Why LangChain? Prompt Templates and Chat Models

Calling an LLM API directly (as in the RAG and AI Agent courses) works, but every real application ends up rebuilding the same pieces: swappable prompts, a consistent interface across model providers, and a way to chain steps together. LangChain is a Python framework that standardizes those pieces so you're not reinventing them for every project.

Chat models: one interface, many providers

from langchain_openai import ChatOpenAI

model = ChatOpenAI(model='gpt-4o-mini', api_key='YOUR_API_KEY')
response = model.invoke('What is a large language model?')
print(response.content)

This needs a real API key and network access, so it's read-only here. The point of ChatOpenAI (and its siblings, ChatAnthropic, ChatGoogleGenerativeAI, and others) is that they all implement the same .invoke() interface, swapping providers later means changing one constructor call, not rewriting your application logic.

Prompt templates

Hardcoding a prompt as a plain string works until you need to reuse it with different inputs, a PromptTemplate (or ChatPromptTemplate for chat models) separates the fixed wording from the variables that change per call:

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ('system', 'You are a helpful assistant that explains topics simply.'),
    ('human', 'Explain {topic} in two sentences.'),
])

messages = prompt.invoke({'topic': 'recursion'})
print(messages)

{topic} is a placeholder filled in by whatever dictionary you pass to .invoke(), the same template can be reused with {'topic': 'closures'} or {'topic': 'binary search'} without touching the wording itself.

Why this matters over raw API calls

• Consistency: the same .invoke() shape works whether the underlying model is from OpenAI, Anthropic, or a local model.
• Reusability: a prompt template is a first-class object you can store, version, and reuse across an app, instead of a string interpolated in twelve different places.
• Composability: templates and models are designed to be chained together, which is exactly what the next lesson builds on.