Building Agents for Real-World Applications with CrewAI & AutoGen
This is a comprehensive course that teaches you how to leverage AI-powered agents to automate complex tasks. From meeting preparation and online research to customer support and trip planning, you’ll learn to build autonomous agents for various real-world use cases using CrewAI and Autogen Studio.
Lesson 1
Build Advanced Multi-Agent Systems with Crew AI
Discover how Crew AI serves as an exciting alternative to Microsoft's Autogen as we show you how to set up your own agent crews quickly and efficiently. Imagine creating a scenario where Donald Trump and John McClane team up to solve a superhero mystery at Trump Tower using a multi-agent team that taps into the Marvel Superhero API.
In this video, you'll learn:
- The key differences between Crew AI and Microsoft's Autogen
- How to set up agent definitions with specific roles, goals, and backstories
- How to integrate external APIs, like the Marvel Superhero API, into your agents
- How to assemble your agent crew and manage tasks hierarchically
- Tips on reducing code complexity and achieving the same results with less code
Lesson 3
Revolutionize Your Online Research and Writing with Crew AI
Follow along step-by-step as I use Crew AI and DuckDuckGo Search to transform your approach to online research, design a beautiful UI with Streamlit, and automatically generate captivating articles.
In this video, you'll learn:
- How to set up a virtual environment and install Crew AI and DuckDuckGo Search libraries
- How to create and configure AI agents for research and writing tasks using Crew AI
- How to automate the discovery of the latest trends on any topic
- How to design an interactive user interface with Streamlit
- How to generate well-written, informative articles in seconds
- Tips for efficiently utilizing AI agents to accelerate your research and writing processes
Lesson 4
Automate Pizza Ordering with AI
Discover how to use artificial intelligence to transform the way pizza orders are processed and delivered, saving costs and enhancing the customer experience. In this video, I'll show you how to utilize Crew AI for autonomous agent management, convert speech to text accurately with OpenAI's Whisper, and interact with SQL databases for real-time data handling.
In this video, you'll learn:
- How to create and configure agents for order reception and processing
- How to use OpenAI's Whisper to transcribe customer audio orders
- How to interact with a SQLite database to check ingredient availability
- How to design a user-friendly interface with Streamlit
- How to record and process audio orders directly from the microphone
Full Video & Source Code
Lesson 5
Automate Your Trip Planning with AI
Discover how to harness the power of Crew AI to automate your trip planning, saving you hours of research and scheduling. In this video, I'll show you how to design and implement agents and tasks tailored for comprehensive trip planning from scratch. Curious about how to simplify your travel planning with the latest AI technology? Let's dive in and discover together.
In this video, you'll learn:
- How to define agents and tasks for trip planning
- How to create custom tasks for identifying the best city, gathering in-depth city guides, and planning detailed itineraries
- How to implement agents with specific roles, goals, and backstories, including a city selection expert, local expert, and travel concierge
- How to integrate web search capabilities into your agents using tools like SERP API
- How to run the Crew AI process and generate a comprehensive travel plan
Lesson 6
Automate Email Filtering with AI
Tired of manually sorting through your inbox to find important emails? In this video, I'll show you how to automate your email management using Crew AI, Langchain, and LangGraph. We'll build an AI agent that filters out emails that need responses from those that don't, saving you time and boosting your productivity.
In this video, you'll learn:
- How to create an AI agent with a specific role and goal for email filtering using Crew AI
- How to connect your AI agent with Gmail by setting up credentials and accessing the Gmail API
- How to fetch emails from your inbox and provide essential information to the agent for processing
- How to assemble your agent and tasks into a crew and initiate the email filtering process
- Tips on authenticating with Gmail and securely handling API credentials
Lesson 7
Automate Email Responses with AI
Ever feel overwhelmed by emails that need immediate attention? In this video, I'll show you how to extend our email automation system using Crew AI, Langchain, and the Gmail API. We'll create new agents that not only identify urgent emails but also draft personalized responses, making your email management more efficient than ever.
