Fantastic tutorial. You have a very clear teaching style. I think this is sort of long term memory and agentic learning is still kind of flakey but this is a pretty solid solution. I'd love to see more videos around agents/memory/tools etc from you. Best, Jacob
Amazing content! I think this is the best teachable agent overview I've found. I really appreciate the walking through the code. Its very informative but also reminds me as an experienced developer that this is all just code at the end of the day, and I know how to read it 😅. Please keep it up!!! 🙏🙏
hey thats a really great and well explained video! I really like to try how its working. Do you have an idea how to split the chat, in a chatlog, where the complete dialog is saved and a chat interface, where you di not see the system and the internal dialog, only the messages between user and bot?
How do we track or load the chat history for a user specific? Like for example, let us say if the chat prompt is opened and used by me as a user in my system, then whatever I interact with the Agents then AutoGen Teachable agent should load the chat history only with respect of me in the window.
I appreciate you're taking a deeper dive into Autogen's Teachable Agent. Thank you! Using chromadb, can we assume documents can be loaded into the db for future reference by a teachable agent?
You are welcome! That's a great question. While I haven't done exactly what you describe, I don't really see any reason that would not work. I'd have to do some testing to confirm that - might be a good topic for a future video!
10 месяцев назад
Andy this really great! I am trying now to integrate the teachable agent in a team of agents controlled by an admin agent. I would like to use panel to have a minimal ui for the bot. Do You have any idea? All the best
Thank you! That sounds like a fun project, but unfortunately I am not familiar with Panel. I will say that I've created some Blazor interfaces around the Open AI API itself, and that was a great learning experience. I'd be excited to see what you come up with, please keep me in the loop!
Any idea how scalable this is compared to memgpt? I wonder what the size and performance is of this after a few months of use as it remembers more and more facts. Anyone have any guesses based on the architecture?
Good question, I'm not entirely sure. Please keep in mind that this is based on no evidence at all, but I'd have to imagine they'd be at least somewhat equivalent; they both basically vectorize important "memories" that they identify, and store them locally. I have absolutely stored entire directories of PDFs in a vector db, and while it took quite a while to process the documents, retrieval never seemed to lag much. If you did something similar, but in the context of a conversation, I'm not sure you'd notice much of a performance hit as time went by. But with that said, I have not had a long running conversation like you're describing, so it's hard to really be sure.
Fantastic tutorial. You have a very clear teaching style. I think this is sort of long term memory and agentic learning is still kind of flakey but this is a pretty solid solution. I'd love to see more videos around agents/memory/tools etc from you. Best, Jacob
Good video, thank you!
Thank you for making these videos despite the health struggles 🙏
You are welcome!
Thanks, this makes things very practical in my mind.
You are welcome! I'm glad it was helpful!
Thanks for sharing this! Easy to follow great content in these exciting times :)
You're welcome! Thank you for the kind words. Things certainly are getting exciting!
Amazing content! I think this is the best teachable agent overview I've found. I really appreciate the walking through the code. Its very informative but also reminds me as an experienced developer that this is all just code at the end of the day, and I know how to read it 😅. Please keep it up!!! 🙏🙏
Wow what a great compliment, thank you so much! I'm glad this was helpful!
awesome. best tutorial out there.
Thank you!!
You are welcome!
hey thats a really great and well explained video! I really like to try how its working. Do you have an idea how to split the chat, in a chatlog, where the complete dialog is saved and a chat interface, where you di not see the system and the internal dialog, only the messages between user and bot?
How do we track or load the chat history for a user specific? Like for example, let us say if the chat prompt is opened and used by me as a user in my system, then whatever I interact with the Agents then AutoGen Teachable agent should load the chat history only with respect of me in the window.
I appreciate you're taking a deeper dive into Autogen's Teachable Agent. Thank you!
Using chromadb, can we assume documents can be loaded into the db for future reference by a teachable agent?
You are welcome!
That's a great question. While I haven't done exactly what you describe, I don't really see any reason that would not work. I'd have to do some testing to confirm that - might be a good topic for a future video!
Andy this really great! I am trying now to integrate the teachable agent in a team of agents controlled by an admin agent. I would like to use panel to have a minimal ui for the bot. Do You have any idea? All the best
Thank you! That sounds like a fun project, but unfortunately I am not familiar with Panel. I will say that I've created some Blazor interfaces around the Open AI API itself, and that was a great learning experience. I'd be excited to see what you come up with, please keep me in the loop!
Any idea how scalable this is compared to memgpt? I wonder what the size and performance is of this after a few months of use as it remembers more and more facts. Anyone have any guesses based on the architecture?
Good question, I'm not entirely sure. Please keep in mind that this is based on no evidence at all, but I'd have to imagine they'd be at least somewhat equivalent; they both basically vectorize important "memories" that they identify, and store them locally. I have absolutely stored entire directories of PDFs in a vector db, and while it took quite a while to process the documents, retrieval never seemed to lag much. If you did something similar, but in the context of a conversation, I'm not sure you'd notice much of a performance hit as time went by. But with that said, I have not had a long running conversation like you're describing, so it's hard to really be sure.