Download the full Project files for this project at my Patreon along with 250+ other projects:www.patreon.com/posts/add-memories-to-109823010 Special Live coding event August 10th: www.patreon.com/posts/special-event-2-109557463 Learn to code fast with AI assistance with my 1000x MasterClass: www.patreon.com/posts/1000x-dev-103326330\ Discounts for my Patrons: www.patreon.com/posts/announcing-perks-108997398 Search 200+ echohive videos and code download lnks:www.echohive.live/ Auto Streamer: www.autostreamer.live/ Fastapi course: www.patreon.com/posts/learn-fastapi-26-95041684 Chat with us on Discord: discord.gg/PPxTP3Cs3G Follow on twitter(X) : twitter.com/hive_echo
I haven't quite figured out a good method of using images in memory. In context Windows it isn't able to refer to the base 64 images that are sent every time. So you end up having to create an index of sort and resubmit each one that you want to keep in conversation even though it's available in context. So that means then there's another portion of the class that has to know that you want to continue talking about the images that you've shared and then only refer to the ones that you're talking about but then processing all of the ones that you are sharing across all the context. Then there's a possibility of creating a method of removing images and context. I swear everything in a database so instead of storing the base 64 translated image for each message sent in you instead create an index of sorts and you refer to the image and then load the base 64 image into the message each time which would mean multiple in one or all messages but I haven't found a great way of keeping everything in there and then possibly mentioning a photo by index number or something more natural so that you don't have to step outside of the conversational tone and instead make it to where the assistant itself can figure out what you're talking about after processing all the images each time. It's more about something like building a memory system that could potentially see people that you've labeled in a certain way and then know who it is you're talking about without having to refer to them by name to pull out of memory information about the individual by name. That I can do it's the constant input of images and then relating that to data about labeled people and things within images. I've been thinking through this problem for quite a while and memory has tended to be the hardest part about these systems. I do appreciate all the work you're doing and can't wait to be able to get back to 12-hour coding sessions of my own. Everything's been a little too busy for me.
Will take a look at it. You can also check out the Agent AGI I built: Agent AGI builds its own tools, takes multiple actions, improves itself ua-cam.com/video/apXWaJ7ruyQ/v-deo.html
Download the full Project files for this project at my Patreon along with 250+ other projects:www.patreon.com/posts/add-memories-to-109823010
Special Live coding event August 10th: www.patreon.com/posts/special-event-2-109557463
Learn to code fast with AI assistance with my 1000x MasterClass: www.patreon.com/posts/1000x-dev-103326330\
Discounts for my Patrons: www.patreon.com/posts/announcing-perks-108997398
Search 200+ echohive videos and code download lnks:www.echohive.live/
Auto Streamer: www.autostreamer.live/
Fastapi course: www.patreon.com/posts/learn-fastapi-26-95041684
Chat with us on Discord: discord.gg/PPxTP3Cs3G
Follow on twitter(X) : twitter.com/hive_echo
I haven't quite figured out a good method of using images in memory. In context Windows it isn't able to refer to the base 64 images that are sent every time. So you end up having to create an index of sort and resubmit each one that you want to keep in conversation even though it's available in context. So that means then there's another portion of the class that has to know that you want to continue talking about the images that you've shared and then only refer to the ones that you're talking about but then processing all of the ones that you are sharing across all the context. Then there's a possibility of creating a method of removing images and context. I swear everything in a database so instead of storing the base 64 translated image for each message sent in you instead create an index of sorts and you refer to the image and then load the base 64 image into the message each time which would mean multiple in one or all messages but I haven't found a great way of keeping everything in there and then possibly mentioning a photo by index number or something more natural so that you don't have to step outside of the conversational tone and instead make it to where the assistant itself can figure out what you're talking about after processing all the images each time. It's more about something like building a memory system that could potentially see people that you've labeled in a certain way and then know who it is you're talking about without having to refer to them by name to pull out of memory information about the individual by name. That I can do it's the constant input of images and then relating that to data about labeled people and things within images. I've been thinking through this problem for quite a while and memory has tended to be the hardest part about these systems. I do appreciate all the work you're doing and can't wait to be able to get back to 12-hour coding sessions of my own. Everything's been a little too busy for me.
Can you build a script similar to agent zero?
Will take a look at it. You can also check out the Agent AGI I built: Agent AGI builds its own tools, takes multiple actions, improves itself
ua-cam.com/video/apXWaJ7ruyQ/v-deo.html
hi thank you for making this ,this is quite useful and informative can you access to the github repo link?
Thank you 🙏 I am glad you find it useful. The project files are available to my Patreon supporters. The link is in the description.