LLM Development Basics
This course is for developers with development experience but no prior knowledge of the LLM field. The course aims to help developers systematically master the core principles, component composition, security risks of large language models (LLM), and the practical application of mainstream development frameworks (LangChain & LangGraph). From theoretical introduction to project demonstration, the course proceeds step by step, taking into account both technical breadth and engineering implementation.
Day 1: Concept Establishment and Basic Theory
From traditional AI to Generative Models
Study Content +
1.1 LLM Theoretical Basis
The evolution from traditional AI to generative AI
Core principles of LLMs and the most important technical concepts
An overview of mainstream LLM providers and the open-source ecosystem
Common model categories and representative application scenarios
1.2 Core Components and Key Concepts
Fundamental elements such as prompts, models, agents, tools, and memory
Agent protocols including MCP, A2A, and AG-UI
Analysis of typical LLM application architectures and case studies
1.3 Tools and Hands-On Demonstrations
Demonstrations of common tools such as ChatGPT, Claude, and Gemini
Hands-on experience with interaction patterns such as MCP, Canvas, and Deep Search
Exploring LLM use cases in development with tools like GitHub Copilot and Cursor
1.4 Security and Compliance Fundamentals
An overview of potential LLM-related risks
A detailed explanation of the OWASP Top 10 for LLMs (2025) with practical recommendations
Learning Objectives +
- โ Understand the development history of AI from classical methods to generative models
- โ Master the basic operating principles and core components of large language models
- โ Understand the characteristics of mainstream LLM providers and how to choose among them
- โ Become familiar with typical LLM application scenarios and security risks
Day 2: Development Frameworks and Project Preparation
A development path from theory to practice
Study Content +
2.1 LLM Project Development Workflow
How to plan an LLM application project
A complete development workflow from requirements analysis to model integration
Practical paths for data preparation, model access, and service deployment
2.2 Overview of Mainstream Development Frameworks
A comparison of the core architectures of LangChain and LangGraph
Detailed explanation of the core modules used to build agents, including tools, memory, and executors
Applicable scenarios for common protocol support such as MCP, A2A, and AG-UI
2.3 Framework Practice
Get started with LangChain and LangGraph using official demos
Implement basic conversational tasks by combining agents, tools, and memory
2.4 Preparation for the Practice Project
Introduce the practice environment and starter code structure
Clarify project goals and task breakdown to establish a smooth development rhythm
Learning Objectives +
- โ Master the complete development workflow and planning approach for LLM projects
- โ Understand the characteristics of mainstream development frameworks and when to use them
- โ Learn how to build a fully functional agent system
- โ Become familiar with standard protocols and communication mechanisms for LLM applications
Day 3: Integrated Practice and Hands-On Labs
A complete journey from concepts to practical application
Study Content +
3.1 Lab 1: A Basic Chatbot
Goal: build a minimum viable LLM chatbot
Steps: prompt design -> API integration -> message interaction
3.2 Lab 2: Building a Simple Agent
Goal: build an agent with basic reasoning capabilities
Steps: define the agent -> integrate tools -> manage context
3.3 Lab 3: Calling External Tools
Goal: implement an agent that can call external tools such as search and calculators
Steps: define the tools -> build the invocation chain -> integrate returned results
3.4 Lab 4: MCP Server Practice
Goal: build an agent service that follows the MCP protocol
Steps: understand the protocol -> build the service -> collaboratively debug and test the interface
3.5 Wrap-Up and Q&A
Review the main knowledge points covered over the three days
Answer common questions and suggest next-step learning paths
Learning Objectives +
- โ Master the complete development workflow of LLM applications through real project practice
- โ Learn how to build complex multi-agent collaboration systems
- โ Develop practical skills for integrating and invoking a variety of tools
- โ Understand how the MCP protocol is applied in real projects
- โ Establish a complete LLM project development framework and a set of best practices