An intuitive, AI-guided modeling platform that helps non-expert analysts build, explore, and deploy machine learning models—without writing a line of code.

This project aims to democratize modeling by combining powerful ML automation, storytelling-driven user experience, and visual workflows, wrapped in a seamless SaaS front end. The long-term vision includes a fully open-source deployment engine and intelligent model lifecycle management.

• Drag-and-drop data ingestion (CSV/XLSX)
• Automated EDA & visual insights
• AI-powered model recommendations
• Narrative-driven workflow guidance
• Feature engineering & preprocessing
• Feature Store - Centralized feature repository with versioning and reusability
• Model training and explainability tools

Future phases will include:

• Open-source deployment layer
• Automated model monitoring & retraining
• Data integration & streaming pipelines

narrative-modeling-app/
├── apps/
│   ├── frontend/         # Next.js + Tailwind UI
│   ├── backend/          # FastAPI backend for ML orchestration
│   └── mcp/              # MCP server for advanced data processing
├── scripts/              # Utility scripts for development
├── docs/                 # Project documentation
├── .github/              # GitHub Actions / CI workflows
├── README.md
└── .gitignore

• Frontend: Next.js, Tailwind CSS, NextAuth v5 (Auth), React Flow
• Backend: FastAPI, Python, Pydantic, Beanie ODM
• Modeling: scikit-learn, pandas, XGBoost, SHAP
• Database: MongoDB Atlas (cloud-hosted) with Redis caching
• Storage: AWS S3
• Auth: NextAuth with Google/GitHub providers
• Dev Tools: GitHub (issues + Actions CI/CD), uv (Python), Docker

⚠️ Project is in active solo development. Contributions and deployment tooling will be part of Phase 2+.

cd apps/backend
uv sync  # Install dependencies
uvicorn app.main:app --reload

cd apps/frontend
npm install
npm run dev

• Backend: Copy .env.example to .env and configure
  • MongoDB Atlas connection required (no local MongoDB needed)
  • Set MONGODB_URI to your Atlas connection string
  • Configure AWS S3 credentials for file storage
• Frontend: Copy .env.local.example to .env.local and configure
  • Set NEXT_PUBLIC_API_URL to backend URL
• Development: Set SKIP_AUTH=true to bypass authentication (requires ENVIRONMENT=development or test — backend startup fails hard otherwise)

Actively developed. Core platform is functional end-to-end: data ingestion, automated EDA, feature engineering, model training/explainability, and a production inference surface.

• ✅ Test suite — 1,400+ backend tests passing at 85%+ coverage
• ✅ CI/CD — GitHub Actions is the merge gate (backend lint/typecheck/tests, frontend lint/build/jest, MCP, integration, and an e2e smoke run)
• ✅ Model training & explainability — real trained-model surface under /api/v1/ml/, with SHAP-based interpretability
• ✅ Production deployment — model deploy + inference API
• ✅ MongoDB Atlas — cloud-hosted, with Redis caching and AWS S3 storage

For comprehensive documentation, see DOCUMENTATION_INDEX.md

Quick links:

• Local Development Setup
• Current Sprint (Sprint 12)
• Sprint History
• User Stories
• Production Deployment
• Production API Guide

This project is licensed under the GNU Affero General Public License v3.0 or later (AGPL-3.0-or-later).

Copyright © 2025-2026 Noatak Enterprises, LLC, dba Bria Strategy Group

Because the AGPL covers network use (§13), the complete corresponding source code of the deployed service is available at https://github.com/frankbria/narrative-modeling-app. See LICENSE for the full terms.

Frank Bria
Building solo with help from ChatGPT & GitHub Copilot
frankbria.com