Skip to content

AI products and agents, Open source

AI Legal Document Analyzer

Structured data and semantic search for legal documents

A NestJS service that turns legal PDFs and HTML into structured metadata and vector embeddings, queryable over REST and GraphQL.

View on GitHub (opens in new tab)
The Documents GraphQL query from the Legal Document Analyser README (paginated documents with id, title, summary, jobs and meta) typed into an editor, next to its REST endpoints: POST /docs/upload, GET /docs and GET /docs/:id.
GraphQL and REST API from the README

Stack

  • NestJS
  • TypeScript
  • LangChain
  • OpenAI
  • Milvus
  • GraphQL
  • Redis
Status
Open source
Type
AI products and agents
Source
GitHub

The problem

Legal documents arrive as long PDFs and HTML pages, and the data inside them is hard to query.

What it does

The service ingests legal PDF and HTML files, extracts structured metadata with OpenAI and LangChain, stores semantic vectors in Milvus, and exposes REST and GraphQL APIs for querying the processed documents.

How it's built

TypeScript on Node.js and NestJS, OpenAI through LangChain, Milvus as the vector database, Redis, and GraphQL alongside REST.

Key features

  • Ingests legal documents as PDF or HTML
  • Extracts structured metadata with OpenAI and LangChain
  • Stores semantic vectors in Milvus
  • Query processed documents over REST or GraphQL

Screens

  • System diagram from the repository README: REST upload goes to DocsController, which stores file info in PostgreSQL and publishes METADATA and SUMMARY jobs to BullMQ and Redis; the metadata processor writes vectors to Milvus, and GraphQL and REST read the results.
    System diagram from the README

Contact

Want something like this built?

Get in touchExperience