A fast-paced introduction to agents in LangChain. It is packed with examples and animations to get the main points across as simply as possible. OpenAI function calls will be covered in a future video.
💌 Link to the newsletter
practical-ai-builder.beehiiv....
⏳ Timestamps
00:00 Intro
00:04 What can LLMs do?
00:19 Agents dependencies
00:23 What can an agent do?
00:33 Table of content
01:00 Action Agents / Thought-Action-Observation Loop Explained
1:08 Thought Step Explained
1:20 Agent's Prompt / What's inside?
1:34 Agent's Output Parse / It's importance
1:50 Thought Step Explained (continued)
2:00 The Role of the AgentExecutor
2:18 Observation / Tool output
2:30 Agent Loop Stopping Condition
2:37 Agent Code Example (NYC Real Estate Mortgage Deposit)
3:02 Agent Types
3:12 ReAct DocStore Agent
3:21 Self-ask with Search Agent
3:46 Comparison between (Zero Shot | Conversational | Chat Zero Shot | Chat Conversational | Chat Conversational | Structured Chat Zero Shot) - React Description Agent
3:52 Agent Type for Chat Models
4:00 Agent Type for Chat History
4:05 Agent Type for Multi-Argument Tools
4:11 Other Agent Frameworks (Plan-and-execute, Autonomous, Generative)
4:19 Plan-and-Execute Agent (Why?)
4:37 Action Agent Control Flow
4:46 Plan-and-Execute Control Flow
4:54 The Planner
5:02 The Executor
5:04 PROs/CONs Plan-and-execute
5:24 Sales Report Agent Code Example (Gmail, SQLite DB)
6:00 Results: Sales Report Agent Code Example (Gmail, SQLite DB)
6:10 Plan-and-Execute Inspiration
6:20 Autonomous Agent
6:23 BabyAGI Introduction
6:30 BabyAGI Explained Graphically
7:05 Consensus on BabyAGI (Useful or nah?)
7:17 Generative Agent
7:30 Tools & Toolkits (Command to find available tools + toolkits)
7:37 Toolkit Example Overview: GmailToolkit
7:44 Creating a Custom Tool
7:53 Using Pydantic w/ Custom Tool
7:57 Decorator Custom Tool
8:00 Structured Custom Tool (Multi-arguments)
8:05 Prompts & Output Parsers
8:15 Check out LangChain's Prompts Source Code
8:23 OpenAI Function Calls (Postponed)
8:33 Thank you
🔗 Links
Source code: github.com/edrickdch/langchai...
LangChain: python.langchain.com/docs/mod...
ReAct Paper: arxiv.org/abs/2210.03629
MRKL Paper: www.ai21.com/blog/jurassic-x-...
Self-ask with search Paper: python.langchain.com/docs/mod...
BabyAGI: github.com/yoheinakajima/babyagi
Generative Agents Paper: arxiv.org/abs/2304.03442
Plan-and-solver Paper: github.com/AGI-Edgerunners/Pl...
💬 If you're looking for a PDF Chat Tool: pdf.ai/?via=edrick
🙏 Support the channel with a donation: paypal.me/edrickdch
Негізгі бет Ғылым және технология LangChain Agents: A Simple, Fast-Paced Guide
Пікірлер: 9