Eureka — AI Research Ideation Platform

Turn a research topic into breakthrough ideas in minutes — not weeks.

Eureka is an AI-powered research companion that analyzes hundreds of scientific papers, maps knowledge gaps, synthesizes novel ideas, and evaluates them for novelty and feasibility — all in one flow. Built for researchers, students, and innovators who need to move fast without sacrificing depth.

---

What Eureka Does

You start with a single topic. Eureka's multi-agent pipeline handles the rest:

1. Scout — Fetches and reads the latest papers from arXiv on your topic
2. Gap Analyst — Identifies underexplored areas and white spaces in the literature
3. Innovator — Synthesizes breakthrough ideas at the intersection of known research
4. Critic — Evaluates each idea for novelty, feasibility, and coherence
5. Coherence Validator — Ensures the full idea set is consistent and well-structured

The result: a ranked list of ideas with scores, citations, validation plans, and a visual knowledge map — ready to export and continue in your workflow.

---

Key Features

• Real-time reasoning trace — Watch each agent think and work live
• Ranked idea cards — Sorted by novelty + feasibility, click to deep-dive
• Knowledge map — Interactive graph of clusters and papers, drag/zoom/filter
• Live ROI estimate — See how much time Eureka saves vs. manual research
• Export to TRAE — One-click copy idea as prompt, or download JSON — continue your work in AI agents
• Profile settings — Set your mode (academic/product/develop/review) and context

---

Quick Start

Run locally

cd D:\PROJECT\trae_project\repo_gh
.\run-dev.ps1

Then open http://localhost:5174

Backend runs on port 8000, frontend on 5174. For alternative ports: .\run-dev.ps1 -BackendPort 8000 -FrontendPort 5173

Using Eureka

1. Enter a research topic (or pick an example chip on the landing page)
2. Click Run — watch the Reasoning Engine work in real time
3. Explore Ideas — click any card to see full detail and scores
4. Switch to Knowledge Map — pan/zoom the interactive graph
5. Export ideas to TRAE to continue building

Mock vs Real Backend

• Mock mode (default): Works immediately, no setup needed. Uses sample data.
• Real mode: Set VITE_USE_MOCK=false in frontend/.env then restart. Requires backend running on port 8000.

---

Tech Stack

• Frontend: React 19 + Vite + TypeScript + Tailwind CSS
• Backend: FastAPI + Python
• Graph: D3-force for interactive knowledge maps
• SSE: Real-time streaming from backend agents

---

Project Structure

repo_gh/
├── backend/        FastAPI backend (O1)
├── frontend/      React frontend (O2)
├── shared/        Shared TypeScript types
├── docs/          Specs and plans
└── run-dev.ps1    One-command launcher

---

Contributing

This project is developed as a collaboration between O1 (backend) and O2 (frontend). See docs/ for specs and plans.