Open grayerbeard opened 2 months ago
SystemSculpt
commented
2 months ago Yep, I have YouTube video transcription in the works as part of the upcoming new Data module, will have that + Omnivore + Readwise + Reader integration to start with, maybe Twitter too if I have time to add that before I release the initial version of it :)
grayerbeard
commented
2 months ago I think that is great and there is just one request I have just thought of…..
If you could figure out how to answer a question about your plug in using AI from within Obsidian.. “How do I …”, “I want to ….. but instead … how do I fix” etc. You could build in an attempt at this and ask users to tell you if they don’t get a good answer so then make it self improve. Its an important step towards the totally intuitive AI where learning how to set up a plug in is totally painless!! (Would be a great help for my decaying 74 year old brain!!)
By the way I got the local LLM working really well yesterday, I was using Llama 3 and it was really fast. My only difficulty was issues with context windows and long questions getting short answers. When using OpenAI you have a context window setting. In LM Studio I used to be able to mess with that and some Models had big windows – you have done a lot of testing so at some point some sort of write up – or a remake of some of your old videos – looking at how to tackle your Plug in plus LM Studio – which Models best for what etc would be great (and how to tweak) . – A future wish though you need to get the functionality going first.
Regards
David Torrens
Grayerbeard
Shropshire UK
From: SystemSculpt @.> Sent: Sunday, April 28, 2024 2:02 AM To: SystemSculpt/obsidian-systemsculpt-ai @.> Cc: David Torrens @.>; Author @.> Subject: Re: [SystemSculpt/obsidian-systemsculpt-ai] AI Using Template Suggestion (Issue #4)
Yep, I have YouTube video transcription in the works as part of the upcoming new Data module, will have that + Omnivore + Readwise + Reader integration to start with, maybe Twitter too if I have time to add that before I release the initial version of it :)
— Reply to this email directly, view it on GitHub https://github.com/SystemSculpt/obsidian-systemsculpt-ai/issues/4#issuecomment-2081271424 , or unsubscribe https://github.com/notifications/unsubscribe-auth/ADUCWEMQV6BARLTRWAT2Q2DY7RDBHAVCNFSM6AAAAABGW7TSS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAOBRGI3TCNBSGQ . You are receiving this because you authored the thread. https://github.com/notifications/beacon/ADUCWEMTCYE5MHUPCYSCD3DY7RDBHA5CNFSM6AAAAABGW7TSS6WGG33NNVSW45C7OR4XAZNMJFZXG5LFINXW23LFNZ2KUY3PNVWWK3TUL5UWJTT4BWXIA.gif Message ID: @. @.> >
I used my fabric_get_instructions template to get instructions from a video. Worked OK with this procedure;
Result ; I now have a Note File With the required result at the end but still have the template at start.
SUGGESTION: Would prefer to somehow do something more like this:
Put the YouTube transcript in a Note and enter its File Name (This could also be some other prompt for AI) Enter somewhere what Template to use Then Just Hit a Hotkey and the AI would generate a new file based on the Transcript and the Template, with an automaticly generated title.
I suspect you maybe working on something like this but I feel current procedure is improvable..
Template Used: fabric_get_instructions.md
Result: Setting Up SystemSculpt AI with BRAT on Obsidian A Step-by-Step Tutorial.md