Native RAG Support
complete
A
Ali
Simplified workflow to create RAGs (knowledge stacks) and use them.
I would like to be able to create a knowledge stack (RAG) using TypingMind interface. I would specify a local folder or local PDFs on my computer and typingmind would use either openAI's embedding model or a local embedding model to create a RAG. Then, when creating a chat, I would be able to call upon a local knowledge stack.
knowledge stack and files should be local since knowledge stack can contain large files
All should be done through typingmind's UI without using endpoints etc. Currently, typingmind supports RAGs, but it's through a complicated API. It should all be done locally using the typingmind interface.
It shouldn't just input the whole document as input. That would consume a lot of tokens like in agents system instructions.
N
Ngoc Nguyen
complete
this is now available! https://x.com/TypingMindApp/status/1902006304484540753
J
Jpsala
I'm waiting for this! Nice!
N
Nathalie Salles
That update would be a dream. I have over 1,500 articles for my dissertations that I cannot process and this RAG could be a life savior for my work
N
Ngoc Nguyen
planned
Aitan Roubini
I’ve been trying to figure out if I could simulate this by writing (or get the AI to write) a myphpadmin style plugin to create MySQL databases to draw from and write to. That way I could even leverage my own LAMP server.
Obviously a built in RAG would be 1000% preferable, but much like everyone here I’m hungry for a robust and advanced memory function.
I’ve been trying so many different (mostly process-based) approaches to having TM retain core processes and directives outside of the System Instruction, and none of them would work as well as a built-in RAG function.
Take a marketing plan for example. It’s highly changeable and something that the system instruction needs to interact with, and adhere to.
It literally would be the one killer feature that takes this already indispensable application and makes it a gold-standard.
Is cost the issue? Would creating this feature cause TM’s storage footprint to skyrocket? I’m already saving all of my chats, so how much more could a RAG function eat up?
Worth noting though is that I have to look elsewhere to see if there’s a reasonably priced and safe solution that can be plugged into TM. I’d prefer not to have to pay extra, but would consider it if it was reasonable. Either than or utilize some other cloud infrastructure that I’m already paying for.
R
Robert coorey
If there was a WYSIWYG RAG editor inside TypingMind as well where you could update the RAG live that would be great.
Diego Sala
This would be a hugely useful update for me