Applied AI / Nuclear document workflow

NuclearWorkflowAI

A public local-stack applied AI demo for source-bounded document review using prepared public CNSC reactor licensing guidance.

GitHub repository

Status Public repository and local Docker Compose demo

Stack Next.js, FastAPI, Postgres, Redis, Weaviate, Minio, Ollama

Focus Source-bounded answers, saved findings, review notes, audit logs, usage limits

Scope No public uploads, OCR, ingestion, index rebuilds, project creation, or unlimited model calls

NuclearWorkflowAI Ask page showing a cited answer from public CNSC guidance — Screenshot from the NuclearWorkflowAI repository.

Project notes

The repository includes the local Docker Compose demo, product notes, architecture documentation, screenshots, checks, and public seed-data placeholders.

The design starts from traceable document work, not chat.
The demo is designed around cited source excerpts, saved findings, and summary export rather than a free-form chatbot experience.
The workflow uses prepared public CNSC guidance.
Users can browse prepared records, open source text, run search, ask grounded questions, save useful findings, and create a short review note.
The app includes RAG, GraphRAG, and OmniRAG modes.
Ask answers cite source records, refuse unsupported questions, and log model runs, retrieval traces, prompts, and responses for later inspection.
The app is built as a full local stack.
Docker Compose starts the Next.js app, FastAPI service, Postgres, Redis, Weaviate, Minio, and Ollama for a runnable local demo.
The public workflow is deliberately controlled.
Public users cannot upload files, trigger OCR, run ingestion, rebuild indexes, create projects, or make unlimited model calls.
The repository documents checks and demo readiness.
The repo includes readiness scripts, service checks, API contract notes, architecture notes, a demo script, and a portfolio review guide.