Azure OpenAI + Log Analytics – Supported Audit Logs & Setup Guidance

CHIEN YU HSIEN 20 信譽點數
2025-06-20T05:41:54.48+00:00

We are planning to integrate our internal system with Azure OpenAI API, and aim to enable auditing and monitoring using Log Analytics.

Could you please confirm whether the following audit items are supported natively through Azure OpenAI diagnostics or logging features?

Audit Item Description
Who made the request? Identify which account or system identity (e.g., Managed Identity, Subscription Key) initiated the call
Who made the request? Identify which account or system identity (e.g., Managed Identity, Subscription Key) initiated the call
When and from where? Timestamp and source IP or region
What was requested? Prompt content or API type (e.g., chat, embedding)
Which model and how many tokens? Model name (e.g., GPT-4o) and token usage per request
Anomalies detected? Abuse detection via Azure Defender or throttling/overuse alerts
Log retention Default retention period and how to extend (e.g., to 1–7 years via Storage/Archive)

If any of these are not directly supported, are there alternative tools or best practices (e.g., API Management, Application Insights, custom logging) to achieve similar visibility?

Additionally, we would greatly appreciate:

Step-by-step setup instructions to enable and query these logs

Links to relevant documentation or sample queries

Any recommended templates or dashboards (e.g., Workbooks)

Thank you for your assistance

Azure
Azure

雲端運算平台和基礎結構，可透過 Microsoft 管理的資料中心全球網路來建置、部署和管理應用程式與服務。

您的回答

答案可由問題作者標示為「已接受」，而由仲裁者標示為「推薦」，這可協助使用者知道答案解決了作者的問題。