured logging with severity levels
- System metrics endpoint
- Usage analytics dashboard

## Testing

- Vitest with jsdom environment
- React Testing Library for component tests
- MSW for API mocking
- V8 coverage provider with HTML reports
- 20 test files covering stores, hooks, API routes

## CI/CD

- GitHub Actions deployment workflow
- Vercel automatic deployments on push to main
- Preview deployments per PR
- 3 Vercel cron jobs:
  - Stale generation cleanup (every 5 min)
  - Guest session expiry (every 30 min)
  - Credit reconciliation (every 5 min)

## Use Cases

### Launch a SaaS
Skip the first 18 months of infrastructure. Go straight to the product that matters. All the billing, auth, AI orchestration, and deployment infrastructure is ready.

### Internal Dev Platform
Give your engineering team an AI-powered builder without building the builder. Full sandbox environment, code generation, and deployment pipeline included.

### Catch Up to Competition
Competitors have a year-long head start. You have their entire stack — ready to fork. DOM selection, planning mode, live preview — the features users expect.

### Agency White-Label
Reskin it, rebrand it, ship it to clients. One purchase, unlimited projects. Full source code ownership means complete control over branding and features.

### AI Coding Workflows
Code generation, agentic pipelines, sandboxed execution — the plumbing is done. Add your own models, tune prompts, extend the workflow.

### Design-to-Code Tools
Visual editing, live preview, DOM selection — ship the creative workflow your users want. WebContainer sandbox provides a complete browser-based IDE.

## Purchase

- $199 (30% off, normally $285)
- One-time payment, no subscription
- Full source code ownership
- Instant access after purchase
- No vendor lock-in, no attribution required
- Deploy anywhere: your cloud, your data center, air-gapped if needed

Purchase: https://shop.platform-digital.com/b/eVqbJ1eU5bMW8Dn0wHfQI03
Website: https://www.platform-digital.com
Documentation: https://www.platform-digital.com/docs