1.0.0-beta.14 (2025-02-21)
Features Added
-
For
ChatCompletionOptions
:store
(bool), which enables saving outputs for FT (distillation)metadata
(string/string map), used withstore
reasoning_effort
(enum string,low
,medium
,high
) to configure how hardo1
thinksdeveloper
message role (system
replacement foro1
and newer models) now available. Other than the name, it's identicalprediction
added for predicted outputsmodalities
added ("text" | "audio") for audio input supportaudio
added for audio content options (voice, format, etc.)- user role message content part representation for audio input added
- AZURE-ONLY User Security Context (for Defender for Cloud Integration)
- Azure chat request bodies now have a user_security_context property
- These are JSON objects that have application_name, end_user_id, end_user_tenant_id, and source_ip, all optional string properties
-
For
ChatCompletions
:- audio output added as
ChatMessageAudioContentItem
- parity usage updates:
PromptTokensDetails
getsAudioTokens
CompletionTokensDetails
getsAcceptedPredictionTokens
,RejectedPredictionTokens
, andAudioTokens
- audio output added as
Other Changes
Dependency Updates
- Upgraded
azure-core
to version1.55.2
. - Upgraded
azure-core-http-netty
to version1.15.10
.