AutoGen RAG Multi-Agents: How I created Multiple AI Agents to Respond Accurately

🚀 Dive deep into the world of AI with our latest tutorial on creating an advanced Group Chat Manager using RAG (Retrieval-Augmented Generation) and AutoGen! In this comprehensive guide, we'll explore how to harness the power of multiple AI agents, including a Boss, Python Engineer, Project Manager, and Code Reviewer, to work together seamlessly, providing context-aware responses and collaboration. 🤖✨AutoGen RAG Group Chat: Creating Multiple AI Agents to Respond Accurately
If you like this video:
Tweet something positive about these tutorials in / @mervinpraison
"@MervinPraison ......................."
Whether you're an AI enthusiast or a professional looking to integrate sophisticated AI functionalities into your projects, this video is your go-to resource. Learn about the two methods of invoking RAG: Retriever Agents and Function Calling, and see how to store embeddings in Chroma DB for enhanced performance. 🛠️🧠
🔑 Key Takeaways:
Understand RAG and its role in multi-agent AI systems.
Step-by-step instructions on setting up your AI agents for group chat management.
Insights into using Chroma DB for storing embeddings.
Practical examples of both Retriever Agents and Function Calling methods.
Don't forget to subscribe and hit the bell icon to stay updated on all things AI! 👍💡 Like this video to help spread knowledge and share with others who might find it useful. Let's embark on this AI journey together!
🔗 Resources and Documentation:
AutoGen Tutorial: • AutoGen Tutorial
Patreon: / mervinpraison
Ko-fi: ko-fi.com/mervinpraison
Discord: / discord
Twitter / X : / mervinpraison
Code: mer.vin/2024/02/autogen-rag-g...
Timestamps:
0:00 - Introduction to RAG in AutoGen
0:46 - What is RAG for Multi-Agents?
1:17 - How to Set Up Your AI Agents
2:23 - Creating the Group Chat Manager
3:29 - Comparing With and Without RAG Agents
4:47 - Running the Code: Without RAG vs With RAG
5:51 - Implementing RAG Using Function Calling
7:11 - Final Overview and Running the Function Calling RAG
8:03 - Conclusion and Call to Action
#AutoGenRAG #MultiAgents #GroupChat #Tutorial
#autogen #autogenTutorial #MicrosoftAutogen #AutogenMicrosoft #AIAgents #AI #CreateAIAgents #HowToInstallAutogen #AutoGen #AutogenQuickstart #HowToUseAutogen #AutogenInterface #AutogenAI #AutogenAssistant #AutonomousAgents #AutogenTutorials #AutogenInstallation #AutogenRAG #AutogenRAGMultiAgents #AutogenRAGAIAssistants #AIAssistants #RAGAIAgents #AutogenRAGAIAgent #AIAgent #AIAssistant #AutoGenRAG #AutogenRAGTutorial

[@squibtechnologies]

Keep it going, Mervin! I like how you showed multiple methods here, easy to follow.

[@mikew2883]

Amazing stuff Mervin! 👏

[@MervinPraison]

Thank you

[@JosefEngelfrost]

Thank you for your videos, to me they make a great complement to the AutoGen notebooks and other written sources. It's much easier to follow along if I've seen someone do it first. I would love to hear you elaborate on the difference between using function calling and RAG via retrieval agent. How do you experience the difference? Do you have use cases when you would prefer one over the other?

[@MervinPraison]

Thank you 😊 Simple easy difference is LLM Response for Function calling is JSON object LLM Response for RAG is Text Both are different

[@JosefEngelfrost]

@@MervinPraison Yes, they are in different format, but I'm more interested in how they work on a higher level, i.e. the impact function calling has on the group chat and the result vs providing RAG through a specific agent.

[@EddieAdolf]

Can this be done in Autogenstudio UI yet?

[@SaiTeja-go6lw]

Could you share video related to invoice processing documents for agent rag cases

[@MervinPraison]

Soon I will do an example use case, you could modify that based on your above requirement

[@akshaypachbudhe3319]

how to implement this RAG using local PDF documents instead of web URLs (if possible in ollama)??

[@RobOttly]

friend i ask of you why autogen when most sources report crewAi > autogen? Perhaps some reason led for you this conclusion? Looking to your channel in await for local RAG implementation 🤞 you make very many good video ❤ I wishing for crewAi advanced series with local RAG very much like this video but no chatgpt

[@MervinPraison]

Every project has its own advantages. It's best to understand what each tool offers at a high level so you can select the right one for your use case. For simple applications and if you're a beginner, CrewAI is a good choice. If you're an advanced user who needs more customization options, consider AutoGen. Your decision should also be based on the specific features you want in your application. For example, if you want to implement RAG (Retrieval-Augmented Generation) with your AI agents and CrewAI doesn't currently support it natively, AutoGen would be an easier choice.

