I’m building a meal planning PickAxe and am planning to build a knowledge base of several hundred recipes uploaded in CSV format. I’d also like to have users of the PickAxe upload their own recipe base (also in CSV). Is there a way to have the PickAxe prioritize the users list ahead of the default knowledge base?
Your Pickaxe should prioritize end-user uploaded files by default as they are added to the context window rather than the KB.
To optimize your Pickaxe for end-user uploads, make sure to increase your maximum input length. This help post explains exactly how the end-user uploads work.
Thank you for the detail! I might prefer it to be turned into vectors for RAG lookup. If they upload several dozen / hundred recipes, it can search through the chunks for relevant ingredients. I’ll mess around and see what works best. Thanks again!
If the end-user upload file EXCEEDS the maximum user input, then it is turned into vector embeddings so that it can be read via RAG. That is the wonderful flexibility of the system.