AI / LLM
Summary of AI，LLM，ML and their relationship
AI (Artificial Intelligence) contains ML (Machine Learning), which contains LLM (Large Language Models) and LLMs are a specific type of ML model focused on language.— like nested Russian dolls.
AI（人工智能）包含 ML（机器学习），ML 再包含 LLM（大语言模型）, 而LLM是ML中专门处理语言的一类模型。——就像俄罗斯套娃一样层层嵌套。
Harness Engineering and its 12 core Modules
The management framework that controls, coordinates, and enables AI agents to work reliably using memory, tools, workflows, and safety controls.
Agent Harness 是管理 AI Agent 的“控制系统”，负责记忆、工具调用、工作流、安全和整体协调。
What is Databricks Genie
Databricks Genie is Databricks’ natural-language analytics interface. It allows business users to ask questions about governed enterprise data in plain English, instead of writing SQL.
Databasic Agent Bricks Encosystem
Databricks Genie
Databricks Genie is Databricks’ natural-language analytics interface. It allows business users to ask questions about governed enterprise data in plain English, instead of writing SQL. more …
Since 2026, the AI industry has converged on a shared stack for building intelligent agents:
- LLM – the brain (OpenAI, Anthropic, Meta)
- MCP – the standard protocol for tools & data (Anthropic)
- Agent Skills – reusable capabilities (Archetype AI)
- Harness – the runtime orchestration layer (Microsoft, others)
These four pieces are now widely adopted, but nobody had given them a single, simple name.
Now， AgentForge is that name.
AgentForge = LLM + MCP + Agent Skills + Harness
It’s the complete approach to building production‑ready agents — not just calling an LLM.
Why “Forge”? Because you don’t plug these pieces together. You forge them: heat, harnness, shape, harden. Until they work as one.
LLM Foundation
RAG 2.0
Memory System
MCP protocols (Model Context Protocol)
Agent Capability Design
Harness
Evaluation
Azure AI Ecosystems
Deployment & CI/CD
Integration
Build production agents on LangChain/LangGraph, forged not just chained.
We will discuss the AI/LLM Engineering Stack in detail — designed in the order of building a complete AI system from the ground up.
1️⃣ OpenAI API : Learn to call LLM directly
2️⃣ LangChain: Learn to orchestrate LLM with a framework
3️⃣ RAG: Learn to inject external knowledge into LLM
4️⃣ Agent: Learn to let LLM make autonomous decisions
5️⃣ Tool System: Learn to connect LLM to external capabilities
6️⃣ Complete System Integration: Combine everything into a real working product
This step, we will learn to call LLM directly.
This step, we will learn how to orchestrate LLM with a framework – LangChain.
LangChain is an open-source framework for building applications powered by Large Language Models (LLMs). It provides tools, abstractions, and integrations that make it easier to connect LLMs with external data, APIs, memory, and other components — turning a simple AI chat into a powerful application.
LangChain Six Core Modules
LangChain 框架
├── 1. Models → 调用各种 LLM
├── 2. Prompts → 管理提示词
├── 3. Chains → 串联组件
├── 4. Memory → 存储对话历史
├── 5. Tools → 调用外部能力
└── 6. Agents → 自主决策