[@JosefEngelfrost]

I don't think Mervin is "team autogen", he makes videos about both, and anything else that's interesting :D I've played around with a little of both and it's not as simple as one is better than the other. I like CrewAI better, but autogen is more mature and let's me do things CrewAI just doesn't support (yet). Regarding local RAG, I haven't tried it but it should be as simple as switching out GPT with e.g. Ollama. Just follow these intructions, and then add the instructions on how to switch GPT for a local LLM on top of it. If you try it, it would be interesting to hear if it works, and if the result is good.

[@RevMan001]

I found the same thing, I started with crewAI (it's simple for a quick crew using chatgpt4) but for more advanced integrations, it just doesn't seem evolved enough right now. Autogen seems to have a head start and with Microsoft behind it, it'll probably stay that way. I'm sure crewAI will get better, but I don't have time to wait. @@JosefEngelfrost

[@IlliaKniaziev]

Hi, is it possible to place couple of url's for rag use, in my case i have different books that i want to use, and in my case or i give all of book to all agents, or maybe predefine specific url to specific agent (idk), if its possible could you share some videos or maybe tell where to find info about it

[@IlliaKniaziev]

yes, im using autogen to automize this process

[@rs832]

Mervin, thank you for these videos! I would LOVE to see this for CrewAi -- any chance thats in the pipeline for the very near future?

[@MervinPraison]

Yes it’s in my pipeline :)

[@rs832]

@@MervinPraison This is great news! Is there someplace the pipeline can be viewed? If not, perhaps you can share when the video is expected to be released? Thank you - we are looking forward to this journey!

[@planplay5921]

I think the open source technology we have now is enough to start a company with good profits.

[@MervinPraison]

Yes

[@loicbaconnier9150]

If you think so why didn't use open source model in spite of openai in your example ? Will be more interresting with open source like mixtral. And with the limitation of their function calling

[@christopheboucher127]

Hi Mervin, do you think is possible to implement this Rag Function as skill in autogen studio ? You can just reply yes or no or maybe make a video ;) Thx a lot for all your precious video

[@MervinPraison]

Yes, Its possible. But CrewAI don’t provide that feature by default. I hope they will do it soon. But still you should be able to implement it using Tools in CrewAI. It’s in my Pipeline.

[@christopheboucher127]

CrewAI ? I was talking about autogen studio and skills:)

[@MervinPraison]

​@@christopheboucher127 Ops sorry for that. I've been receiving lot of comments about CrewAI and I mixed that here. "Do you think is possible to implement this Rag Function as skill in autogen studio ?" Yes. Sure, I will look in to it and will try to make a video on that.

[@RetiredVet]

Mervin, I may have missed the discussion in the video, but you import AssistantAgent and with coder you assign it: "coder = AssistantAgent" but with pm you use autogen.AssistantAgent: "pm = autogen.AssistantAgent" May I ask if there is any reason for doing this? Is there any difference?

[@MervinPraison]

You are absolutely correct. I should have used one or the other to stop confusion. Both are the same, based on how u import If you use import autogen Then you should use autogen.AssistantAgent If you use from autogen import AssistantAgent Then you should use AssistantAgent I did both without realising 🤦‍♂️ do one or the other to stop confusion. But it should work anyways.

[@RetiredVet]

@@MervinPraison That is what I thought, but when someone who knows so much more than I do about a language, I don't assume. Thanks for the response.

[@YeasinRafio]

Hello, can you please suggest me open source llm for agents, because most of them I heard are not good at using tools

[@rs832]

my experience with mistral 7b & mixtral 8x7b has been pretty solid, to the point where I haven't had the desire to look for others allowing me to focus on other tasks.

[@MervinPraison]

Yes I recommend Mixtral 8x7b

[@fortuneasy]

Marvin, in open IA key, can I use gemini api?

[@lokeshart3340]

Same here can we use something free like mistral

[@fortuneasy]

i will try by myself and back here@@lokeshart3340

[@MrJekyllDrHyde1]

I do not see the point of multi-agent AIs. The use case you presented can be done with regular RAGs with ONE verbose prompt. Why bother making different agents, tasks etc ?