How does Docalysis.com work? I'm presuming you take the user input then search the documents for matching words to get snippets then add the snippets to a prompt (along with the original query) you send to an LLM.

What does the cost look like? Are you running your own llama.cpp or just using the ChatGPT API?