Add file upload feature
complete
MindMac
Add feature to allow user upload files (pdf, doc, docx, txt, …) then ask questions about these documents.
Update:
- 7 March 2024: File upload now is in beta phase and available from the version 1.9.0, please check it out and give your feedback. Thank you all!
MindMac
complete
Closed as the file upload feature is already on the app. Please share your feedback and suggestions to make this feature better. Thank you all for your support.
Shawn Ashman
Really cool I can wait to give it a spin :) thanks a lot to the MindMac team !!
g
gerbull001
MindMac Great to hear - where do the files get uploaded to?
MindMac
gerbull001: Files will be split into chunks and uploaded to OpenAI/GoogleAI/MistralAI to obtain embedding vectors. If you prefer to keep your files locally, you can configure the system to utilize the Ollama model as a document analyzer.
g
gerbull001
MindMac Amazing, I look forward to using it. Thank you
BP
MindMac is there a way to pretokenize the input files so that they can be used later? I was thinking of having phi-2 split up a document into logical parts of 250 words each and then loading it into MindMac and using Ollama to answer questions about the document. A Hacker News user describe this workflow here: https://news.ycombinator.com/item?id=39614273 .... Or just thinking through this a little more, maybe add a way to edit the prompt for the embedding model? That way you could control how the file is broken up.
MindMac
BP: Thanks for your suggestion. We will consider to add feature that allows users connect external vector databases, then ask about these data.
MindMac
Finally, we're excited to announce the addition of a document upload feature in the version 1.9.0, currently in beta. This allows you to upload files and inquire about their content. It seamlessly supports PDF, plaintext, and some code file formats. Preliminary testing with Microsoft Office files (Word, PowerPoint, Excel) has shown promise, and we encourage you to provide feedback on this functionality. Please upgrade on your end. Thank you all!
S
Sanjeev
MindMac I am primarily using to for pdfs, and feel it could improve a bit. Many times, my response is that the information cannot be found in the chunks provided. So it would be great to be able to inspect the full chunk that was fed into the LLM.
F
Flyable7855
cannot wait :)
MindMac
Flyable7855: Very soon :D, I'm testing the feature.
D
Davo
+1!
Christopher Davis
More and more each day, this becomes a hindrance to my work with MindMac. Please bump this up on your priority list. Thank you.
g
gerbull001
This will be amazing once this is done - Gino Goossens: same - move away from GPT+. API RAG is more accurate in my experience. MindMac: Presumably this will be an integration with the openAI assistants API?
O
Oscar Neto
yeah - such important feature!
MindMac
in progress
C
Carlos
MindMac: ❤️
G
Gino Goossens
MindMac: Yes! Can't wait so I can completely move away from GPT+. Thanks!
Akhilesh Sabharwal
MindMac: ❤️
Load More
→