When you were using PPQ, were you getting errors about context? Or how did you diagnose that PPQ wasn't working well with large context? What tool were you using? Cline? I think you will find that Claude Code rate limits you once you burn a certain amount, but let me know.

The error was paying 20-30 cents per query late into a chat and >$30/month. It was just an economic decision. PPQ actually performs fairly well with the large contexts (over the API). I think the document search could be better. You only get 3 results, and it overall feels limited. Claude Projects are insane with how much data you can add in as files, and works just like search-based deep research, but with the documents. I only use PPQ online or from Open WebUI. And I only use Claude on the web app. Can't trust the AI on the bare metal.