AI Agents with LangChain & Autonomous Workflows · 课时

顺序链与简单链

学习构建用于按顺序执行任务的基础链，将一个步骤的输出传递给下一步骤作为输入

顺序链与简单链 是 CoddyKit 上的免费 AI Agents with LangChain & Autonomous Workflows 课时。这是第 2 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 AI Agents with LangChain & Autonomous Workflows 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 AI Agents with LangChain & Autonomous Workflows 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Chains: Steps in Order

In LangChain, chains allow you to combine multiple Large Language Model (LLM) calls and other utilities into a single, coherent application.

Think of them as a series of steps where the output of one step becomes the input for the next. This creates a powerful, automated workflow.

Sequential Chains are a specific type designed for tasks that require a strict, ordered execution of steps.

Meet the LLMChain

Before we build complex sequential chains, let's look at the basic building block: the LLMChain.

An LLMChain combines an LLM (like GPT-4) with a PromptTemplate. It takes an input, formats it using the template, sends it to the LLM, and returns the LLM's response.

LLM: The language model that generates text.
PromptTemplate: Defines how user input is structured for the LLM.

Running a Single LLMChain

Here's how you can create and run a simple LLMChain. We'll ask it to suggest a catchy name for a new tech product.

Notice how the PromptTemplate defines what we expect the LLM to do, and the chain executes it.

import os
from langchain.chains import LLMChain
from langchain_core.prompts import PromptTemplate
from langchain_openai import ChatOpenAI

# IMPORTANT: Set your OpenAI API key as an environment variable
# os.environ["OPENAI_API_KEY"] = "YOUR_API_KEY_HERE"
# For local testing, ensure it's set before running.

def main():
    # 1. Define the LLM
    llm = ChatOpenAI(temperature=0.7)

    # 2. Define the Prompt Template
    prompt_template = PromptTemplate(
        input_variables=["product_type"],
        template="Suggest a catchy name for a new {product_type}."
    )

    # 3. Create the LLMChain
    name_chain = LLMChain(llm=llm, prompt=prompt_template)

    # 4. Run the chain
    result = name_chain.invoke({"product_type": "AI assistant"})
    print(f"Suggested Name: {result['text']}")

if __name__ == "__main__":
    main()

Introducing SimpleSequentialChain

What if you need to perform multiple steps, where each step builds on the previous one? That's where SimpleSequentialChain comes in.

It takes a list of chains and executes them in order. The single output from the first chain becomes the single input for the second chain, and so on.

It's ideal for straightforward, linear workflows.

Input-Output Flow

SimpleSequentialChain manages the flow automatically:

You provide an initial input to the first chain.
The first chain processes it and produces an output.
This output automatically becomes the input for the second chain.
This continues until the last chain, whose output is the final result of the SimpleSequentialChain.

It's like an assembly line for AI tasks!

First Step: Name Generation

Let's build a two-step sequential chain. Our first step will be to generate a product name, similar to our previous LLMChain example.

We define an LLMChain for this purpose. We'll make sure its output is clearly defined using output_key for the next step.

import os
from langchain.chains import LLMChain
from langchain_core.prompts import PromptTemplate
from langchain_openai import ChatOpenAI

# IMPORTANT: Set your OpenAI API key as an environment variable
# os.environ["OPENAI_API_KEY"] = "YOUR_API_KEY_HERE"

def main():
    llm = ChatOpenAI(temperature=0.7)

    # Chain 1: Generate a product name
    prompt_name = PromptTemplate(
        input_variables=["product_type"],
        template="Suggest a catchy, short name for a new {product_type}. Response should only be the name."
    )
    chain_name = LLMChain(llm=llm, prompt=prompt_name, output_key="product_name")

    # This chain will output a dictionary like {'product_type': '...', 'product_name': '...'}
    # The 'product_name' will be passed to the next chain.
    result = chain_name.invoke({"product_type": "smartwatch"})
    print(f"Input: {result['product_type']}")
    print(f"Output (product_name): {result['product_name']}")

if __name__ == "__main__":
    main()

Second Step: Slogan Generation

Now, let's create the second LLMChain. This chain will take the product_name generated by the first chain and create a slogan for it.

