(P4478/name) AI Agent Development Course: Tool Calling, Memory Management, LangGraph Framework, E-commerce Customer Service Real Business Scenario Cases

(P4478/description) Course Introduction:
(P4478/description) This Agent project practical course covers both theoretical principles and real e-commerce customer service implementation. The theory section covers AI Agent basics, execution loops, prompt constraints, context engineering, structured output, ToolCalling, AgenticRAG, short/long-term memory, LangGraph framework, multi-agent collaboration, security permissions, and observability evaluation. The business practice uses e-commerce customer service as a case study, starting from chat service setup, step by step solving hallucinations, inaccurate retrieval, tool errors, context overflow, and unauthorized operations, completing refund/return flows and peak stress drills, and outlining launch boundaries and project roadmap. Suitable for technical learners wanting to implement agents.

(P4478/description) Course Outline:
(P4478/description) 01_What is an AI Agent
(P4478/description) 02_Models and Messages
(P4478/description) 03_Agent Execution Loop
(P4478/description) 04_E-commerce Business Agent Scenario Map
(P4478/description) 05_How Prompts Constrain Business Roles
(P4478/description) 06_Context Engineering Basics
(P4478/description) 07_Structured Output
08_Streaming and User Experience
09_ToolCalling Basics
10_Tool Parameters, Return Values and Error Handling
11_ToolRuntime, ToolNode and LangGraph Tool Nodes
12_First create_agent
13_Why RAG is Needed
14_Document Chunking and Metadata
15_Citation Recall and Anti-Hallucination
16_AgenticRAG Introduction
17_Short-term Memory
18_Long-term Memory and Privacy Boundaries
19_Runtime and Context
20_Why LangGraph is Needed
21_State Nodes and Edges
22_FunctionalAPI Workflows
23_Persistence, Recovery and Time Travel
24_Human-in-the-loop and Interrupt
25_Error Handling and Fault Tolerance
26_Observability and Trace
27_Evaluation and Testing
28_MCP Security and Permission Boundaries
29_ClaudeCode Engineering Mindset
30_Context Engineering and Project Memory
31_Tool Permission Sandbox and Security Boundaries
32_Hooks, Skills, MCP and Extension Layering
33_Subagents Multi-Agent Collaboration and Governance
Lesson 01: Customer Service Ran Away, Boss Tells You to Answer the Phone
Lesson 02: Build Chat Service, Let Messages Flow In
Lesson 03: First Version AI Chats Well But Can't Answer Business Questions
Lesson 04: AI Says I'll Handle It But After-sales Got No Return Request
Lesson 05: AI Promises Compensation, You Stuff All Rules into Prompt
Lesson 06: Prompt Grows Like a Wall, Fear Breaking It When Changing One Spot
Lesson 07: Month-end Bill Arrives: See Token Burn per Round
Lesson 08: Problem Isn't Model Dumb, It's Your Crude Input
Lesson 09: Split After-sales Rules and Tag Each Segment
Lesson 10: First Vector Retrieval: Find Closest Rule to User Question
Lesson 11: Boss Finally Sees Answer Source
Lesson 12: Hit Doesn't Mean Correct, Can't Fabricate Without Basis
Lesson 13: User Asks About Earphone Promotion, Vector Retrieval Got Wrong Rule
Lesson 14: Too Many Similar Rules, Top-k Might Not Be Right
Lesson 15: Why RAG Is Needed
Lesson 16: Document Chunking and Metadata
Lesson 17: Citation Recall and Anti-Hallucination
Lesson 18: AgenticRAG Introduction
Lesson 19: Short-Term Memory
Lesson 20: Long-Term Memory and Privacy Boundaries
Lesson 21: Runtime and Runtime Context
Lesson 22: Why LangGraph Is Needed
Lesson 23: State, Nodes, and Edges
Lesson 24: Functional API Workflows
Lesson 25: Persistence, Recovery, and Time Travel
Lesson 26: Human-in-the-loop and Interrupt
Lesson 27: Error Handling and Fault Tolerance
Lesson 28: Observability and Trace
Lesson 29: Evaluation and Testing
Lesson 30: MCP Security and Permission Boundaries
Lesson 31: ClaudeCode Engineering Mental Model
Lesson 32: Context
Lesson 33: User Claims VIP, Doesn't Mean They Really Are
Lesson 34: History, Tool Results, RAG Chunks All Crowded Together, Model Starts Confusing
Lesson 35: Context Too Long, Model Starts Forgetting Middle Content
Lesson 36: Users, Tools, Knowledge Base Can All Bring Dirty Instructions
Lesson 37: Boss Asks Why It Answered That, Can't Just Say LLM Thinks So
Lesson 38: After Changing Prompt, Shipping Improved but Refunds Broke
Lesson 39: Don't Randomly Change After Failure, Bad Reviews Aren't Free
Lesson 40: Boss Checks Bill Again, Start Cost Tiering
Lesson 41: Night Before Peak Sale, Full Dress Rehearsal of E-commerce Customer Service Agent
Lesson 42: First Wave of Peak Sale: Shipping and Promotion Rules Must Be Right
Lesson 43: Second Wave: User Wants Refund, AI Must Stop and Wait for Approval
Lesson 44: Third Wave: Service Glitches, Unauthorized Queries, Knowledge Misses All Hit
Lesson 45: Boss Asks Can We Launch, Write Boundaries First
Lesson 46: Boss Asks Can It Be Stronger, Explain Roadmap and Project Expression