LangChain vs direct API calls - Practice Questions

Challenge - 5 Problems

❓ component_behavior

intermediate

What is the main difference in output when using LangChain's chain vs direct API call?

Consider a LangChain chain that calls an LLM with a prompt template versus a direct API call to the same LLM with the same prompt. What difference in output behavior would you expect?

LangChain

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

llm = OpenAI(temperature=0)
prompt = PromptTemplate(template="Translate '{text}' to French.", input_variables=["text"])
chain = LLMChain(llm=llm, prompt=prompt)

result_chain = chain.run(text="Hello")

# Direct API call
result_direct = llm("Translate 'Hello' to French.")

A LangChain output is a Python dict, direct API call returns a string.

B LangChain output includes extra metadata, while direct API call returns only the raw text.

C LangChain output is always JSON, direct API call returns plain text.

D Both outputs are identical strings because LangChain just wraps the API call with formatting.

💡 Hint

Think about what LangChain adds on top of the API call.

LangChain's LLMChain formats the prompt and calls the LLM API under the hood. The output string is the same as the direct API call with the same prompt and parameters.

❓ state_output

intermediate

What happens to state when using LangChain chains vs direct API calls?

When you run multiple prompts in sequence, how does LangChain manage state compared to direct API calls?

LangChain

from langchain import ConversationChain
from langchain.llms import OpenAI

llm = OpenAI(temperature=0)
conv = ConversationChain(llm=llm)

response1 = conv.predict(input="Hello!")
response2 = conv.predict(input="How are you?")

# Direct calls
resp1 = llm("Hello!")
resp2 = llm("How are you?")

A Direct API calls keep conversation history; LangChain chains do not.

B Both LangChain and direct API calls keep conversation history automatically.

C LangChain ConversationChain keeps conversation history automatically; direct API calls do not.

D Neither LangChain nor direct API calls keep conversation history automatically.

💡 Hint

Think about how conversation context is preserved.

LangChain's ConversationChain stores previous messages internally to provide context. Direct API calls are stateless unless you manually include prior messages.

📝 Syntax

advanced

Which code snippet correctly creates a LangChain chain that calls OpenAI with a prompt template?

Identify the correct code to create a LangChain LLMChain with a prompt template that asks for a summary.

A

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

prompt = PromptTemplate(template="Summarize: {text}", input_variables=["text"])
llm = OpenAI()
chain = LLMChain(llm=llm, prompt=prompt)

B

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

prompt = PromptTemplate(template="Summarize: {text}", input_variables=["text"])
llm = OpenAI()
chain = LLMChain(prompt, llm)

C

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

prompt = PromptTemplate(template="Summarize: {text}", input_variables=["text"])
llm = OpenAI()
chain = LLMChain(llm=llm)

D

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

prompt = PromptTemplate(template="Summarize: {text}", input_variables=["text"])
llm = OpenAI()
chain = LLMChain(prompt=prompt)

💡 Hint

Check the order and presence of required arguments in LLMChain constructor.

LLMChain requires both llm and prompt arguments. The correct order is llm first, then prompt.

🔧 Debug

advanced

Why does this LangChain code raise a KeyError when running a chain?

Given this code, why does running chain.run() raise a KeyError?

LangChain

from langchain import LLMChain, PromptTemplate
from langchain.llms import OpenAI

prompt = PromptTemplate(template="Answer: {question}", input_variables=["question"])
llm = OpenAI()
chain = LLMChain(llm=llm, prompt=prompt)

result = chain.run()

A Because OpenAI LLM is not initialized with an API key.

B Because <code>run()</code> requires the 'question' argument but none was provided.

C Because <code>LLMChain</code> does not support <code>run()</code> method.

D Because the prompt template is invalid and missing a colon.

💡 Hint

Check the input variables required by the prompt template.

The prompt template expects a 'question' variable. Calling run() without arguments causes a KeyError for missing 'question'.

🧠 Conceptual

expert

Which advantage does LangChain provide over direct API calls for building complex applications?

Choose the best advantage LangChain offers compared to making direct API calls to language models.

A LangChain automatically manages prompt templates, chains, memory, and integrations, simplifying complex workflows.

B LangChain reduces API latency by caching all responses locally.

C LangChain replaces the need for API keys by using open-source models only.

D LangChain guarantees 100% accurate outputs from language models.

💡 Hint

Think about what LangChain adds beyond just calling the API.

LangChain provides tools to build chains, manage memory, use prompt templates, and integrate with other services, making complex app building easier.