How NeuroGraph Works

Deep dive into the architecture powering our Brain Connectivity GraphRAG system.

Standard RAG vs GraphRAG

Standard RAG

Question
Vector Search (Similarity)
Top-K Text Chunks
LLM Prompt
Missing Multi-hop Context

NeuroGraph (GraphRAG)

Question
Vector Search + Entity Extraction
Graph Traversal (N-hops)
Merged Context (Graph + Vectors)
Comprehensive Multi-hop Answer

Knowledge Graph Schema

Node Types

Type Color Description Example
brain_region Cyan Anatomical structure in the brain Hippocampus, Amygdala
tract Magenta White matter pathway Fornix, Corpus Callosum
function Green Cognitive or motor function Memory Consolidation
disorder Red Neurological/psychiatric condition Alzheimer's Disease

Relationship Types

Relationship Source Node Target Node Example
connects_to brain_region brain_region Hippocampus → Amygdala
projects_via brain_region tract Hippocampus → Fornix
enables brain_region function Amygdala → Emotion Processing
affected_by brain_region disorder Hippocampus → Alzheimer's

Tech Stack

Python & FastAPI

High-performance asynchronous backend API

NetworkX

In-memory graph processing and multi-hop traversal

ChromaDB

Vector database for semantic similarity search

Google Gemini

Advanced LLM for entity extraction and synthesis

Three.js & 3D Force Graph

WebGL-accelerated interactive 3D visualizations