LangChain / RAG / Vector DBs · Lesson

Building Custom Output Parsers

Transform raw LLM text into reliable structured data with custom and built-in LangChain output parsers.

Building Custom Output Parsers is a free LangChain / RAG / Vector DBs lesson on CoddyKit — lesson 4 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the LangChain / RAG / Vector DBs learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.

Why Parse Output?

LLMs return free-form text, but applications need structured data: objects, lists, enums. An output parser converts the raw string into something your code can use safely.

The Parser Interface

A LangChain output parser implements two key methods.

parse(text) turns the string into your type
get_format_instructions() tells the model how to format its reply

A Minimal Custom Parser

Subclass BaseOutputParser and implement parse. This one splits a comma-separated reply into a list.

from langchain_core.output_parsers import BaseOutputParser

class CommaListParser(BaseOutputParser):
    def parse(self, text):
        return [t.strip() for t in text.split(",")]

print(CommaListParser().parse("a, b, c"))  # ["a", "b", "c"]

Format Instructions

Override get_format_instructions so the prompt nudges the model toward a parseable shape. The instructions are injected into your prompt template.

def get_format_instructions(self):
    return "Reply with items separated by commas, no numbering."

Pydantic Output Parser

For rich objects, LangChain ships a PydanticOutputParser. You define a schema and it generates instructions plus validation automatically.

from pydantic import BaseModel

class Person(BaseModel):
    name: str
    age: int

# parser = PydanticOutputParser(pydantic_object=Person)

Wiring a Parser into a Chain

With LCEL you pipe the model output straight into the parser using the | operator.

chain = prompt | llm | CommaListParser()
result = chain.invoke({"topic": "fruits"})

Handling Malformed Output

Models sometimes ignore the format. A robust parser validates and raises a clear error, or attempts a best-effort recovery, instead of crashing downstream.

def parse(self, text):
    try:
        return json.loads(text)
    except json.JSONDecodeError:
        raise ValueError("Model did not return valid JSON")

The OutputFixingParser

Wrap any parser in an OutputFixingParser. When parsing fails, it sends the bad output and the error back to an LLM to repair it.

Streaming and Parsers

Some parsers support incremental parsing of streamed tokens. For structured types this is hard, so streaming parsers often emit partial objects as fields complete.

Validation Beyond Types

You can add business rules inside parse: ranges, allowed values, required combinations. Reject anything that would corrupt your application state.

def parse(self, text):
    n = int(text.strip())
    if not 1 <= n <= 5:
        raise ValueError("Rating must be 1-5")
    return n

Putting It Together

Define the parser, expose format instructions, inject them into the prompt, and pipe everything in a chain. The result is reliable structured output.

prompt = template.partial(
    format=parser.get_format_instructions()
)
chain = prompt | llm | parser
chain.invoke({"input": "..."})

Quick Check

Test your understanding of output parsers.

Recap

You built custom output parsing:

Implement parse and get_format_instructions
Use PydanticOutputParser for rich schemas
Pipe parsers into chains with |
Wrap with OutputFixingParser for resilience

Frequently asked questions

Is the “Building Custom Output Parsers” lesson free?

Yes — the full text of “Building Custom Output Parsers” is free to read here on the web, and the LangChain / RAG / Vector DBs course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the LangChain / RAG / Vector DBs course, upgrade to CoddyKit PRO.

What will I learn in “Building Custom Output Parsers”?

Transform raw LLM text into reliable structured data with custom and built-in LangChain output parsers. You practise LangChain / RAG / Vector DBs with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.

Do I need any experience to start LangChain / RAG / Vector DBs?

No prior experience is required. LangChain / RAG / Vector DBs on CoddyKit is structured for beginners through advanced learners; this is — lesson 4 of 4, so you can start here or from the beginning and move at your own pace.

How long does the “Building Custom Output Parsers” lesson take?

Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.

Can I write and run code in this LangChain / RAG / Vector DBs lesson?

Yes. Every LangChain / RAG / Vector DBs lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.

All lessons in this course

Developing Custom Document Loaders
Integrating Custom Embedding Models
Extending Retrieval Chains with Custom Logic
Building Custom Output Parsers

← Back to LangChain / RAG / Vector DBs