• Stars
    star
    412
  • Rank 105,024 (Top 3 %)
  • Language
    TypeScript
  • License
    MIT License
  • Created over 1 year ago
  • Updated over 1 year ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Classify and extract structured data with LLMs

🏷 LLMParser

LLMParser is a simple and flexible tool to classify and extract structured data from text with large language models.

📖 Full documentation available here

npm package Build Status

Why?

While LLMs are extremely powerful, producing reliable JSON output is challenging.

LLMParser aims to solve this by enforcing a consistent JSON input and output format for classifying and extracting text with LLMs.

What can you do?

LLMParser is a fairly general purpose tool. You can use it to extract job titles from LinkedIn profiles, dishes from restaurant menus, or even classify reviews as positive or negative. Here are some more examples:

  • Extracting name, school, current job title from resumes
  • Classifying corporate contracts as NDA, MSA, etc. and extracting important fields like effective date and counterparty name
  • Extracting place names from Apple Notes

Install

npm install llmparser

Usage

Quick note: this library is meant for server-side usage, as using it in client-side browser code will expose your secret API key. Go here to get an OpenAI API key.

import { LLMParser } from llmparser;

const categories = [
  {
    name: "MSA",
    description: "Master service agreement",
  },
  {
    name: "NDA",
    description: "Non disclosure agreement",
    fields: [
      {
        name: "effective_date",
        description: "effective date or start date", // instruction for LLM
        type: "string"
      },
      {
        name: "company",
        description: "name of the company",
        type: "string"
      },
      {
        name: "counterparty",
        description: "name of the counterparty",
        type: "string"
      }
    ]
  }
]

const parser = new LLMParser({
  categories,
  apiKey: process.env.OPENAI_API_KEY
})

const ndaText = await loadPDFAsText("src/nda.pdf") // get text of PDF
const extraction = await parser.parse(ndaText);

Classified as an NDA and extracted 3 fields.

{
  "type": "NDA",
  "confidence": 1,
  "source": "This is a Mutual Non-Disclosure Agreement (this “Agreement”), effective as of the date stated below (the “Effective Date”), between Technology Research Corporation, a Florida corporation (the “Company”), and Kevin Yang (the “Counterparty”).",
  "fields": {
    "effective_date": {
        "value": "2022-01-11T06:00:00.000Z",
        "source": "Effective date of January 11, 2022",
        "confidence": 1
    },
    "company": {
        "value": "Technology Research Corporation",
        "source": "between Technology Research Corporation, a Florida corporation",
        "confidence": 0.9
    },
    "counterparty": {
        "value": "Kevin Yang",
        "source": "and Kevin Yang (the “Counterparty”)",
        "confidence": 0.9
    }
  }
}