usemoss/moss — reverse-engineered prompt

Build me a fast semantic search tool for AI agents, like a retrieval layer that apps can embed directly instead of calling a slow vector database. I want developers to be able to create an index, add plain text documents with ids and optional metadata, load that index, then ask natural language questions and get the best matching documents back really quickly.

It should support both semantic search and keyword matching, include simple metadata filters, and work from Python and TypeScript examples. Please include a small demo app where someone can paste support docs, search them, and see scores and response time. Also include a simple command line tool for creating, loading, and querying indexes.

The main goal is that it feels easy for people building chatbots, copilots, voice agents, or RAG apps. Keep the setup simple with project id and key style config, add clear README instructions, and include sample scripts for quick testing. Look up current docs online if you need to.