官方技术资料:Microsoft(第 14 页)
官方技术资料 · Microsoft:共 312 条,本页为第 196–210 条。
Set up human evaluation for your agents (preview)
Learn how to set up human evaluation for your Microsoft Foundry agents, create templates, and analyze results to improve agent performance.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:model-evaluation · llm
链接复核时间:2026-09-24
Log end user feedback (preview)
Learn how to log end user feedback such as thumbs up/down or rating scales from your AI application using OpenTelemetry semantics in Microsoft Foundry.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Optimize agent prompts by using Prompt Optimizer (preview)
Learn how to use Prompt Optimizer in Microsoft Foundry to automatically improve your agent's system instructions using AI-driven prompt engineering best practices.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Configure tracing for AI agent frameworks (preview)
Debug issues and monitor AI agent performance in production by configuring OpenTelemetry tracing for LangChain, LangGraph, Semantic Kernel, and OpenAI Agents SDK.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Review agent interactions with Trace Replay (preview)
Learn how to use Trace Replay in Microsoft Foundry to navigate, filter, and analyze conversation traces from agent runs.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Set up tracing in Microsoft Foundry
Learn how to set up tracing in Microsoft Foundry to debug AI agent runs and monitor behavior by sending telemetry to Azure Monitor Application Insights with OpenTelemetry.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Annotate traces with human feedback (preview)
Learn how to annotate traces with thumbs up or thumbs down feedback in Microsoft Foundry to track quality signals directly on production traces.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Restrict access to sensitive content in Microsoft Foundry traces (preview)
Learn how to protect sensitive generative AI trace data in Microsoft Foundry by routing it to a dedicated table and restricting access with Azure role-based access control.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
Convert agent traces into evaluation datasets (preview)
Learn how to use data generation in Microsoft Foundry to turn production agent traces into evaluation and fine-tuning datasets.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:model-evaluation · llm
链接复核时间:2026-09-24
Abuse monitoring
Learn about the abuse monitoring capabilities of Azure OpenAI
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:observability · deployment
链接复核时间:2026-09-24
System message design
Learn how system messages shape Azure OpenAI chat responses, with best practices for control, safety, and consistency.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:prompt-engineering · system-prompt
链接复核时间:2026-09-24
Groundedness detection filter
Learn how to use groundedness detection in Azure OpenAI in Microsoft Foundry Models to validate LLM responses against source documents for RAG scenarios.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:guardrails · governance
链接复核时间:2026-09-24
Personally identifiable information (PII) filter
Learn about the Personally Identifiable Information (PII) filter for identifying and flagging known personal information in large language model outputs.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:guardrails · governance
链接复核时间:2026-09-24
Prompt Shields in Microsoft Foundry
Learn how Prompt Shields in Microsoft Foundry detect and block user prompt attacks and document attacks on your model deployments and agents.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:prompt-injection · guardrails · governance
链接复核时间:2026-09-24
Protected material detection filter
Learn about the Protected Material Detection Filter for identifying and flagging known protected text and code content in large language model outputs.
证据用途:用于理解、实现、选型、可靠性、安全或迁移判断。
使用边界:只直接证明 Microsoft 的实现与声明;预览功能和产品生命周期使用前应复核
关联概念:guardrails · governance
链接复核时间:2026-09-24