medpromptjs

A set of base classes for making GenAI application development easy. Implements the LLM-in-the-Loop CQL execution on unstructured data.

Installation

npm i --save medpromptjs

Usage

BaseLLM

BaseLLM is a base class for LLMs that communicate with a remote API. It extends the LangChain LLM class and provides configurable parameters for model inference.

import { BaseLLM } from 'medpromptjs';

const llm = new BaseLLM({
  baseUrl: 'http://localhost:8080/api/chat',
  model: 'llama2',
  apiKey: 'your-api-key', // optional
  temperature: 0.1,
  maxOutputTokens: 512,
});

const response = await llm.invoke('Hello, how are you?');

BaseEmbedding

BaseEmbedding is a base class for embeddings that communicate with a remote API. It extends the LangChain Embeddings class.

import { BaseEmbedding } from 'medpromptjs';

const embedding = new BaseEmbedding({
  baseUrl: 'http://localhost:8080/api/embeddings',
  model: 'nomic-embed-text',
  apiKey: 'your-api-key', // optional
});

// Embed multiple documents
const embeddings = await embedding.embedDocuments(['doc1', 'doc2']);

// Embed a single query
const queryEmbedding = await embedding.embedQuery('search query');

Give us a star ⭐️

If you find this project useful, give us a star. It helps others discover the project.

Related projects

• FHIRy - FHIR to pandas dataframe
• kedro-multimodal - Template for multi-modal machine learning in healthcare using Kedro
• ckblib - A library for clinical knowledge graphs

Contributing

• PR welcome

Contributers

• Bell Eapen |
• My Blog post