What is a chain in LangChain - Visual Explanation

Concept Flow - What is a chain in LangChain

Start Chain

↓

Input Received

↓

Process Step 1: Call LLM or Tool

↓

Process Step 2: Use Output as Input

↓

... Repeat for all steps ...

↓

Final Output Produced

↓

Chain Ends

A chain in LangChain takes an input, processes it step-by-step through linked components, and produces a final output.

Execution Sample

LangChain

from langchain.chains import SimpleSequentialChain
from langchain.llms import OpenAI

llm = OpenAI()
chain = SimpleSequentialChain(llm=llm)
result = chain.run("Hello")

This code creates a simple chain that sends input to an LLM and returns the output.

Execution Table

Step	Action	Input	Process	Output
1	Start Chain	Hello	Receive input	Hello
2	Call LLM	Hello	LLM generates response	Hi! How can I help you?
3	Chain Output	Hi! How can I help you?	Return final output	Hi! How can I help you?
4	End Chain	-	-	-

💡 Chain ends after producing the final output from the LLM.

Variable Tracker

Variable	Start	After Step 1	After Step 2	Final
input_text	-	Hello	Hello	Hello
llm_response	-	-	Hi! How can I help you?	Hi! How can I help you?
chain_output	-	-	-	Hi! How can I help you?

Key Moments - 2 Insights

Why does the chain pass the output of one step as input to the next?

Because chains link multiple steps, each step uses the previous output as input to continue processing, as shown in execution_table row 2.

Is the chain just a single function call?

No, a chain can have multiple linked steps, but SimpleSequentialChain runs them one after another, as seen in the concept_flow.

Visual Quiz - 3 Questions

Look at the execution_table, what is the output after Step 2?

A "Chain Ends"

B "Hello"

C "Hi! How can I help you?"

D "Start Chain"

💡 Hint

Check the 'Output' column in execution_table row 2.

At which step does the chain produce the final output?

A Step 3

B Step 2

C Step 1

D Step 4

💡 Hint

Look at the 'Action' and 'Output' columns in execution_table row 3.

If the input to the chain changes, which variable in variable_tracker changes first?

A llm_response

B input_text

C chain_output

D None

💡 Hint

See variable_tracker row for 'input_text' and its values after Step 1.

Concept Snapshot

LangChain chains link multiple steps to process input.
Each step uses the previous step's output as input.
Chains can call LLMs, tools, or other logic.
SimpleSequentialChain runs steps one after another.
Final output is returned after all steps complete.

Full Transcript

A chain in LangChain is like a sequence of steps that take an input and process it step-by-step. First, the chain receives the input. Then it sends this input to a language model or tool. The output from that step becomes the input for the next step. This continues until all steps finish, and the chain returns the final output. For example, a SimpleSequentialChain sends the input to an LLM and returns the response. Variables like input_text and llm_response change as the chain runs. Understanding this flow helps beginners see how LangChain organizes complex tasks into simple linked steps.

Practice

1. What is a chain in LangChain?

easy

A. A single prompt sent to a language model

B. A database used to store language model outputs

C. A sequence of steps linking language model calls to perform a task

D. A tool to visualize language model responses

Solution

1. Step 1: Understand the purpose of a chain

   A chain connects multiple language model calls and prompts to complete a task step-by-step.

2. Step 2: Compare options

   Only A sequence of steps linking language model calls to perform a task describes this linking of steps. The other options describe unrelated concepts.

3. Final Answer:

   A sequence of steps linking language model calls to perform a task -> Option C

4. Quick Check:

   Chain = linked steps for tasks [OK]

Hint: Chains link multiple steps to solve tasks [OK]

Common Mistakes:

• Thinking a chain is just one prompt
• Confusing chains with data storage
• Assuming chains are visualization tools

2. Which of the following is the correct way to create a simple chain in LangChain?

easy

A. chain = LLMChain(llm=llm, prompt=prompt)

B. chain = Chain(llm, prompt)

C. chain = create_chain(llm, prompt)

D. chain = LLMChain(prompt)

Solution

1. Step 1: Recall LangChain syntax for creating a simple chain

   The correct syntax uses named parameters like llm= and prompt= when creating an LLMChain.

