Smart Document Router

The Smart Document Router is an open source document processing data layer.

• It ingests unstructured docs through REST APIs and integrations from faxes, emails, and ERPs.
• It processes documents at scale with OCR and LLMs
• And it chunks, embeds, and organizes documents into queriable knowledge bases

The Document Router is designed to work standalone or with a human-in-the-loop, and can process medical, insurance, financial, supply chain, and legal documents.

It acts as a system of record for the extraction schemas and prompts, and it is portable over all major clouds and LLM providers.

A Document Agent is available to configure prompts and extractions, and to review processed results.

Tech stack

• NextJS, NextAuth, MaterialUI, TailwindCSS
• FastAPI
• MongoDB
• Pydantic
• LiteLLM
• OpenAI, Anthropic, Gemini, Vertex AI for GCP, AWS Bedrock, xAI, OpenRouter...

PyData Boston DocRouter Slides (Feb '24) have more details about tech stack, and how Cursor AI was used to build the DocRouter.

User Experience

Example Deployment

Presentations

• Smart Document Router Slides from Boston PyData, Spring 2025
• DocRouter.AI: Adventures in CSS and AI Coding, Summer 2025

Docs

• Installation
  • Local Development Setup
  • Docker Setup
  • AWS Setup
• Development
  • Environment Variables Guide
  • Database Migrations Guide