🎓 AI Support Fund Analysis Service

Upload documents such as resident registration, income certificates, enrollment certificates, and transcripts, and let the AI analyze your eligibility for support funds and recommend relevant programs.

🏗 Project Overview

• Project Name: AGI-Hackaton
• Objective: Upload documents → Extract information using Upstage API → Recommend optimal government support or scholarship programs using an AI recommendation system.

🧩 Project Structure

AGI-Hackaton
├── backend
│   ├── main.py
│   ├── database.py
│   ├── domain
│   │   ├── student
│   │   └── scholarship
│   ├── llm_router.py
│   ├── llm_model.py
│   └── ...
├── frontend
│   └── ...

⚙ Key Features

📑 Document Upload and Analysis

• Resident Registration, Income Certificates → Extract general user information (name, age, income, etc.) (Planned)
• Enrollment Certificates, Transcripts → Extract academic and grade information.

🤖 Upstage API

• Document Parsing: Extract text from PDFs.
• InfoExtract: Structure key information (income, major, grades, etc.) from documents.
• Powered by langchain_upstage + solar-pro models for advanced Korean document processing.

🌐 AI-Based Program Recommendation

• Analyze user information and document data using trained AI models (langchain_upstage, HuggingFaceEmbeddings, etc.).
• Provide personalized recommendations for programs (scholarships, government support funds).

🔑 Tech Stack

• Backend: FastAPI, SQLAlchemy, SQLite (async engine)
• AI Model: LangChain, HuggingFace, Upstage
• Frontend: React (TypeScript)
• Infra / DevOps: uvicorn, pip, GitHub

🚀 How to Run

cd backend
python -m venv venv
.\venv\Scripts\activate  # Mac/Linux: source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --reload

📦 Usage Examples

1. Student User

   • Upload enrollment and transcript documents via the /upload/student endpoint → Save to DB.
   • Get AI recommendations via the /recommend/recommend-programs endpoint.

2. Final Results with Additional Input

   • Use the /recommend/final-result endpoint to combine session data, recommended programs, and additional user input for a final response.

🧭 API Endpoints

• POST /create-session: Generate a session using a UUID.
• POST /upload/scholarship: Upload scholarship documents → Save to DB.
• GET /recommend/recommend-programs: Recommend scholarships/support programs based on user information.
• POST /recommend/final-result: Provide final recommendations based on additional user input.

📸 Example Screen

🤝 Contribution Guidelines

1. Fork the repository → Create a new branch → Implement features.
2. Submit a Pull Request (PR) to propose changes.
3. Review the code and merge after approval.