Skip to content

Structured Output Parser.md

Latest commit

 

History

History
152 lines (105 loc) · 4.44 KB

Structured Output Parser.md

File metadata and controls

152 lines (105 loc) · 4.44 KB

🗃️ Structured Output Parser – Notes

Notes on the StructuredOutputParser — a schema-based parser that returns fixed-field dictionaries.

Note: StructuredOutputParser is an older LangChain abstraction for schema-based outputs. It has largely been superseded by JsonOutputParser and PydanticOutputParser in newer versions.

Table of Contents

Related Notes

What is Structured Output Parser?

A Structured Output Parser forces an LLM to return output in a pre-defined structure (schema) with:

  • Fixed fields
  • Fixed data types
  • Validated output

It ensures the model output is machine-readable and reliable.

Why Do We Need Structured Output?

Free-text output is:

  • Unpredictable
  • May miss fields
  • May change format

Real applications (APIs, forms, agents) need:

  • Consistent keys
  • Correct data types
  • Validation

👉 Structured Output Parsers solve this.

Types of Structured Output Parsers in LangChain

Parser Schema Definition Output Validation
StructuredOutputParser ResponseSchema dict ❌
JsonOutputParser Format instructions dict ⚠️ Partial
PydanticOutputParser Pydantic model Object ✅

StructuredOutputParser: Core Concept

What it does

  1. Defines expected fields using ResponseSchema
  2. Injects formatting instructions into the prompt
  3. Parses LLM output into a Python dictionary

Usage Example

from langchain.output_parsers import StructuredOutputParser, ResponseSchema
from langchain_core.prompts import PromptTemplate
from langchain_ollama import ChatOllama

schemas = [
    ResponseSchema(name="topic", description="Main topic"),
    ResponseSchema(name="summary", description="Short explanation"),
    ResponseSchema(name="difficulty", description="Easy / Medium / Hard"),
]

parser = StructuredOutputParser.from_response_schemas(schemas)

format_instructions = parser.get_format_instructions()

prompt = PromptTemplate(
    template="Explain {topic}\n{format_instructions}",
    input_variables=["topic"],
    partial_variables={"format_instructions": format_instructions}
)

llm = ChatOllama(model="llama3")
chain = prompt | llm | parser

result = chain.invoke({"topic": "Black Holes"})
# → {"topic": "Black Hole", "summary": "...", "difficulty": "Medium"}
print(type(result))  # → <class 'dict'>

When to Use It

Use StructuredOutputParser when you need:

  • Fixed fields in the output
  • Consistent format without strict type validation
  • Legacy LangChain compatibility

Examples:

  • Question–answer format
  • Summary + keywords
  • Explanation + difficulty level
  • Tool input formatting

Limitations

  • ❌ Does not strictly validate data types
  • ❌ Model may still hallucinate values
  • ❌ Less powerful than PydanticOutputParser

👉 For strict control → PydanticOutputParser

Comparison with Other Parsers

Parser Structure Validation Use Case
StrOutputParser ❌ ❌ Simple text
JsonOutputParser ⚠️ ❌ Loose JSON
StructuredOutputParser ✅ ❌ Fixed fields
PydanticOutputParser ✅ ✅ Production

Memory Trick 🧠

  • Text → StrOutputParser
  • JSON → JsonOutputParser
  • Fields → StructuredOutputParser
  • Types → PydanticOutputParser

One-line definition: Structured Output Parser enforces a predefined response schema on LLM output, returning structured data as a dictionary.