Skip to main content

@decision Decorator

The @decision decorator marks methods as LLM-powered decision points. Decisions are the primary way agents interact with language models.

Basic Usage​

from waxell_sdk import agent, decision

@agent(name="classifier")
class Classifier:

    @decision
    def classify_intent(self, ctx):
        """Classify the user's intent."""
        return ctx.llm.classify(
            ctx.input.message,
            categories=["question", "complaint", "request", "feedback"]
        )

Decision Types​

Classification​

@decision
def classify(self, ctx):
    return ctx.llm.classify(
        text=ctx.input.text,
        categories=["positive", "negative", "neutral"]
    )

Generation​

@decision
def generate_response(self, ctx):
    return ctx.llm.generate(
        prompt=f"Respond to: {ctx.input.message}",
        max_tokens=500
    )

Structured Output​

from pydantic import BaseModel

class Analysis(BaseModel):
    sentiment: str
    confidence: float
    topics: list[str]

@decision
def analyze(self, ctx):
    return ctx.llm.generate(
        prompt=f"Analyze: {ctx.input.text}",
        response_model=Analysis
    )

Governance​

Decisions are subject to governance policies:

  • Token Limits: Maximum tokens per decision
  • Model Restrictions: Allowed models per agent
  • Content Filters: Block sensitive content

Next Steps​