swapnil18800/weblens — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me WebLens, an AI web research assistant that answers questions with evidence people can check. I want a simple web app where a user types a question, sees the assistant start responding quickly, and can expand a reasoning trace that shows how the question was understood, what searches ran, what pages were read, and which passages supported the final answer.

The answer should include clear numbered citations, retrieved passage previews, source links, tags, and suggested follow up questions. It should handle easy factual or math questions without doing a web search, but use live web research for anything current or complex. Please include caching so repeated questions can reload previous answers and reasoning traces.

Use the existing Python backend and React frontend approach from this repo, with environment variables for search, database, tracing, and model keys. Add setup instructions so I can run it locally and deploy it. Look up current docs online if you need to.

Want more depth? Deep Reverse