AI各层级的易混淆概念

一个完整层级图

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
产品层
└── Assistant(ChatGPT / Copilot)

智能体层
└── Agent
├── Planner
├── Memory
├── Tools
└── Skills

能力编排层
├── Skill(流程封装)
├── Chain(固定步骤)
└── Function Calling

工具层
└── Tools(API / DB / Search)

模型层
└── LLM(GPT / Claude / LLaMA)

Assistant = UI壳子
Agent = 决策大脑
Skill = 固定能力流程
Tool = 外部执行器
Model = 语言与推理引擎
Infra = 运行与部署系统

更详细的内容

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
🟣 产品交互层(Product / Experience Layer)

├── Assistant(AI助手产品)
│ ├── ChatGPT / Copilot / Claude App
│ ├── Web / Mobile UI
│ ├── 对话管理(session / thread)
│ └── 权限 / 用户系统

├── Copilot(半自动辅助)
│ ├── IDE插件(代码补全)
│ ├── Office助手
│ └── 人类主导 + AI辅助

└── Multi-Agent Apps(多智能体应用)
├── Research App
├── AI Analyst
└── AI Dev Team

🟡 智能体编排层(Agent Orchestration Layer)

├── Agent(智能体核心)
│ ├── Goal / Objective(目标)
│ ├── Reasoning Engine(推理)
│ ├── Action Selector(动作选择)
│ └── Loop Controller(循环执行)

├── Planner(任务规划器)
│ ├── Task decomposition(任务拆解)
│ ├── Dependency graph(依赖关系)
│ └── Step ordering(步骤排序)

├── Executor(执行器)
│ ├── Tool invocation
│ ├── Skill execution
│ └── Result validation

├── Memory System(记忆系统)
│ ├── Short-term memory(上下文窗口)
│ ├── Long-term memory(向量数据库)
│ └── Episodic memory(事件记录)

└── Reflection / Critic(反思系统)
├── Self-check
├── Error correction
└── Retry strategy

🔵 能力编排层(Cognitive / Workflow Layer)

├── Skill(技能封装)
│ ├── Data analysis skill
│ ├── Report generation skill
│ └── Code generation skill

├── Chain(固定流程)
│ ├── Prompt Chain
│ ├── Tool Chain
│ └── RAG Chain

├── Workflow Engine
│ ├── DAG execution(有向无环图)
│ ├── condition branching
│ └── parallel execution

└── Prompt Engineering Layer
├── System prompt
├── Few-shot examples
└── Role prompting

🟢 工具与外部能力层(Tooling Layer)

├── Tools(工具/API)
│ ├── REST APIs
│ ├── gRPC services
│ ├── DB query tools
│ ├── Web search tools
│ └── Code execution tools

├── Function Calling Protocol
│ ├── JSON schema
│ ├── parameter validation
│ └── structured output

├── RAG System(检索增强生成)
│ ├── embedding model
│ ├── vector database
│ ├── retriever
│ └── reranker

└── External Systems
├── SaaS APIs (Slack, Jira)
├── Internal microservices
└── data warehouse

🟠 模型能力层(Model Layer)

├── Foundation Model
│ ├── GPT-4 / GPT-5 系列
│ ├── Claude 系列
│ ├── LLaMA 系列
│ └── Mistral 等

├── Instruction-tuned Model
│ ├── RLHF / DPO 训练
│ └── chat optimization

├── Multimodal Models
│ ├── text + image
│ ├── text + audio
│ └── text + video

└── Reasoning Models
├── chain-of-thought
├── tree-of-thought
└── planning enhanced models

🔴 基础设施层(Infra Layer)

├── Inference Engine
│ ├── vLLM
│ ├── TensorRT-LLM
│ └── Ollama

├── Serving Layer
│ ├── API Gateway
│ ├── Load Balancer
│ └── Rate Limiting

├── Data Layer
│ ├── Vector DB (Pinecone / Milvus)
│ ├── Relational DB
│ └── Object Storage

├── Observability
│ ├── logging
│ ├── tracing
│ └── prompt monitoring

└── Security Layer
├── prompt injection protection
├── tool sandboxing
└── data isolation
Notice: 正常情况下,这里会有一个基于utteranc.es的留言系统,如果看不到,可能要想想办法才能看到。

Powered by Hexo and Hexo-theme-hiker

Copyright © 2012 - 2026 tiaobug.com All Rights Reserved.

鲁ICP备2024124237号-1