2. Step 2: Check each option

   chain = LLMChain(llm=llm, prompt=prompt) matches the correct syntax. The other options use incorrect function or class names or miss the llm parameter.

3. Final Answer:

   chain = LLMChain(llm=llm, prompt=prompt) -> Option A

4. Quick Check:

   Use named parameters for LLMChain [OK]

Hint: Use named parameters when creating chains [OK]

Common Mistakes:

• Omitting required parameters
• Using wrong class or function names
• Passing parameters without names

3. Given this code snippet, what will be the output of result?

from langchain.chains import LLMChain
llm = SomeLLM()
prompt = "Translate English to French: {text}"
chain = LLMChain(llm=llm, prompt=prompt)
result = chain.run({"text": "Hello"})

medium

A. "Bonjour"

B. "Hello"

C. An error because of missing input

D. "Translate English to French: Hello"

Solution

1. Step 1: Understand what the chain does

   The chain uses the prompt to translate English text to French by calling the language model with the input text.

2. Step 2: Analyze the input and expected output

   The input text is "Hello", so the chain should return the French translation "Bonjour".

3. Final Answer:

   "Bonjour" -> Option A

4. Quick Check:

   Chain translates input text correctly [OK]

Hint: Chain output matches prompt task with input [OK]

Common Mistakes:

• Expecting the original text as output
• Confusing prompt string with output
• Assuming missing input causes error

4. Identify the error in this LangChain code snippet:

from langchain.chains import LLMChain
llm = SomeLLM()
prompt = "Summarize: {text}"
chain = LLMChain(llm=llm)
result = chain.run({"text": "This is a long article."})

medium

A. Calling run() without arguments

B. Incorrect input dictionary key

C. Using LLMChain instead of ComplexChain

D. Missing prompt parameter when creating the chain

Solution

1. Step 1: Check chain creation parameters

   The chain is created without the required prompt parameter, which is necessary for the chain to work.

2. Step 2: Verify input and method calls

   The input dictionary key matches the prompt placeholder, and run() is called with arguments, so no error there.

3. Final Answer:

   Missing prompt parameter when creating the chain -> Option D

4. Quick Check:

   Prompt is required when creating a chain [OK]

Hint: Always provide prompt when creating a chain [OK]

Common Mistakes:

• Forgetting to pass prompt parameter
• Assuming input keys can be arbitrary
• Calling run() without inputs

5. You want to build a LangChain that first translates English text to French, then summarizes the French text. Which approach correctly uses chains to achieve this?

hard

A. Call the language model twice manually without chains

B. Create two chains: one for translation, one for summarization, then link them sequentially

C. Use a single chain with a prompt that asks for translation and summary at once

D. Create a chain that only summarizes English text directly

Solution

1. Step 1: Understand chaining multiple tasks

   To perform two steps in order, create separate chains for each task and link them so output of first is input to second.

2. Step 2: Evaluate options for chaining

   Create two chains: one for translation, one for summarization, then link them sequentially correctly describes linking two chains sequentially. Use a single chain with a prompt that asks for translation and summary at once tries to do both in one prompt, which is less modular. Call the language model twice manually without chains skips chains. Create a chain that only summarizes English text directly misses translation step.

3. Final Answer:

   Create two chains: one for translation, one for summarization, then link them sequentially -> Option B

4. Quick Check:

   Use multiple linked chains for multi-step tasks [OK]

Hint: Link chains sequentially for multi-step tasks [OK]

Common Mistakes:

• Trying to do multiple tasks in one prompt
• Not linking chain outputs properly
• Skipping chain usage for multi-step flows