Sequential chains in LangChain

Introduction

Sequential chains help you run multiple steps one after another, passing results from one step to the next. This makes complex tasks easier by breaking them into simple parts.

When you want to process data step-by-step, like cleaning text then summarizing it.

When you need to combine different language models or tools in order.

When you want to build a workflow that depends on previous answers.

When you want to keep your code organized by separating tasks.

When you want to reuse small chains to build bigger ones.

Syntax

LangChain

from langchain.chains import SequentialChain

chain = SequentialChain(
    chains=[chain1, chain2, ...],
    input_variables=["input1", "input2"],
    output_variables=["output1", "output2"]
)
result = chain.run({"input1": value1, "input2": value2})

You create a SequentialChain by giving it a list of smaller chains.

Input variables are the starting data you provide, output variables are what you want back at the end.

Examples

This example runs chain1 then chain2 in order, passing the output of chain1 to chain2.

LangChain

from langchain.chains import SequentialChain

# Define two simple chains
chain1 = SomeChain()
chain2 = AnotherChain()

# Create a sequential chain
seq_chain = SequentialChain(
    chains=[chain1, chain2],
    input_variables=["text"],
    output_variables=["summary"]
)

result = seq_chain.run({"text": "Hello world"})

Here, three chains run one after another to answer a question step-by-step.

LangChain

seq_chain = SequentialChain(
    chains=[chainA, chainB, chainC],
    input_variables=["question"],
    output_variables=["answer"]
)

answer = seq_chain.run({"question": "What is AI?"})

Sample Program

This program first summarizes the input text, then translates the summary to French using two chains connected sequentially.

LangChain

from langchain.llms import OpenAI
from langchain.chains import LLMChain, SequentialChain
from langchain.prompts import PromptTemplate

# Create a prompt template for first chain
template1 = "Summarize this text: {text}"
prompt1 = PromptTemplate(input_variables=["text"], template=template1)

# Create first chain
llm = OpenAI(temperature=0)
chain1 = LLMChain(llm=llm, prompt=prompt1, output_key="summary")

# Create a prompt template for second chain
template2 = "Translate this summary to French: {summary}"
prompt2 = PromptTemplate(input_variables=["summary"], template=template2)

# Create second chain
chain2 = LLMChain(llm=llm, prompt=prompt2, output_key="french_translation")

# Create sequential chain
seq_chain = SequentialChain(
    chains=[chain1, chain2],
    input_variables=["text"],
    output_variables=["summary", "french_translation"]
)

# Run the chain
result = seq_chain.run({"text": "Langchain helps you build chains of language models."})
print(result)

Important Notes

Make sure each chain's output keys match the next chain's input variables.

Sequential chains keep your workflow clear and easy to debug.

You can nest sequential chains inside other chains for more complex flows.

Summary

Sequential chains run multiple chains one after another.

They pass outputs from one chain as inputs to the next.

This helps build clear, step-by-step workflows.

Practice

1. What is the main purpose of a SequentialChain in Langchain?

easy

A. To create a single chain that never passes outputs

B. To run multiple chains all at the same time independently

C. To run multiple chains one after another, passing outputs as inputs

D. To stop chains from running automatically

Solution

1. Step 1: Understand SequentialChain behavior

   A SequentialChain runs chains in order, passing output from one to the next.

2. Step 2: Compare options to this behavior

   Only To run multiple chains one after another, passing outputs as inputs describes this step-by-step passing of outputs between chains.

3. Final Answer:

   To run multiple chains one after another, passing outputs as inputs -> Option C

4. Quick Check:

   SequentialChain = run chains sequentially with output passing [OK]

Hint: Sequential means one after another with output passing [OK]

Common Mistakes:

• Thinking chains run in parallel
• Believing outputs are not passed
• Confusing SequentialChain with single chain

2. Which of the following is the correct way to create a SequentialChain with two chains named chain1 and chain2?

easy

A. SequentialChain([chain1, chain2])

B. SequentialChain(chains=[chain1, chain2], input_variables=["input"], output_variables=["output"])

C. SequentialChain(chain1, chain2)

D. SequentialChain(chains=chain1 + chain2)

Solution

1. Step 1: Recall SequentialChain constructor

   It requires a list of chains and lists of input and output variable names.

