Photo by Author | Canva and Chat GPT
. Introduction
Got hub New programming has become a platform for the early people who want to learn languages, ideas and skills. With the growing interest in Aging AI, the platform is rapidly displaying real projects that focus on “agent workflows”, which makes it an ideal environment for learning and building.
Is a remarkable source Microsoft/AI-AGENTS-BIGHERSIn which a 12 lesson course covering the basic principles of construction of AI agents is presented. Each lesson is designed to stand on its own, which allows you to start at any point that is in line with your needs. It also offers multi -language support, which ensures wider access to learners. Each lesson of this course includes examples of code, which I can meet code_samples Folder
More that it uses the course Azure Ai Foundry And Gut Hub Model Catalog To communicate with language models. This includes numerous AI agent framework and services like Azure AI Agent ServiceFor, for, for,. Cementic kernelAnd Autogen.
We will review each lesson in detail to facilitate your decision -making process and provide a clear overview of what you will learn. This guide acts as a helpful source for early individuals who can feel uncertain about choosing the starting point.
. 1. Introduction to cases of AI agents and agents
This tutorial introduces AI agents-a system-driven system by big language models (LLM) that feel more than their environment, tools and knowledge, and act-and-key agent types (simple/model-based reflex, purpose/utility, learning, learning, classification, classification, and multiplied.
You will learn when the agents will open the open, multi -phase, and correctional tasks, and the basic buildings of the agent solution: explaining tools, functions and behaviors.
. 2. Ai agent Framework detection
This tutorial searches the AI ​​agent’s framework with pre -constructed components and abstraction that allows you to quickly, repetition and deploy agents by standardizing shared challenges and promoting scales and developer performance.
You will compare Microsoft autojin, cementary kernel, and organized Azure AI agent service, and learn when the use of standstone tools is merged when it is integrated with your existing Azure ecosystem.
. 3. Understanding AI agent design samples
This tutorial introduces AI Agent Design Rule, which is a human -focused (UX) approach to building customer -focused agent experiences between the hereditary ambiguity of productive AI.
You will learn what the principles are, with practical guidelines to use them, and with their examples, emphasizing agents that fill human abilities, facilitating the difference of knowledge, facilitating cooperation, and helping people become better versions through auxiliary, intended interaction.
. 4. The toll use design sample
This tutorial introduces a sample of the tool use design, which allows LLM -powered agents to access external tools such as functions and control to the API, which enables them to take action only by producing text.
You will learn about key use matters, including dynamic data retrieval, code implementation, workflow automation, customer support integration, and content production/modification. In addition, this tutorial will cover the necessary building blocks of this design pattern, such as well -defined tool schemes, routing and selection logic, processed sandboxing, memory and observations, and error dealing (including timeout and re -attempting methods).
. 5. AgentCarg
This lesson describes the collective breed (RAG) from the agent recovery, which explains a multi -faceted and reasonable approach that is driven by large language models (LLM). In this approach, the model between device/function calls and structural output, evaluates the results, evaluates the results, improves questions, and repeats the process until satisfactorily receives a response. It often uses a maker checker loop to increase accuracy and recover from damaged questions.
You will learn about situations where agent rags have increased the first scenario, such as the first, first scenario of accuracy and the expansion tool, such as API calls. In addition, you will find out how reliable and increase the results of the reasoning process and the use of repentance loops.
. 6. Building reliable AI agents
This lesson teaches you how to enforce the best methods of security and privacy by designing a strong system message message framework (meta gestures, basic indicators, and refreshing), and teaching reliable AI agents by providing the user’s standard experience.
You will learn to identify the risks and reduce them, such as quick/round injection, access to unauthorized system, service overloading, poison pollution in knowledge base, and wrinkled errors.
. 7. Planning design sample
This tutorial focuses on planning design for AI agents. Start by explaining a clear overall purpose and setting the standard of success. After that, break the complex tasks in order and management all the tasks.
Use structural output formats to ensure reliable, machine -readable reactions, and enforce the event -powered orchestry to deal with dynamic tasks and unexpected inputs. Advocate agents with appropriate tools and appropriate tools and guidelines for their use.
