Azure AI Foundry¶

# Install Azure SDK
pip install openai azure-identity

# Azure OpenAI with GPT-4
from openai import AzureOpenAI
from azure.identity import DefaultAzureCredential

client = AzureOpenAI(
    api_version="2024-02-15-preview",
    azure_endpoint="https://myresource.openai.azure.com",
    azure_ad_token_provider=DefaultAzureCredential().get_token  # (1)!
)

response = client.chat.completions.create(
    model="gpt-4-turbo",  # (2)!
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain containerization"}
    ],
    max_tokens=800,
    temperature=0.7
)

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

# Streaming responses
response = client.chat.completions.create(
    model="gpt-4-turbo",
    messages=[{"role": "user", "content": "Write a poem"}],
    stream=True  # (1)!
)

for chunk in response:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end='', flush=True)

# Function calling (tools)
import json

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get weather for a location",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"},
                "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}
            },
            "required": ["location"]
        }
    }
}]  # (1)!

response = client.chat.completions.create(
    model="gpt-4-turbo",
    messages=[{"role": "user", "content": "What's the weather in NYC?"}],
    tools=tools,
    tool_choice="auto"  # (2)!
)

# Check if function called
if response.choices[0].message.tool_calls:
    tool_call = response.choices[0].message.tool_calls[0]
    function_args = json.loads(tool_call.function.arguments)
    print(f"Calling {tool_call.function.name} with {function_args}")  # (3)!

# On Your Data (RAG without infrastructure)
response = client.chat.completions.create(
    model="gpt-4-turbo",
    messages=[{"role": "user", "content": "What are our Q4 results?"}],
    extra_body={
        "data_sources": [{
            "type": "azure_search",  # (1)!
            "parameters": {
                "endpoint": "https://mysearch.search.windows.net",
                "index_name": "financial-docs",
                "authentication": {
                    "type": "api_key",
                    "key": "search-key"
                }
            }
        }]
    }
)

print(response.choices[0].message.content)
print("\nCitations:", response.choices[0].message.context)  # (2)!

# Prompt Flow (visual workflow designer)
# Define in Azure portal, execute via SDK

from azure.ai.ml import MLClient
from azure.identity import DefaultAzureCredential

ml_client = MLClient(
    DefaultAzureCredential(),
    subscription_id="sub-id",
    resource_group_name="rg-name",
    workspace_name="workspace-name"
)

# Execute prompt flow
result = ml_client.promptflow.invoke(
    flow="customer-support-flow",  # (1)!
    data={"question": "How do I reset my password?"}
)

print(result['answer'])

# Content Safety (moderation)
from azure.ai.contentsafety import ContentSafetyClient
from azure.core.credentials import AzureKeyCredential

safety_client = ContentSafetyClient(
    endpoint="https://mysafety.cognitiveservices.azure.com",
    credential=AzureKeyCredential("key")
)

from azure.ai.contentsafety.models import AnalyzeTextOptions

result = safety_client.analyze_text(
    AnalyzeTextOptions(text="User input text here")
)  # (1)!

# Check severity (0-6, higher = more severe)
if result.hate_result.severity > 4:
    print("Content blocked: hate speech")  # (2)!

# Custom model fine-tuning
# 1. Prepare training data (JSONL format)
# 2. Upload to Azure Blob Storage
# 3. Create fine-tuning job

from azure.ai.ml.entities import FineTuningJob

job = FineTuningJob(
    model="gpt-3.5-turbo",
    training_file="azureml://datastores/workspaceblobstore/paths/train.jsonl",
    validation_file="azureml://datastores/workspaceblobstore/paths/val.jsonl",
    hyperparameters={
        "n_epochs": 3,
        "batch_size": 8,
        "learning_rate_multiplier": 0.1
    }
)

fine_tuned = ml_client.jobs.create_or_update(job)  # (1)!

# Batch processing (cost-effective for bulk)
# Upload JSONL with multiple requests
batch_input = [
    {"custom_id": "req-1", "method": "POST", "url": "/chat/completions", "body": {...}},
    {"custom_id": "req-2", "method": "POST", "url": "/chat/completions", "body": {...}}
]

# Submit batch job (50% discount vs real-time)
batch_job = client.batches.create(
    input_file_id="file-123",
    endpoint="/v1/chat/completions",
    completion_window="24h"  # (1)!
)

# Retrieve results later
results = client.batches.retrieve(batch_job.id)

// C# SDK (first-class Azure support)
using Azure.AI.OpenAI;
using Azure.Identity;

var client = new OpenAIClient(
    new Uri("https://myresource.openai.azure.com"),
    new DefaultAzureCredential()  // (1)!
);

var chatOptions = new ChatCompletionsOptions
{
    DeploymentName = "gpt-4-turbo",
    Messages =
    {
        new ChatRequestSystemMessage("You are helpful"),
        new ChatRequestUserMessage("Explain async/await")
    },
    MaxTokens = 800
};

var response = await client.GetChatCompletionsAsync(chatOptions);
Console.WriteLine(response.Value.Choices[0].Message.Content);  // (2)!