A programming framework for agentic AI
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Structured output using GPT-4o models\n",
"\n",
"This cookbook demonstrates how to obtain structured output using GPT-4o models. The OpenAI beta client SDK provides a parse helper that allows you to use your own Pydantic model, eliminating the need to define a JSON schema. This approach is recommended for supported models.\n",
"\n",
"Currently, this feature is supported for:\n",
"\n",
"- gpt-4o-mini on OpenAI\n",
"- gpt-4o-2024-08-06 on OpenAI\n",
"- gpt-4o-2024-08-06 on Azure"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's define a simple message type that carries explanation and output for a Math problem"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"from pydantic import BaseModel\n",
"\n",
"\n",
"class MathReasoning(BaseModel):\n",
" class Step(BaseModel):\n",
" explanation: str\n",
" output: str\n",
"\n",
" steps: list[Step]\n",
" final_answer: str"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"\n",
"# Set the environment variable\n",
"os.environ[\"AZURE_OPENAI_ENDPOINT\"] = \"https://YOUR_ENDPOINT_DETAILS.openai.azure.com/\"\n",
"os.environ[\"AZURE_OPENAI_API_KEY\"] = \"YOUR_API_KEY\"\n",
"os.environ[\"AZURE_OPENAI_DEPLOYMENT_NAME\"] = \"gpt-4o-2024-08-06\"\n",
"os.environ[\"AZURE_OPENAI_API_VERSION\"] = \"2024-08-01-preview\""
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"import os\n",
"from typing import Optional\n",
"\n",
"from autogen_core.models import UserMessage\n",
"from autogen_ext.models.openai import AzureOpenAIChatCompletionClient\n",
"\n",
"\n",
"# Function to get environment variable and ensure it is not None\n",
"def get_env_variable(name: str) -> str:\n",
" value = os.getenv(name)\n",
" if value is None:\n",
" raise ValueError(f\"Environment variable {name} is not set\")\n",
" return value\n",
"\n",
"\n",
"# Create the client with type-checked environment variables\n",
"client = AzureOpenAIChatCompletionClient(\n",
" azure_deployment=get_env_variable(\"AZURE_OPENAI_DEPLOYMENT_NAME\"),\n",
" model=get_env_variable(\"AZURE_OPENAI_MODEL\"),\n",
" api_version=get_env_variable(\"AZURE_OPENAI_API_VERSION\"),\n",
" azure_endpoint=get_env_variable(\"AZURE_OPENAI_ENDPOINT\"),\n",
" api_key=get_env_variable(\"AZURE_OPENAI_API_KEY\"),\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"{'steps': [{'explanation': 'Start by aligning the numbers vertically.', 'output': '\\n 16\\n+ 32'}, {'explanation': 'Add the units digits: 6 + 2 = 8.', 'output': '\\n 16\\n+ 32\\n 8'}, {'explanation': 'Add the tens digits: 1 + 3 = 4.', 'output': '\\n 16\\n+ 32\\n 48'}], 'final_answer': '48'}\n"
]
},
{
"data": {
"text/plain": [
"MathReasoning(steps=[Step(explanation='Start by aligning the numbers vertically.', output='\\n 16\\n+ 32'), Step(explanation='Add the units digits: 6 + 2 = 8.', output='\\n 16\\n+ 32\\n 8'), Step(explanation='Add the tens digits: 1 + 3 = 4.', output='\\n 16\\n+ 32\\n 48')], final_answer='48')"
]
},
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Define the user message\n",
"messages = [\n",
" UserMessage(content=\"What is 16 + 32?\", source=\"user\"),\n",
"]\n",
"\n",
"# Call the create method on the client, passing the messages and additional arguments\n",
"# The extra_create_args dictionary includes the response format as MathReasoning model we defined above\n",
"# Providing the response format and pydantic model will use the new parse method from beta SDK\n",
"response = await client.create(messages=messages, extra_create_args={\"response_format\": MathReasoning})\n",
"\n",
"# Ensure the response content is a valid JSON string before loading it\n",
"response_content: Optional[str] = response.content if isinstance(response.content, str) else None\n",
"if response_content is None:\n",
" raise ValueError(\"Response content is not a valid JSON string\")\n",
"\n",
"# Print the response content after loading it as JSON\n",
"print(json.loads(response_content))\n",
"\n",
"# Validate the response content with the MathReasoning model\n",
"MathReasoning.model_validate(json.loads(response_content))"
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.11.7"
}
},
"nbformat": 4,
"nbformat_minor": 2
}