Finally a good tutorial on AG studio! Next suggested content: how to integrate AGS in a frontend able to manage input/output like text and files (pdf, images). For example: external frontend > input (text) >> Autogen Studio flow >>> external frontend >>>> output (downloadable pdf with text and images inside). #AutoGenStudioChallenge
This is a really useful guide! There are also several excellent substitute tools available for anyone who would like to investigate more possibilities for creating AI bots.
Great content brother! You earned my SUB and thumbs up... I'm obsessed with AG. With the UI, it really helps me not having hardly any coding skills. Game changer. Thanks bub!
Hello, while creating a model I'm facing this error Error occurred while testing model: Error code: 404 - {'error': {'message': 'The model `gpt-4-1106-preview` does not exist or you do not have access to it.', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}} Please help me to fix this issue
isn't the purpose of the groupchat_assistant to decide the order? I think you need to write a prompt in the system message to describe the workflow and how the agents work together which includes the order.
Yes you can use the system message to give the logic which groupchat_assistant will follow to select the agents, the problem is when an agent can't fulfill its job because the code it has generated require a certain package to be downloaded or other issues. This forces groupchat_assistant of going back and forth between agents. So unless you're going to lay out the whole logic in the system Message it's better not to fill it with a strict workflow to follow. in general there are 3 components that help the groupchat_assistant in selecting agents and laying out the logic. 1. speaker selection method: this will define the way different agents are going to be called. 2. system message: as you said. 3. the context of the question itself since groupchat itself has a model that can understand and put a plan in place, this is what I've used in the video.
This is a nice video and very useful. In my applications I'm looking for the 'system' to have ALL the customer pdF Invoices uploaded, or better yet, as an SalesOrder table in a database. this seems like a lot of work for just one customer and one email. Is there a way to create Agents that could filter out which customer order? Etc.
Thanks, really useful video but I had to laugh at the end. For a second I thought you might recognise the huge risk posed by the "as a token of our gratitude" sentence, but you totally brushed it off because "it didn't offer any actual coupons"... you really missed an opportunity to acknowledge the limitations, stress the importance of testing + refining prompts and give any words of caution around applying this in mission-critical scenarios. The email was also way too long 😉 I'm sure I don't need to say this but I hope you revoked that API key before uploading this video...
You know Email marketing is a $7Bn industry. But the point is not what you're getting as an output. The idea is to understand how agents interact with with each other and how UserProxy and ChatManager send and receive tasks. There were 2 different advanced use cases I wanted to show but creating them took too much time therefore I couldn't include them in an introductory video. Also we always receive Thank You emails after a purchase, and this use case if connected to a database can be a great way to upsell a customer. So if you think about it ain't that crappy after all.
Finally a good tutorial on AG studio! Next suggested content: how to integrate AGS in a frontend able to manage input/output like text and files (pdf, images). For example:
external frontend > input (text) >> Autogen Studio flow >>> external frontend >>>> output (downloadable pdf with text and images inside).
#AutoGenStudioChallenge
Hi Reda, this tutorial is really quite useful. I finally understood all those sections in the docs using a real life example. Many thanks!
Superbe Mr Marzouk ! Merci.
You're most welcome 🙏🙏.
Et de rien.
As always, Bravo! Really interesting !
Thanks for sharing.
Thank you so much Amine, appreciate the kind words!!
This is a really useful guide! There are also several excellent substitute tools available for anyone who would like to investigate more possibilities for creating AI bots.
Great content brother! You earned my SUB and thumbs up... I'm obsessed with AG. With the UI, it really helps me not having hardly any coding skills. Game changer. Thanks bub!
Thank you, it really means a lot 🙏🙏
great video as well as the troubleshooting
Appreciate it 🙏🙏
Hello, while creating a model I'm facing this error
Error occurred while testing model: Error code: 404 - {'error': {'message': 'The model `gpt-4-1106-preview` does not exist or you do not have access to it.', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}
Please help me to fix this issue
Bravo well done incredible video I now understand auto-gen thanks for this🎉🎉🎉
Thank you for the appreciation 🙏
isn't the purpose of the groupchat_assistant to decide the order? I think you need to write a prompt in the system message to describe the workflow and how the agents work together which includes the order.
Yes you can use the system message to give the logic which groupchat_assistant will follow to select the agents, the problem is when an agent can't fulfill its job because the code it has generated require a certain package to be downloaded or other issues. This forces groupchat_assistant of going back and forth between agents.
So unless you're going to lay out the whole logic in the system Message it's better not to fill it with a strict workflow to follow.
in general there are 3 components that help the groupchat_assistant in selecting agents and laying out the logic.
1. speaker selection method: this will define the way different agents are going to be called.
2. system message: as you said.
3. the context of the question itself since groupchat itself has a model that can understand and put a plan in place, this is what I've used in the video.
I test it only with userproxy and agent but still it sometimes stays unresponsible state
You did a great video. How do you use Azure OpenAI credentials? Azure needs more than just the API key to work.
This is a nice video and very useful. In my applications I'm looking for the 'system' to have ALL the customer pdF Invoices uploaded, or better yet, as an SalesOrder table in a database. this seems like a lot of work for just one customer and one email. Is there a way to create Agents that could filter out which customer order? Etc.
I like waht you did, but for no code people this is so hard because we dont know what we should install for windows etc.. really really nice video
You're most welcome!
Hi , Very good content .. however I am not able to see group chat agent while creating new workflow.. can you help me on that
I think you just need to restart Visual Studio inorder for the Autogen to pickup the OPEN_API_KEY environment variable
How to use or connect this to a real live application. meaning how to publish or call the agent from a live app?
Dear sir, can I use my local LLM models instead of OpenAI API?
Yes you can, I did myself but I think as of now they simply don't perform as well.
microsoft.github.io/autogen/blog/2023/07/14/Local-LLMs/
@@redamarzouk thank you sir. I switched to CrewAI for convenience.
what about langgraph for agents, any thoughts on it
how to enable groupchat agent
Thanks, really useful video but I had to laugh at the end. For a second I thought you might recognise the huge risk posed by the "as a token of our gratitude" sentence, but you totally brushed it off because "it didn't offer any actual coupons"... you really missed an opportunity to acknowledge the limitations, stress the importance of testing + refining prompts and give any words of caution around applying this in mission-critical scenarios. The email was also way too long 😉
I'm sure I don't need to say this but I hope you revoked that API key before uploading this video...
19:40 "it stayed in line and did not hallucinate" - I swear you AI bros are starting to make me think I'm hallucinating. For real? Check yourself.
I do not how to make it local not using openai keys no llm studio, just oobabooga self host
Do you do not explain alot..
You mean I explain lot or
I don't explain enough?
Using AI to send out crappy emails seems like such a waste of what AI is and will be.
You know Email marketing is a $7Bn industry. But the point is not what you're getting as an output. The idea is to understand how agents interact with with each other and how UserProxy and ChatManager send and receive tasks.
There were 2 different advanced use cases I wanted to show but creating them took too much time therefore I couldn't include them in an introductory video.
Also we always receive Thank You emails after a purchase, and this use case if connected to a database can be a great way to upsell a customer.
So if you think about it ain't that crappy after all.