App Compiler 🚀

An AI-powered multi-stage pipeline that works like a compiler — takes a natural language prompt and outputs a complete, validated, executable app configuration.

What it does

Input: "Build a CRM with login, contacts, dashboard, role-based access, and payments"

Output: Complete validated JSON config with UI + API + DB + Auth schemas that is proven executable via runtime checks.

Pipeline Architecture

Natural Language
      ↓
Stage 1 — Intent Extraction
Parses entities, roles, features, assumptions
      ↓
Stage 2 — Architecture Design
Entity relationships, flows, page structure
      ↓
Stage 3 — Schema Generation (4 parallel)
DB Schema · API Schema · UI Schema · Auth Schema
      ↓
Stage 4 — Validation + Repair Engine
5 cross-layer checks · Surgical repair (not brute retry)
      ↓
Runtime Executor
SQLite in-memory · API endpoint validation

Key Features

• Multi-stage pipeline — not a single prompt
• Surgical repair engine — detects which layer failed and fixes only that layer
• Cross-layer validation — UI fields map to API endpoints which map to DB columns
• Runtime execution proof — DB schema runs in SQLite, API endpoints validated
• Evaluation framework — 20 test cases with real metrics

Eval Results

Metric	Result
Real prompts (10/10)	100% success
Edge cases (5/10)	50% success
Runtime success	70%
Avg latency	93s
Total repairs	11

Tech Stack

• Backend — Python, FastAPI, Pydantic v2
• Frontend — React, Vite
• AI Model — Groq API (llama-3.1-8b-instant)
• Runtime — SQLite in-memory
• Validation — Pydantic v2 cross-layer consistency checks

Project Structure

app-compiler/
├── pipeline/
│   ├── stage1_intent.py        # Intent extraction
│   ├── stage2_architecture.py  # Architecture design
│   ├── stage3_schemas.py       # Schema generation
│   └── stage4_refinement.py    # Validation + repair
├── schemas/                    # Pydantic contracts
├── validators/                 # Cross-layer checks
├── repair/                     # Targeted repair engine
├── runtime/                    # Execution simulator
├── eval/                       # Evaluation framework
├── api/                        # FastAPI backend
└── ui/                         # React frontend

How to Run Locally

Backend

git clone https://github.com/Aryan-0-07/app-compiler
cd app-compiler
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
echo GROQ_API_KEY=your_key_here > .env
uvicorn api.main:app --reload --port 8000

Frontend

cd ui
npm install
npm run dev

Open http://localhost:5173

How it Works

1. Type a natural language prompt describing your app
2. Click Generate App Config
3. The 4-stage pipeline runs automatically
4. Results show in tabbed UI — Summary, DB, API, UI, Auth, Log
5. Log tab shows runtime execution proof

Evaluation

Run the full evaluation framework:

python eval/runner.py

View results at http://localhost:8000/eval/results