Seezer.ai

AI-powered website builder — describe what you want, get a production-ready React app.

Seezer.ai converts plain English prompts into fully functional, build-verified React applications using a multi-agent pipeline, isolated sandbox execution, and real-time streaming.

---

What It Does

Type a prompt like "Build a SaaS landing page with a hero section, pricing cards, and CTA" — Seezer.ai plans, builds, validates, and verifies a complete React project, ready to deploy.

• Generates complete React applications from natural language
• Runs code safely inside isolated E2B sandboxes
• Streams live build logs via WebSockets
• Verifies production builds automatically before marking them ready
• Persists project state and chat history across sessions

---

Architecture

User Prompt
     │
     ▼
API Server (Express)
     │
     ▼
Redis Queue
     │
     ▼
Worker Process
     │
     ▼
Agent Pipeline
  ├── PlannerAgent       — Analyzes context, generates structured build plan
  ├── BuilderAgent       — Creates/modifies React components and routing
  ├── CodeValidatorAgent — Fixes imports, syntax errors, dependencies
  └── AppCheckerAgent    — Runs npm build + lint, fixes compilation errors
     │
     ▼
E2B Sandbox (isolated execution)
     │
     ▼
WebSocket Live Logs  →  Database (PostgreSQL)

---

Multi-Agent Pipeline

Agent	Responsibility
PlannerAgent	Reads existing project files, generates a structured build plan, decides which files to create or modify
BuilderAgent	Writes React components, manages routing, uses sandbox tools for file operations
CodeValidatorAgent	Detects broken imports, fixes syntax errors, ensures dependency correctness
AppCheckerAgent	Runs npm run build, fixes compilation errors, executes linting, confirms production readiness

Each agent operates with tool-based AI reasoning — not just text generation.

---

Tech Stack

Backend: Node.js, Express, Redis, PostgreSQL, Prisma ORM, JWT Auth

AI & Execution: OpenAI Agent SDK, E2B Sandbox, structured prompt engineering

Real-Time: WebSockets for live build log streaming

---

Getting Started

Prerequisites

• Node.js + pnpm
• Docker (for PostgreSQL and Redis)
• OpenAI API key
• E2B API key

1. Clone

git clone https://github.com/yourusername/seezer-ai.git
cd seezer-ai

2. Install dependencies

pnpm install

3. Configure environment

Create a .env file:

DATABASE_URL=postgresql://postgres:password@localhost:5432/postgres
REDIS_URL=redis://localhost:6379
JWT_SECRET=your_secret
OPENAI_API_KEY=your_openai_key
E2B_API_KEY=your_e2b_key
E2B_TEMPLATE_ID=code-interpreter-v1

4. Start infrastructure

# PostgreSQL
docker run -p 5432:5432 -e POSTGRES_PASSWORD=password postgres

# Redis
docker run -p 6379:6379 redis

5. Run

pnpm dev

---

How a Build Works

1. User submits a prompt via the API
2. Job is pushed to Redis and picked up by a worker
3. Agent pipeline executes inside an E2B sandbox
4. Build logs stream to the client via WebSocket in real time
5. Final build is verified — project marked ready and state saved to PostgreSQL

---

Key Design Decisions

Incremental updates — agents read existing project files before making changes, preserving context across edits.

Sandboxed execution — all code runs inside E2B, never on the host.

Automated build verification — AppCheckerAgent runs npm run build and lint on every generation. Projects are only marked ready when the build passes.

Redundancy prevention — the pipeline detects and skips unnecessary package installs.

Persistent memory — project context is stored between sessions so follow-up prompts build on prior work.

---

Security

• All code execution is sandboxed via E2B — no arbitrary host access
• JWT-based authentication on all endpoints
• Tool invocations are controlled and scoped
• Build verification runs before any project is marked deployable

---

Roadmap

• Versioned project snapshots
• One-click deployment (Vercel / Netlify integration)
• Persistent vector memory for long-term context
• Prompt template library
• Rate limiting and billing system

---

Author

Nitin Yadav — B.Tech CSE, Backend & AI Systems Engineer

Built with a focus on real infrastructure: distributed job processing, multi-agent orchestration, secure sandbox execution, and production build automation.