One-sentence headline summary

The Lightfeed Extractor is a TypeScript library that leverages LLMs and Playwright to automate robust, structured web data extraction for production-grade data pipelines and retail intelligence.

Key points

• Integrates with Playwright to support local, serverless, and remote browser environments with built-in anti-bot and proxy configurations.
• Uses Zod schemas to enforce structured data output, featuring a "JSON recovery" utility to sanitize and repair failed or malformed LLM responses.
• Converts complex HTML into LLM-ready markdown, with options to clean URLs, remove tracking parameters, and isolate main content.
• Pairs with the @lightfeed/browser-agent to enable AI-driven navigation, allowing for complex interactions like searching and pagination.
• Compatible with major LLM providers via LangChain, including OpenAI, Google Gemini, Anthropic, and local models via Ollama.

This library simplifies the development of reliable web scraping pipelines by combining AI-driven data parsing with resilient browser automation. It provides developers with a production-ready toolkit to handle common extraction challenges like dynamic content, schema validation, and anti-bot detection.