To improve the final results, permanently evaluate the results, measure the performance, and repeat.
. 8. Multi -agent design pattern
This tutorial explains multi -agent design pattern, which involves connecting a number of special agents to cooperate toward the common goal. This approach is especially effective for complex, cross domain, or parallel tasks that benefit from the distribution of labor and integrated hand -offs.
In this tutorial, you will learn about the basic building blocks of this design pattern: an orchestrator/controller, characterized agents, joint memory/state, communication protocol, and routing/hand -off strategies, which include setting, harmony and group chat samples.
. 9. Metangation design sample
This tutorial introduces meta -identity, which can be considered as “thinking about thinking” for AI agents. Meta Identification allows these agents to monitor their reasoning process, to explain their decisions and to adopt their views and past experiences.
You will learn planning and diagnosis techniques, such as reflection, criticism, and maker checker patterns. These methods themselves promote correction, help identify mistakes, and help prevent endless reasoning loop. In addition, this technique will increase transparency, improve the quality of reasoning, and support better adaptation and impression.
. 10. AI agent in production
This tutorial shows how the “black box” agents can be converted into a “glass box” system by implementing a strong observation and diagnosis technique. You will create runs as a trace (representing the end -end tasks at the end) and as the Snins (applications for specific steps included in the language models or tools) as using platforms Langfuse And Azure Ai Foundry. This approach will enable you to analyze debugging and root causes, manage delays and costs, and confidence, safety and audit.
You will learn what aspects should be evaluated, such as output quality, safety, tool call success, delay, and cost, and strategies to enhance performance and effectiveness.
. 11. Using the Agentic Protocol
This tutorial introduces the Agentic Protocol that standardize the integrated and cooperative methods of AI agents. We will find three major protocols:
Model Context Protocol (MCP)Which provides access to tools, resources and indicators, provides access to the client server, acts as a “Universal Adapter” for context and abilities.
Agent Protocol from Agent (A2A)Which ensures safe, interconnected and task delegation between agents while completing the MCP.
Natural Language Web Protocol (NLWEB)Which enables websites’ Natural Natural Language interface, allowing agents to discover and communicate web content.
In this tutorial, you will learn about the purpose and benefits of each protocol, how they enable large language models (LLMs) to interact with tools and other agents, and where each one fits in large architecture.
. 12. Context Engineering for AI agents
This lesson introduces the context of engineering, which is the discipline of providing agents to the right information, the right form, and at the right time. This approach enables them to effectively plan their next steps, moving beyond one -time quick writing.
You will learn how context engineering is different from quick engineering, as it includes ongoing, dynamic curse instead of static instructions. In addition, you will understand why strategies such as writing, selection, compressing, and isolated information are essential for reliability, especially considering the limits of compulsive contexts.
. The final views
These Gut hub course AI provides everything you need to start building agents. This includes comprehensive lessons, short videos, and the Rinable Uzar Code. You can find titles in any sequence and run the samples using the Gut Hub Model (available for free) or Azure AI Foundry.
In addition, you will have the opportunity to work with Microsoft’s Azure AI agent service, cement kernel, and autogen. This course is a community -driven and open source. The contribution is welcomed, the problems are encouraged, and it is licensed for you for a fork and extension.
Abid Ali Owan For,,,,,,,,,, for,, for,,,, for,,,, for,,, for,,,, for,,,, for,,,, for,,, for,,, for,,, for,,, for,,,, for,,, for,,, for,,,, for,,, for,,,, for,,, for,,, for,,,, for,,, for,,, for,,,, for,,, for,,,, for,,, for,,,, for,,, for,,,, for,,, for,,,, for,,,, for,,,, for,,,, for,,,, for,,,, for,,,, for,,, for,,, for,,, for,,, for,,,,, for,,,, for,,,, for,,,, for,, for,.@1abidaliawan) A certified data scientist is a professional who loves to create a machine learning model. Currently, he is focusing on creating content and writing technical blogs on machine learning and data science technologies. Abid has a master’s degree in technology management and a bachelor’s degree in telecommunications engineering. Its vision is to create AI products using a graph neural network for students with mental illness.