Cortex - Enterprise Intelligence Platform
Overview
Most analytics teams run two stacks that never talk to each other. Dashboards and SQL on one side, a pile of LLM scripts nobody owns on the other. Cortex is one platform for both.
The BI half is Superset for dashboards, Dremio for ad-hoc queries, and a template-driven report builder. The AI half sits behind a router agent that reads a document and decides what to do with it: sort it, answer questions against it, summarize it, or profile it for quality. Both halves read the same lake, and the whole thing comes up with one docker-compose up.

Architecture
Key Features
BI Platform Services
- Dashboards - KPI views in Superset, refreshed against live data
- Self-Service Analytics - Dremio in front of the lake, so an analyst can answer their own question without waiting on a pipeline
- Report Builder - Fills a template and renders it to HTML, PDF, or DOCX
AI Platform Services
- Smart Router - An LLM ensemble reads a document and sorts it into one of 100+ categories
- RAG Chat - Ask questions of a document set. Answers come back grounded in the passages that were actually retrieved
- Summarization - Pulls key points, entities, and action items out of long documents
- Data Quality - Profiles a dataset, flags anomalies, scores what it finds
Tech Stack
- Frontend - React, Vite, TailwindCSS
- Backend - FastAPI, Python
- Data Lake - MinIO (Bronze/Silver/Gold medallion architecture)
- Databases - PostgreSQL, ChromaDB (vectors), Neo4j (graph)
- BI Tools - Apache Superset, Dremio
- AI/LLM - Ollama (Qwen, Gemma models)
- Infrastructure - Docker, Redis
Everything Runs Local
No part of this calls a hosted model. Ollama serves Qwen and Gemma on the same box, ChromaDB holds the vectors, MinIO holds the files. That was the point: an organization that cannot send internal documents to an external API still gets classification, search, and chat over them.
Quick Start
# Prerequisites: Docker, Ollama with models
ollama pull qwen2.5:14b
# Clone and start
git clone https://github.com/AlharbiAbdullah/cortex
cd cortex
docker-compose up --build -d
# Access the platform
# Frontend: http://localhost:3000
# Superset: http://localhost:8088
# Dremio: http://localhost:9047