Anthropic's Project Feature
planned
Daniel Tango
Spinning this up, a bit: Projects have a "system prompt" that is the same for all chats in that project (I think Perplexity has that, as well).
This could work the same in Bolt: projects contain documents and system prompts. (and RAG :))
Daniel Nguyen
planned
Yeah. Currently BoltAI does support document analysis (attach documents to a chat) but it's per chat.
Daniel Tango
Daniel Nguyen - that's not what DCHA is asking for. I doubt you can support what he wants, as this is a feature not of the Anthropic models, but of their web app, kind of like ChatGPT's code interpreter.
Daniel Nguyen
Daniel Tango Hmm I’m not sure how it is different from the current Document Analysis in BoltAI. As far as I know they don’t use RAG either. So the whole project needs to fit within 200k context.
Daniel Tango
Daniel Nguyen- I don't have a paid Claude account anymore, but maybe I am also mixing it up with Artefacts?
Daniel Nguyen
Daniel Tango ah right. Artifacts is a different feature and it’s trickier to implement in BoltAI. Projects are really just a template of system prompt and ingested documents