หมายเหตุ
การเข้าถึงหน้านี้ต้องได้รับการอนุญาต คุณสามารถลอง ลงชื่อเข้าใช้หรือเปลี่ยนไดเรกทอรีได้
การเข้าถึงหน้านี้ต้องได้รับการอนุญาต คุณสามารถลองเปลี่ยนไดเรกทอรีได้
Tip
An alternate Foundry project quickstart is available: Quickstart: Get started with Microsoft Foundry (Foundry projects).
This quickstart sets up your local environment for hub-based projects, deploys a model, and builds a simple traced/evaluable chat script.
Prerequisites
- Azure subscription.
- Existing hub project (or create one). If not, consider using a Foundry project quickstart.
Set up your development environment
- Install prerequisites (Python, Azure CLI, login).
- Install packages:
pip install azure-ai-inference azure-identity azure-ai-projects==1.0.0b10
Different project types need distinct azure-ai-projects versions. Keep each project in its own isolated environment to avoid conflicts.
Deploy a model
- Portal: Sign in, open hub project.
- Model catalog: select gpt-4o-mini.
- Use this model > accept default deployment name > Deploy.
- After success: Open in playground to verify.
Build your chat app
Create chat.py with sample code:
from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential
project_connection_string = "<your-connection-string-goes-here>"
project = AIProjectClient.from_connection_string(
conn_str=project_connection_string, credential=DefaultAzureCredential()
)
chat = project.inference.get_chat_completions_client()
response = chat.complete(
model="gpt-4o-mini",
messages=[
{
"role": "system",
"content": "You are an AI assistant that speaks like a techno punk rocker from 2350. Be cool but not too cool. Ya dig?",
},
{"role": "user", "content": "Hey, can you help me with my taxes? I'm a freelancer."},
],
)
print(response.choices[0].message.content)
Insert your project connection string from the project Overview page (copy, replace placeholder in code).
Run:
python chat.py
Add prompt templating
Add get_chat_response using mustache template (see chat-template.py sample) then invoke with user/context messages.
Run again to view templated response.
Clean up resources
Delete deployment or project when done to avoid charges.