PromptTemplate structures user inputs into system/user/assistant messages. Output parsers turn LLM text into structured objects (JSON, dataclasses, Pydantic models). This post covers templating syntax, parsing strategies, and error handling.
Prompt Templates
from langchain.prompts import PromptTemplate, ChatPromptTemplate
# Simple template template = "Translate '{input_language}' to '{output_language}': {text}" prompt = PromptTemplate( input_variables=["input_language", "output_language", "text"], template=template )
formatted = prompt.format( input_language="English", output_language="Spanish", text="Hello, world!" ) print(formatted)
Output:
Translate 'English' to 'Spanish': Hello, world!
Chat Templates
from langchain.prompts import ChatPromptTemplate
prompt = ChatPromptTemplate.from_messages([ ("system", "You are a helpful assistant."), ("human", "{user_input}"), ])
formatted = prompt.format_messages(user_input="What time is it?") for msg in formatted: print(f"{msg.type}: {msg.content}")
Output:
system: You are a helpful assistant.
human: What time is it?
Output Parsing
from langchain.output_parsers import PydanticOutputParser
from pydantic import BaseModel
class Answer(BaseModel): answer: str confidence: float
parser = PydanticOutputParser(pydantic_object=Answer) parser_instructions = parser.get_format_instructions() print(parser_instructions)
Output:
The output should be formatted as a JSON instance...
Conclusion
Templates structure inputs; parsers extract structured outputs. Together they enable predictable, composable workflows. Understanding templating and parsing patterns is essential for building reliable LLM applications. Next: we'll explore working with LLMs and Chat Models—how to choose, configure, and switch between providers.