2. Step 2: Check each option's syntax

   Only SequentialChain(chains=[chain1, chain2], input_variables=["input"], output_variables=["output"]) correctly uses named parameters with lists for chains and variables.

3. Final Answer:

   SequentialChain(chains=[chain1, chain2], input_variables=["input"], output_variables=["output"]) -> Option B

4. Quick Check:

   Correct constructor syntax = SequentialChain(chains=[chain1, chain2], input_variables=["input"], output_variables=["output"]) [OK]

Hint: Look for named parameters and list brackets [OK]

Common Mistakes:

• Passing chains without list brackets
• Missing input/output variable lists
• Using plus operator to combine chains

3. Given the following code snippet, what will be the final output printed?

from langchain.chains import SequentialChain

chain1 = SomeChain()  # outputs {'intermediate': 'hello'}
chain2 = SomeChain()  # expects input 'intermediate' and outputs {'final': 'hello world'}

seq_chain = SequentialChain(chains=[chain1, chain2], input_variables=['input'], output_variables=['final'])

result = seq_chain.run({'input': 'start'})
print(result)

medium

A. Error: missing input for chain2

B. 'hello world'

C. 'start'

D. {'final': 'hello world'}

Solution

1. Step 1: Understand chain outputs and inputs

   chain1 outputs {'intermediate': 'hello'}, chain2 uses 'intermediate' input and outputs {'final': 'hello world'}.

2. Step 2: SequentialChain runs chain1 then chain2, passing outputs

   Final result is {'final': 'hello world'}, printed as "{'final': 'hello world'}".

3. Final Answer:

   {'final': 'hello world'} -> Option D

4. Quick Check:

   Output of SequentialChain = {'final': 'hello world'} [OK]

Hint: Final output is dict of output_variables [OK]

Common Mistakes:

• Expecting string instead of dict repr
• Confusing input and output keys
• Assuming error due to input passing

4. What is the error in this code snippet that tries to create a SequentialChain?

seq_chain = SequentialChain(chains=[chain1, chain2], input_variables=['input'])
result = seq_chain.run({'input': 'data'})

medium

A. Missing output_variables parameter causes an error

B. Chains list should be a tuple, not a list

C. Input dictionary keys do not match input_variables

D. run() method requires no arguments

Solution

1. Step 1: Check required parameters for SequentialChain

   Both input_variables and output_variables are required parameters.

2. Step 2: Identify missing parameter

   The code misses output_variables, so it will raise an error.

3. Final Answer:

   Missing output_variables parameter causes an error -> Option A

4. Quick Check:

   output_variables missing = error [OK]

Hint: Always provide input and output variable lists [OK]

Common Mistakes:

• Forgetting output_variables
• Using wrong data type for chains
• Passing arguments incorrectly to run()

5. You want to build a SequentialChain that first extracts keywords from text, then summarizes those keywords. Which approach correctly sets up this workflow?

hard

A. Create two chains: keyword_extractor outputs 'keywords'; summary_chain takes 'keywords' as input; combine with SequentialChain passing these variables

B. Create one chain that does both extraction and summary in one step

C. Run keyword_extractor and summary_chain separately without chaining outputs

D. Use SequentialChain but ignore passing variables between chains

Solution

1. Step 1: Identify the workflow steps

   First extract keywords, then summarize them, so output of first is input of second.

2. Step 2: Use SequentialChain with proper variable passing

   Set keyword_extractor to output 'keywords', summary_chain to input 'keywords', then chain them sequentially.

3. Final Answer:

   Create two chains: keyword_extractor outputs 'keywords'; summary_chain takes 'keywords' as input; combine with SequentialChain passing these variables -> Option A

4. Quick Check:

   SequentialChain passes outputs as inputs [OK]

Hint: Chain outputs must match next chain inputs [OK]

Common Mistakes:

• Trying to do both steps in one chain
• Not passing outputs to next chain
• Ignoring variable names in chaining