Mantis-Basic-Version-of-Cognitive-Cartography

MANTIS V3 – Cognitive Discovery Engine

“A human-in-the-loop AI system that maps knowledge and generates research hypotheses.”

Problem Statement

• Hard to Explore: PDFs and books are dense; finding connections across 100+ documents is impossible manually.
• Linear Reading Fails: Traditional reading doesn't reveal hidden structural links between distinct domains.
• Discovery is Difficult: Generating novel hypotheses requires bridging gaps between isolated knowledge clusters.

Solution Overview

• Cognitive Maps: Visualizes documents as a 2D galaxy of ideas.
• Discovery Bridges: Identifies and visualizes "paths" between disconnected clusters (e.g., Biology ↔ Architecture).
• Hypothesis Generation: AI analyzes the "bridge" to propose novel research ideas connecting the two domains.

---

Features

• Text/PDF/Book Ingestion: Parses and creates intelligent chunks.
• Cognitive Map: Interactive 2D visualization of knowledge space.
• Discovery Bridges: Pink dotted lines showing potential interdisciplinary connections.
• Hypothesis Generation: One-click generation of research proposals.
• Human-in-the-Loop AI: Users guide the discovery by selecting bridges.

Tech Stack

• Frontend: React (Vite)
• Backend: FastAPI (Python)
• AI: Sentence Transformers (all-mpnet-base-v2) + LLM (gpt-oss-120b via OpenRouter)
• Deploy: Vercel (Frontend) + Railway (Backend)

System Architecture

[PDFs] -> /ingest -> [Chunks] -> [Embeddings]
                                      |
                                      v
                                [Clustering] -> [Cognitive Map]
                                      |
                                      v
                                [Bridge Detection] -> [Visual Bridge (Pink Line)]
                                      |
                                      v
[User Click] -> /hypothesize -> [Novel Research Hypothesis]

---

How to Run Locally

1. Backend Setup

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

Backend runs on http://localhost:8000

2. Frontend Setup

cd frontend
npm install
npm run dev

Frontend runs on http://localhost:5173

---

API Contract

---

Deployment Guide

Railway (Backend)

1. Root Directory: backend
2. Start Command: uvicorn main:app --host 0.0.0.0 --port $PORT
3. Add Environment Variable: OPENROUTER_API_KEY

Vercel (Frontend)

1. Root Directory: frontend
2. Environment Variable: VITE_API_BASE=https://your-railway-url.app

---

Demo Flow

1. Upload: Click "Upload Data" -> "Load Sample Data" (e.g., Physics + Music).
2. Visualize: See two distinct clusters on the map.
3. Discover: A pink dotted line appears connecting them.
4. Hypothesis: Click the line -> "Generate Hypothesis".
5. Result: Read the novel concept generated by the AI.

License

MIT License