Structured Outputs

Direct models to return data that conforms to a specific JSON schema.

Structured Outputs ensure that models follow your supplied JSON schema. Portkey supports this for Anthropic models using a unified response_format interface.

Define object schemas using Pydantic (Python) or Zod (JavaScript) to extract structured information from unstructured text.

1. Using Pydantic or Zod (Recommended)

This approach provides type hinting and automatic validation.

Portkey Python SDK

from portkey_ai import Portkey
from pydantic import BaseModel

class Step(BaseModel):
    explanation: str
    output: str

class MathReasoning(BaseModel):
    steps: list[Step]
    final_answer: str

portkey = Portkey(
    api_key="PORTKEY_API_KEY"
)

# Use .parse() for automatic parsing
completion = portkey.chat.completions.parse(
    model="@anthropic-testing/claude-sonnet-4-6",
    messages=[
        {"role": "system", "content": "You are a helpful math tutor. Guide the user through the solution step by step."},
        {"role": "user", "content": "how can I solve 8x + 7 = -23"}
    ],
    response_format=MathReasoning
)

print(completion.choices[0].message.parsed)

2. Using Raw JSON Schema

For cross-language compatibility or dynamic schemas, pass a standard JSON schema directly.

Portkey Python SDK

from portkey_ai import Portkey

portkey = Portkey(
    api_key="PORTKEY_API_KEY"
)

completion = portkey.chat.completions.create(
    model="@anthropic-provider/claude-3-5-sonnet-20241022",
    messages=[
        {"role": "system", "content": "Extract the event information."},
        {"role": "user", "content": "Alice and Bob are going to a science fair on Friday."}
    ],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "event_extraction",
            "schema": {
                "type": "object",
                "properties": {
                    "location": { "type": "string" },
                    "date": { "type": "string" },
                    "participants": { "type": "array", "items": { "type": "string" } }
                },
                "required": ["location", "date", "participants"],
                "additionalProperties": False
            },
            "strict": True
        }
    }
)

print(completion.choices[0].message.content)

For more, refer to Anthropic’s detailed documentation on Structured Outputs here.