Sophiie AI Agents Hackathon 2026

Your Submission

Instructions: Fill out this section in your forked repo. This is what judges will see first.

Participant

Field	Your Answer
Name	Atharva Santosh Mavale
University / Employer	Deakin University

Project

Field	Your Answer
Project Name	StudyMate AI (AI Questioning Tutor for Code Repos)
One-Line Description	An AI-powered Questioning coding tutor that helps students understand code (especially intimidating GitHub repositories) by asking the right questions, giving progressive hints, and guiding learning step-by-step.
Live App Link	https://atharvamavle-hackathon-uiapp-ps07su.streamlit.app/
Demo Video Link	https://youtu.be/CJm-7gzGaEs
Tech Stack	FastAPI Uvicorn, OpenAI ,Python SDK, LangChain
AI Provider(s) Used	OpenAI , GPT

StudyMate — AI Questioning Tutor for Code Repos

An AI-powered Questioning coding tutor that helps students understand code (especially intimidating GitHub repositories) by asking the right questions, giving progressive hints, and guiding learning step-by-step.

This project was built for the Sophiie AI Agents Hackathon 2026.

• Backend: FastAPI (deployed on Render)
• Frontend: Streamlit (deployed on Streamlit Community Cloud)
• LLM: OpenAI (configured via OPENAI_API_KEY)

---

The gap we’re solving (real problem)

You’ve identified a massive gap in AI education that most tools don’t solve well:

Current reality

• Students copy‑paste AI-generated code without understanding.
• They get stuck when code breaks or needs changes.
• Typical chatbots either dump code or give overwhelming explanations.
• Large GitHub repos feel impossible to start (lots of files, complex structure).
• Traditional step-by-step tutorials don’t adapt to a student’s confusion.

Our solution

• A Questioning AI agent that teaches by asking questions instead of giving answers.
• Students learn by doing: the agent helps them reason, not copy.
• Adaptive dialogue based on student level (beginner/intermediate/advanced).
• Progressive hints when needed.

---

Why this is perfect for the hackathon

1. Unique & impactful

• Directly addresses the “AI copy‑paste crisis” in education.
• Focuses on understanding and reasoning, not output.

2. Agentic behavior (not just a chatbot)

• Multi-turn memory (session-based chat).
• Guided questioning (Questioning method).
• Prompt-driven teaching behavior.

3. Production-ready architecture

• Deployed backend + UI.
• Health endpoint and API docs.

---

Features

• 🤖 Questioning dialogue: Probing questions instead of direct answers.
• 💬 Session-based chat: Each user gets a session_id.
• 🧠 Adaptive teaching: Student name + knowledge level influence the tone and depth.
• 🧪 API docs included: FastAPI Swagger at /docs.
• 🚀 Deployed: Backend on Render, UI on Streamlit.

---

Tech stack

Backend

• FastAPI
• Uvicorn
• OpenAI Python SDK
• LangChain ChatOpenAI

Frontend

• Streamlit
• Requests

---

Live application

• Live APP : https://atharvamavle-hackathon-uiapp-ps07su.streamlit.app/
• Backend (Render): https://hackathon-24mr.onrender.com/docs#/default/create_session_session_create_post

---

Project structure

hackathon/
├── agent/                # Core agent logic
     └── __init__.py
     └── core.py
     └── prompts.py
     └── tools.py          
├── api/                   # FastAPI backend
│   └── main.py            # API routes: /session/create, /chat, /health
├── ui/                    # Streamlit UI
│   └── app.py             # Frontend app
├── requirements.txt
├── .gitignore
└── README.md

---

How it works (core flow)

Student selects a GitHub repo (or future: paste text / upload file)
        ↓
Backend creates a session and agent
        ↓
Agent: “What interests you most about this project?”
        ↓
Student answers
        ↓
Agent asks guided questions + gives hints
        ↓
Student builds understanding step-by-step

---

Prerequisites

• Python 3.9+
• Git
• OpenAI API key

---

Installation & local run

1) Clone

git clone https://github.com/atharvamavle/hackathon.git
cd hackathon

2) Create and activate venv

Windows (PowerShell):

python -m venv .venv
. .\.venv\Scripts\Activate.ps1

macOS/Linux:

python3 -m venv .venv
source .venv/bin/activate

3) Install dependencies

pip install -r requirements.txt

4) Set environment variables (local)

Create a local .env file in the repo root:

OPENAI_API_KEY=sk-your-real-key
OPENAI_MODEL=gpt-4o-mini

Important: Do not commit .env.

5) Run backend

uvicorn api.main:app --host 0.0.0.0 --port 8000 --reload

Verify:

• http://localhost:8000/health
• http://localhost:8000/docs

6) Run Streamlit UI

streamlit run ui/app.py

Open the URL Streamlit prints (usually http://localhost:8501).

---

Deploy

Backend on Render

1. Create a Render Web Service connected to this repo.
2. Set environment variables:
   • OPENAI_API_KEY = your real key
   • OPENAI_MODEL = optional
3. Set Health Check Path to: /health
4. Deploy.

UI on Streamlit Community Cloud

In Streamlit app settings → Secrets, add:

API_BASE_URL = "https://hackathon-24mr.onrender.com"

In ui/app.py, set:

API_BASE = st.secrets["API_BASE_URL"].rstrip("/")

---

API endpoints

• GET / — basic status
• GET /health — health + whether OPENAI_API_KEY is configured
• POST /session/create — start a session, returns session_id and greeting
• POST /chat — send a message, returns model response
• GET /session/{session_id}/history — session transcript
• GET /session/{session_id}/progress — progress tracking (if enabled)

---

Troubleshooting

401 Incorrect API key

• Make sure OPENAI_API_KEY is set (Render env vars for production, .env for local).
• Ensure .env is not committed and you are not overriding production env vars.

UI can’t reach backend

• In Streamlit Cloud, localhost won’t work.
• Use API_BASE_URL secret pointing to your Render URL.

---

Future improvements

• Repo cloning + indexing: actually clone GitHub repos server-side.
• Retrieval / RAG grounded in real files and code snippets.
• “Explain this file” and “trace call path” tools using AST parsing.
• Better persistence: database for sessions + progress tracking.
• Streaming responses and richer UI.
• Auth + rate limiting + safer CORS defaults.

---

Author

Atharva Santosh Mavale

Contact: atharvamavale40@gmail.com

---

Status: ✅ Live and deployed

sophiie.com

---