supermemoryai/supermemory ? reverse-engineered prompt

I want you to build a memory and context engine for AI. The core idea is that AI assistants forget everything between conversations, and this system should fix that. It needs to automatically learn from conversations, extract key facts, build user profiles, and deliver the right context at the right time.

It should have two main parts. First, a polished web app for end-users. People can sign up, and the app will give their AI (like Claude or in their VS Code) a persistent memory. It should also be able to connect to their Google Drive, Notion, etc., to pull in knowledge.

Second, this needs to be a full platform for developers. Create a simple API (with both npm and Python packages) so other programmers can easily add memory, RAG, and user profiles to their own agents without worrying about vector databases or chunking strategies.

For the stack, use TypeScript with Remix and Tailwind for the front-end app. The backend should be super fast and scalable, so let's build it on something like Cloudflare Workers and KV.