Skip to main content

Structured outputs

Get typed responses instead of raw strings.

Basic usage​

Define a dataclass and pass it to run():

from dataclasses import dataclass
from pure_agents import Agent

@dataclass
class Sentiment:
    sentiment: str
    confidence: float
    summary: str

agent = Agent()
result = await agent.run(
    "Analyse: I absolutely love this product!",
    output=Sentiment
)

print(result.sentiment)   # "positive"
print(result.confidence)  # 0.95
print(result.summary)     # "Strong positive sentiment..."

Supported types​

Dataclass fields can be:

  • str: strings
  • int: integers
  • float: numbers
  • bool: booleans
@dataclass
class Recipe:
    name: str
    prep_time_minutes: int
    is_vegetarian: bool
    rating: float

How it works​

pureagents automatically:

  1. Converts your dataclass to a JSON schema
  2. Adds instructions to the prompt asking for JSON output
  3. Parses the response into your dataclass

No magic. You can see exactly what's happening with debug=True.

Error handling​

If the LLM returns malformed JSON:

try:
    result = await agent.run("...", output=MyClass)
except json.JSONDecodeError:
    # Handle invalid JSON
    pass

When to use​

  • Extracting structured data from text
  • Building APIs that need typed responses
  • Data pipelines with validated outputs
  • Any time you want guarantees about response format