Notice how its input_variables matches the output_key from the previous chain. This is how SimpleSequentialChain knows how to connect them.

import os
from langchain.chains import LLMChain
from langchain_core.prompts import PromptTemplate
from langchain_openai import ChatOpenAI

# IMPORTANT: Set your OpenAI API key as an environment variable
# os.environ["OPENAI_API_KEY"] = "YOUR_API_KEY_HERE"

def main():
    llm = ChatOpenAI(temperature=0.7)

    # Chain 2: Generate a slogan based on the product name
    prompt_slogan = PromptTemplate(
        input_variables=["product_name"],
        template="Write a creative slogan for a product named {product_name}. Response should only be the slogan."
    )
    chain_slogan = LLMChain(llm=llm, prompt=prompt_slogan, output_key="slogan")

    # Example of running this chain independently
    result = chain_slogan.invoke({"product_name": "ChronoMind"})
    print(f"Input: {result['product_name']}")
    print(f"Output (slogan): {result['slogan']}")

if __name__ == "__main__":
    main()

Putting Chains Together

Finally, we combine our two LLMChains into a SimpleSequentialChain. We pass the initial input to the overall sequential chain, and it handles the rest!

We'll also set verbose=True to see the intermediate steps, which is great for debugging.

import os
from langchain.chains import LLMChain, SimpleSequentialChain
from langchain_core.prompts import PromptTemplate
from langchain_openai import ChatOpenAI

# IMPORTANT: Set your OpenAI API key as an environment variable
# os.environ["OPENAI_API_KEY"] = "YOUR_API_KEY_HERE"

def main():
    llm = ChatOpenAI(temperature=0.7)

    # Chain 1: Generate a product name
    prompt_name = PromptTemplate(
        input_variables=["product_type"],
        template="Suggest a catchy, short name for a new {product_type}. Response should only be the name."
    )
    chain_name = LLMChain(llm=llm, prompt=prompt_name, output_key="product_name")

    # Chain 2: Generate a slogan
    prompt_slogan = PromptTemplate(
        input_variables=["product_name"],
        template="Write a creative slogan for a product named {product_name}. Response should only be the slogan."
    )
    chain_slogan = LLMChain(llm=llm, prompt=prompt_slogan, output_key="slogan")

    # Combine into a SimpleSequentialChain
    overall_chain = SimpleSequentialChain(
        chains=[chain_name, chain_slogan],
        verbose=True # Set to True to see intermediate steps
    )

    # Run the overall chain with the initial input
    final_result = overall_chain.invoke({"product_type": "AI-powered coffee maker"})
    print("\n--- Final Result ---")
    print(f"Product Slogan: {final_result['slogan']}")

if __name__ == "__main__":
    main()

Tracing Chain Execution

When verbose=True, LangChain prints out the steps taken by your chain. This is incredibly useful for:

Understanding how inputs and outputs flow.
Debugging unexpected results.
Seeing which prompts are sent to the LLM.

It helps you visualize the 'assembly line' in action and ensures each step performs as expected.

Quick Chain Check

Consider a SimpleSequentialChain with three LLMChains: Chain A, Chain B, and Chain C, in that order.

Chain A has output_key='topic'.
Chain B has input_variables=['topic'] and output_key='outline'.
Chain C has input_variables=['outline'].

If the SimpleSequentialChain is invoked with {'initial_input': 'AI agents'}, which statement is true?

Recap: Simple Sequential Chains

Great job! In this lesson, you learned about:

The basic LLMChain as a building block.
How SimpleSequentialChain links multiple chains together.
Passing outputs from one step as inputs to the next.
Using output_key and input_variables for flow.
Debugging chains with verbose=True.

Next, you'll discover how to customize chain logic and integrate your own Python functions within LangChain workflows!

常见问题解答

「顺序链与简单链」课时是免费的吗？

是的 — 「顺序链与简单链」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 AI Agents with LangChain & Autonomous Workflows 课程的其余内容，请升级到 CoddyKit PRO。 AI Agents with LangChain & Autonomous Workflows 课程共包含 4 节课。

「顺序链与简单链」这节课中我会学到什么？

学习构建用于按顺序执行任务的基础链，将一个步骤的输出传递给下一步骤作为输入你通过在浏览器中直接运行的动手代码来练习 AI Agents with LangChain & Autonomous Workflows，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 AI Agents with LangChain & Autonomous Workflows 需要有经验吗？

无需任何先前经验。CoddyKit 上的 AI Agents with LangChain & Autonomous Workflows 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 2 节课，共 4 节。

「顺序链与简单链」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 AI Agents with LangChain & Autonomous Workflows 课中编写并运行代码吗？

能。每节 AI Agents with LangChain & Autonomous Workflows 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 AI Agents with LangChain & Autonomous Workflows