xron-vercel

XRON integration for Vercel AI SDK — compress data in LLM messages for up to 80% token reduction.

Installation

npm install xron-vercel xron-format ai

Usage

Compress Data for Prompts

import { compressDataForPrompt } from 'xron-vercel';

const products = [
  { id: 1, name: 'Widget', price: 9.99, category: 'Tools' },
  { id: 2, name: 'Gadget', price: 19.99, category: 'Tools' },
];

const compressed = compressDataForPrompt(products);
// Use directly in your prompt template

Create XRON Messages

import { createXRONMessage } from 'xron-vercel';
import { generateText } from 'ai';

const systemMsg = createXRONMessage('system', products, { level: 2 });

const { text } = await generateText({
  model: yourModel,
  messages: [
    systemMsg,
    { role: 'user', content: 'Which product is cheapest?' },
  ],
});

Middleware

Automatically compress JSON array content in messages:

import { xronMiddleware } from 'xron-vercel';

const messages = [
  { role: 'system', content: JSON.stringify(largeDataset) },
  { role: 'user', content: 'Summarise this data' },
];

const compressed = xronMiddleware(messages);
// The system message content is now XRON-encoded

API

compressDataForPrompt(data, options?)

Compress an array into an XRON string for prompt injection.

createXRONMessage(role, data, options?)

Create a Vercel AI SDK compatible message with XRON-compressed content.

xronMiddleware(messages, options?)

Intercept messages and compress any JSON array content into XRON format.

Options

• level: Compression level — 1, 2, 3, or 'auto' (default: 'auto')

Licence

MIT