In this video, you'll learn:
- How to set up an Email Action Specialist agent to identify emails requiring immediate action
- How to integrate essential tools like Gmail Get Thread and Tavily Search from Langchain community tools
- How to create an Email Writing Agent that drafts clear and effective responses
- How to implement tasks for analyzing entire email threads and drafting personalized replies
Lesson 8
Automate Your Inbox with AI
Looking to automate your email workflow and save valuable time? In this video, I'll show you how to develop an automation system that checks your inbox regularly for new emails, waits if none are found, and answers emails based on the complete thread history in your tone and voice, saving them in Gmail as drafts.
In this video, you'll learn:
- How to implement individual nodes for your graph in a new
nodes.pyfile - How to create a
check_emailsmethod to process and track emails - How to identify unreviewed emails by verifying specific conditions
- How to define conditional logic to handle new emails or wait intervals
- How to integrate your nodes logic into
app.pyand manage state within your application - How to construct a state graph and outline your workflow
- How to compile and visualize the workflow using the
compilemethod and console output - How to adjust your logic to process only new emails moving forward
- How to test the system by processing emails and verifying automated drafts in Gmail
Lesson 9
Run Autogen Agents for Free: Avoid Pitfalls and Boost Performance
Ever wanted to run Autogen agents without incurring high daily costs? In this video, I'll demonstrate methods to run Autogen agents for free and share insights on how to avoid common pitfalls. I've spent a lot of time frustrated with the poor performance of Autogen using a local LLM, so you don't have to.
In this video, you'll learn:
- How to set up Autogen agents to run locally for free
- Tips to avoid common pitfalls when using Autogen with local LLMs
- The capabilities of Autogen applications in automating tasks like customer support
- How to build autonomous agents to manage tasks such as responding to chats and emails
- Step-by-step guidance on setting up everything from scratch to run Autogen efficiently
Lesson 10
Automate Workflows with Llama2 and Autogen Studio
Want to set up Autogen agents to run locally and optimize your AI workflows? In this video, I'll guide you through configuring Autogen Studio with a local Llama2 model, and show how to create autonomous agents that interact in a group chat without writing any code.
In this video, you'll learn:
- How to configure Autogen Studio to work with local models like Llama2
- How to install LM Studio and start a local server for the Llama2 model
- How to create and configure agents with unique system messages and assign them to the local model
- How to set up a group chat workflow between multiple agents
- How to review the communication between Autogen Studio and LM Studio via the console output
Full Video & Source Code
Lesson 11
Master Autogen Skills
Curious how to create and assign skills to agents in Autogen Studio? In this video, I'll demonstrate how to start from a clean state, build an inventory skill from scratch, and use it in a group chat with autonomous agents to answer customer queries.
In this video, you'll learn:
- How to reset Autogen Studio by deleting the SQLite database file for a fresh start
- How to create a new skill that accesses inventory data using a custom script
- How to assign the inventory skill to an agent in a simple chat setup
- How to create a group chat with an inventory manager agent and a customer support agent
- How to configure the workflow so the customer support agent responds based on information provided by the inventory manager
- How to test the workflow with a customer inquiry and review real-time responses
Lesson 12
Creating a Children's Book with GPT-4o Autogen Studio
Excited about the potential of Autogen Agents to create a complete picture book from scratch? In this video, I'll walk you through the process of replicating a Microsoft Autogen Studio example where agents generate both text and images, combining them into a PDF for a children's book.
In this video, you'll learn:
- How to install and start Autogen Studio to work with local agents
- How to set up a GPT-4.0 model in Autogen Studio using your API key
- How to create an agent capable of generating images and PDFs using built-in skills
- How to design a workflow with a user proxy and book agent to generate content for a picture book
- How to create an example image and description using Autogen's image generation skills
- How to refine your agent’s outputs to generate a multi-page PDF with both images and text
- How to troubleshoot common issues, such as missing images in PDFs or incomplete output
