structured-output

Structured Output

Pass output_schema to LlmAgent to get a typed Pydantic object back.

from pydantic import BaseModel, Field
from orxhestra import LlmAgent
from orxhestra.events.event import Event, EventType

class CompanyAnalysis(BaseModel):
    name: str = Field(description="Company name")
    industry: str = Field(description="Primary industry")
    strengths: list[str] = Field(description="Key strengths")
    risks: list[str] = Field(description="Key risks")
    recommendation: str = Field(description="Buy, Hold, or Sell")
    confidence: float = Field(description="Confidence score 0-1")

agent = LlmAgent(
    name="AnalystAgent",
    model=model,
    tools=[get_financials, get_news_sentiment],
    output_schema=CompanyAnalysis,
    instructions="You are a financial analyst.",
)

Accessing the parsed object

async for event in agent.astream("Analyze Apple", ctx=ctx):
    if event.is_final_response():
        analysis = event.data  # CompanyAnalysis instance
        print(f"{analysis.name}: {analysis.recommendation} ({analysis.confidence:.0%})")

How it works

1. PydanticOutputParser.get_format_instructions() is appended to the system prompt.
2. PydanticOutputParser.parse() extracts and validates JSON from the response.
3. If direct parsing fails, with_structured_output() is used as a fallback.
4. Works with streaming and multi-agent compositions.