Why Chains (sequential, router) in Prompt Engineering / GenAI? - Purpose & Use Cases

The Big Idea

What if your AI could think step-by-step and decide the best path all by itself?

The Scenario

Imagine you have to handle many different tasks one after another, like answering questions, summarizing text, and translating languages, all by yourself without any help.

You try to do each task separately and then combine the results manually.

The Problem

This manual juggling is slow and confusing.

You might forget the order, mix up results, or spend too much time switching between tasks.

It's easy to make mistakes and hard to keep track of everything.

The Solution

Chains let you link tasks together so they run smoothly one after another without extra effort.

Routers help decide which task to do next based on the input, making the process smart and automatic.

This way, your work flows naturally and correctly every time.

Before vs After

✗ Before

result1 = answer_question(text)
result2 = summarize_text(result1)
final = translate(result2)

✓ After

chain = SequentialChain([answer_question, summarize_text, translate])
final = chain.run(text)

What It Enables

It makes building smart, multi-step AI workflows easy and reliable, so you can focus on solving problems instead of managing tasks.

Real Life Example

Think of a customer support bot that first understands a question, then finds the best answer, and finally translates it to the customer's language--all automatically.

Key Takeaways

Manual task handling is slow and error-prone.

Chains automate running tasks in order.

Routers smartly choose the right next step.

Practice

1. What is the main purpose of a sequential chain in GenAI?

easy

A. To run all AI steps at the same time

B. To randomly select one AI step to run

C. To run multiple AI steps one after another in order

D. To stop the AI process after the first step

Solution

1. Step 1: Understand sequential chain behavior

   A sequential chain connects AI steps so they run one after another, passing output from one to the next.

2. Step 2: Compare options to definition

   Only To run multiple AI steps one after another in order describes running steps in order, matching the sequential chain purpose.

3. Final Answer:

   To run multiple AI steps one after another in order -> Option C

4. Quick Check:

   Sequential chain = run steps in order [OK]

Hint: Sequential means steps run one after another [OK]

Common Mistakes:

• Thinking sequential means random step selection
• Confusing sequential with parallel execution
• Assuming sequential chains stop early

2. Which of the following is the correct way to define a router chain in GenAI?

easy

A. router = SequentialChain(steps=[step1, step2])

B. router = RouterChain(steps=[step1, step2], router_function=choose_step)

C. router = RouterChain(steps=step1, step2)

D. router = ChainRouter(steps=[step1, step2])

Solution

1. Step 1: Recall router chain syntax

   A router chain requires a list of steps and a router function to decide which step to run.

2. Step 2: Check each option's syntax

   router = RouterChain(steps=[step1, step2], router_function=choose_step) correctly uses RouterChain with steps list and router_function parameter. Others have wrong class names or syntax.

3. Final Answer:

   router = RouterChain(steps=[step1, step2], router_function=choose_step) -> Option B

4. Quick Check:

   RouterChain needs steps list and router_function [OK]

Hint: RouterChain needs steps list and router_function param [OK]

Common Mistakes:

• Using SequentialChain instead of RouterChain
• Passing steps without brackets as list
• Using wrong class names like ChainRouter

3. Given the code below, what will be the output?

def router_func(input_text):
    if 'weather' in input_text.lower():
        return 'weather_step'
    else:
        return 'default_step'

steps = {
    'weather_step': lambda x: 'It is sunny',
    'default_step': lambda x: 'I do not understand'
}

router_chain = RouterChain(steps=steps, router_function=router_func)

result = router_chain.run('What is the weather today?')

medium

A. 'It is sunny'

B. 'I do not understand'

C. Error: router_function missing

D. 'What is the weather today?'

Solution

1. Step 1: Analyze router function behavior

   The router_func checks if 'weather' is in the input text (case-insensitive). Input contains 'weather', so it returns 'weather_step'.

2. Step 2: Determine which step runs

   The router_chain uses 'weather_step' key to run the lambda returning 'It is sunny'.

3. Final Answer:

   'It is sunny' -> Option A

4. Quick Check:

   Input contains 'weather' -> weather_step -> 'It is sunny' [OK]

Hint: Router picks step by input keyword match [OK]

Common Mistakes:

• Ignoring case in input text check
• Confusing step keys with output strings
• Assuming default_step runs always

4. Identify the error in this router chain code snippet:

steps = {
    'step1': lambda x: 'Hello',
    'step2': lambda x: 'Bye'
}

def router_func(input_text):
    if 'hello' in input_text:
        return 'step1'
    else:
        return 'step3'

router_chain = RouterChain(steps=steps, router_function=router_func)
result = router_chain.run('hello there')

medium

A. Lambda functions require two arguments

B. steps dictionary keys are not strings

C. router_function is missing in RouterChain

D. router_func returns a step key not in steps dictionary

Solution

1. Step 1: Check router_func return values

   router_func returns 'step1' if 'hello' in input, else 'step3'. Input contains 'hello', so returns 'step1'.

2. Step 2: Verify steps dictionary keys

   Steps dictionary has keys 'step1' and 'step2', but no 'step3'. Returning 'step3' would cause error if input changed.

3. Final Answer:

   router_func returns a step key not in steps dictionary -> Option D

4. Quick Check:

   Router returns unknown step key 'step3' [OK]

Hint: Router must return keys present in steps dict [OK]

Common Mistakes:

• Ignoring missing step keys in router return
• Assuming lambda needs multiple args
• Forgetting to pass router_function parameter

5. You want to build a GenAI system that first summarizes a text, then translates it to French, but only if the text is longer than 100 words. Which chain setup is best?

hard

A. Use a sequential chain with a router function that skips translation if text is short

B. Use a router chain that chooses between summarization or translation only

C. Use two separate sequential chains running independently

D. Use a sequential chain that always runs summarization then translation

Solution

1. Step 1: Understand task requirements

   The system must summarize first, then translate only if text is longer than 100 words.

2. Step 2: Choose chain type matching conditional flow

   A sequential chain with a router function can run summarization step first, then decide to run translation step based on text length.

3. Step 3: Evaluate other options

   Router chain alone can't enforce sequential order; two separate chains won't coordinate; always running translation ignores condition.

4. Final Answer:

   Use a sequential chain with a router function that skips translation if text is short -> Option A

5. Quick Check:

   Sequential + router for conditional step flow [OK]

Hint: Combine sequential steps with router for conditional logic [OK]

Common Mistakes:

• Using router chain alone without sequence
• Running translation always ignoring condition
• Splitting steps into independent chains