An AI-powered mentor that guides hackathon teams from idea to demo using real-time reasoning, RAG, and agent orchestration.
- Refine ideas and reduce scope
- Plan implementation effectively
- Design demo-ready architecture
- Prepare winning pitches
All in real time, adapting to team skills and constraints.
Hackathon teams often struggle with:
- Overly ambitious or poorly scoped ideas
- Missing critical architecture components
- Limited mentor availability
- Poor planning, leading to incomplete demos
Currently, no system exists that can:
- Think and guide like a mentor
- Understand time, team size, and skill constraints
- Provide step-by-step actionable advice
Our agent solves these problems by:
- Understanding team ideas and constraints
- Detecting risks in scope and feasibility
- Suggesting realistic MVPs and features
- Recommending simple, demo-focused architectures
- Creating step-by-step build plans and timelines
- Generating pitch and demo scripts
- Idea Understanding & Context Analysis – Captures goals, users, tech stack, and constraints
- Interactive Question Loop – Asks clarifying questions until the idea is clear
- Scope & Feasibility Analysis – Evaluates ideas against time, team size, and complexity
- Automatic MVP Generation – Converts big ideas into achievable MVPs
- Skill-Aware Planning – Adapts advice based on team experience
- RAG-Based Guidance – Recommends strategies using real hackathon patterns
- Architecture Recommendation – Outputs clear, demo-ready designs
- Step-by-Step Build Plan – Time-boxed development roadmap
- Real-Time Mentorship Streaming – Shows agent reasoning live
- Failure Prediction – Warns about unrealistic goals or missing demos
- Pitch & Demo Scripts – Ready-to-use judge-friendly scripts
- Agent Reasoning Visualization – Visualizes decisions and reasoning paths
Frontend (React)
│
│ Server-Sent Events (Real-Time Updates)
▼
Backend (FastAPI)
│
├── Mentor Agent
│ ├── Idea Analyzer
│ ├── Clarification Node
│ ├── Feasibility Checker
│ ├── MVP Generator
│ ├── RAG Retriever
│ ├── Planner
│ └── Pitch & Demo Generator
│
└── Vector Database + LLM
SmartHackathonMentor/
├── frontend/ # React UI
├── backend/ # FastAPI & Agent Logic
├── .gitignore
└── README.md
- Node.js (v18+) & npm
- Python 3.11+
- Vector Database or MongoDB
git clone <repo-url>
cd SmartHackathonMentorInstall frontend:
cd frontend
npm installInstall backend:
cd backend
pip install -r requirements.txtSet environment variables:
cp .env.example .env
# Fill in the required valuesFrontend:
cd frontend
npm run devBackend:
cd backend
uvicorn main:app --reload1️⃣ Idea Intake
- Teams submit their hackathon idea
- The agent extracts goals, users, tech stack, and constraints
- Example visual: form or chat interface showing idea input
2️⃣ Clarification & Analysis
- The agent asks clarifying questions
- Feasibility and scope risks are analyzed instantly
- Example visual: question prompts appearing in the UI
3️⃣ MVP & Architecture Generation
- Large or complex ideas are converted into achievable MVPs
- The agent recommends a demo-focused architecture
- Example visual: diagram of simplified system architecture
4️⃣ Planning & Timelines
- Step-by-step build plans created based on team skills
- Time-boxed tasks ensure demo completion
- Example visual: timeline or Kanban board of tasks
5️⃣ Pitch & Demo Support
- Generates ready-to-use pitch and demo scripts
- Helps teams rehearse for judges
- Example visual: sample pitch screen or generated script
We follow Git Flow:
main– production-readydevelop– integration branchfeature/*– new featuresbugfix/*– fixeshotfix/*– urgent fixes
Workflow:
- Create a feature branch from
develop - Implement changes & commit
- Push branch and create PR to
develop - Code review & merge
- Release from
develop→main
Commit Types: feat, fix, docs, refactor, test, chore