🛠️ Synthron – Custom PC Recommendation Platform

Synthron is a full-stack web application that helps users intelligently build custom PC configurations based on their needs (gaming, editing, budget, etc.) using a combination of AI prompt parsing and real-time price scraping from trusted websites.

The system uses React for the frontend and Django + Django REST Framework on the backend, connected via REST APIs.PostgreSQL is used to store User's data and Redis acts as a cache system to temporarily store the results for faster accessibility. It uses an LLM (LLaMA 3.2) locally to interpret user prompts, and Playwright for dynamic scraping of part prices and availability.

🔍 Currently achieving 70–75% scraping predictability/accuracy. Further improvements are planned.

---

✨ Features

• 🧠 AI-driven prompt interpretation (LLaMA 3.2)
• 🔍 Real-time scraping of PC parts from trusted e-commerce websites
• ⚙️ Component agents: CPU, GPU, RAM, Storage, Motherboard
• 🌐 Region-aware pricing and availability
• 🔐 JWT-based authentication with HttpOnly cookie storage
• 📦 Build history: Saved per user and fetchable later
• 🧪 Compatibility checker: Validates part compatibility (e.g., socket types)
• ⚡ Fast, modern UI using React + Tailwind CSS

---

📁 Folder Structure

ai-pc-builder/
├── backend/
│   ├── api/
│   │   ├── agents/
│   │   │   ├── cpu_agent.py
│   │   │   ├── gpu_agent.py
│   │   │   ├── ram_agent.py
│   │   │   ├── storage_agent.py
│   │   │   ├── motherboard_agent.py
│   │   │   ├── llama.py
│   │   │   ├── playwright_scraper.py
│   │   │   └── product_finder.py
│   │   ├── compatibility_checker.py
│   │   ├── serializers.py
│   │   ├── views.py
│   │   ├── urls.py
│   ├── settings.py
│   ├── manage.py
│   ├── .env
│   └── requirements.txt
├── frontend/
│   ├── public/
│   ├── src/
│   │   ├── components/
│   │   ├── pages/
│   │   ├── api.js        
│   │   ├── App.jsx
│   │   └── index.jsx
│   ├── tailwind.config.js
│   ├── vite.config.js
│   ├── package.json
├── README.md
└── .gitignore

---

⚙️ Backend – Django Setup

1. Clone and set up virtual environment:

git clone https://github.com/yourusername/ai-pc-builder.git
cd ai-pc-builder/backend

python -m venv env
source env/bin/activate  # Windows: env\Scripts\activate

pip install -r requirements.txt

2. Create .env file:

SECRET_KEY=your_django_secret_key
DEBUG=True
ALLLOWED_HOSTS=localhost,127.0.0.1
DB_NAME=
DB_USER=
DB_PASSWORD=
DB_HOST=
DB_PORT=5432

3. Run migrations and start server:

python manage.py makemigrations
python manage.py migrate
python manage.py runserver

📦 Backend Dependencies (backend/requirements.txt)

Django>=4.2
djangorestframework
python-dotenv
httpx
playwright
psycopg2-binary

To install Playwright browsers:

playwright install

---

🎨 Frontend – React Setup

1. Navigate to frontend directory:

cd ../frontend

2. Install dependencies:

npm install

3. Run development server:

npm run start

---

⚠️ LLaMA 3.2 Usage

Synthron uses LLaMA 3.2 locally for AI prompt parsing.
To fully use the project, you'll need to download and run the LLaMA 3.2 model locally.
Refer to the llama.py script for how it's integrated, and ensure your machine meets the requirements for running LLaMA models.

---

🔗 How It Works

1. User Input: "I want a gaming PC under $1000 in Usa and i prefer an AMD build."
2. LLM parses intent, budget, and region.
3. Each component agent (CPU, GPU, RAM, etc.) activates.
4. Agents create product-specific search queries.
5. Serper fetches urls off the web.
6. Playwright scrapes trusted e-commerce websites.
7. Components are selected and ranked by:
   • 🔄 Stock availability
   • 💰 Lowest price
   • 🙋 User preference
8. The result is rendered on the frontend with prices and product links.

---

✅ Trusted Sites Scraped

• Amazon
• Newegg
• Microcenter
• Scan UK
• And more…

🗑️ ScrapeGraph API was removed due to limited free-tier usage. All scraping is now handled with Playwright.

---

⚠️ Known Limitations

• 🛒 Scraping accuracy is 70–75% — dynamic rendering and anti-bot measures may impact results.
• ❌ Not all product queries yield valid results (fallback logic is used).
• 🌍 Currency and regional detection are still basic.

---

🔮 Roadmap

• 🧠 Improve scraping reliability via AI heuristics
• 🔌 Add PSU and case suggestions
• 💸 Integrate Amazon Affiliate tracking
• 📊 Admin dashboard for scraping diagnostics
• 🔐 OAuth login support (Google, GitHub)
• 🛠️ Improving LLM integration and scraping engine.

---

🧠 What's Next?

I’ll be actively improving Synthron with:

• ✨ More exciting features and enhancements
• 🛠️ Integration of modern DevOps tooling and workflows
• 🐳 Containerization using Docker ✅
• ⚙️ Implementation of CI/CD pipelines
• 🚀 Cloud deployment on AWS

---

🤝 Want to Contribute?

Contact me on my email: rehansaqib2006@gmail.com

Feel free to open an issue or PR!