v2.22.26

What's Changed

Features

• CosyVoice: Support cancel in finish-task
• MultiModal Dialog: Support payload.parameters when requestToRespond

Bug Fixes

• Omni Realtime: Add onError() callback for better error handling
• Fix potential NullPointerException and synchronization race conditions

Details

• SpeechSynthesizer / SpeechSynthesizerV2: Improved finish-task handling with cancel support
• OmniRealtimeConversation: Added error callback propagation
• MultiModalDialog: Added requestToRespond with payload parameters support
• FullDuplexServiceParam / FullDuplexRequest: Enhanced parameter passing

---

Contributors

songguocola

v2.22.25

✨ What's New
Image Generation: Added color palette support and sketch-based image generation parameters
Multimodal Streaming: Enhanced with incremental/full mode selection and improved reasoning_content handling
Tool Calls: Added support for quark_search and code_interpreter in MultiModalMessageAdapter
Parameter Validation: Now validates mutual exclusivity of stopStrings and stopTokens with clear error messages

🐛 Fixes
Removed hardcoded URLs in TingWu to support DASHSCOPE_HTTP_BASE_URL environment variable
Fixed NPE issues in stream handling
Prevented automatic incremental_output injection to avoid response merging issues

📋 Other Changes
Added NOTICE file for third-party license compliance (EPL-2.0 components)
Code quality improvements and formatting updates

👥 Contributors
luk384090-cloud
lzsweb

v2.22.24

Release Notes — v2.22.24

🆕 Features

• AgentStudio: Add AgentStudio sub-module for managed agents (#236)
• ASR: Add speaker_id parameter to speech recognition results (#241)

🐛 Bug Fixes

• Generation: Fix searchInfo being lost in streaming generation responses (#243)
• Multimodal Agent: Fix bug in multimodal-dialog agent (#241)

👥 Contributors

• @foleydang (#236)
• @songguocola (#241)
• @lzsweb (#243)

v2.18.6

Hotfix for incrementalOutput serialization

What's Changed

• Fix incrementalOutput parameter serialization to support explicit false value
• When incrementalOutput is set to false, the parameter is now correctly sent to the server
• Enables full output mode (non-streaming accumulation) for qwen-vl-plus and other models

Migration

Users on v2.18.0 experiencing streaming output issues can upgrade to v2.18.6 to get full output mode by setting .incrementalOutput(false) in MultiModalConversationParam.builder().

Full Changelog: v2.18.0...v2.18.6

v2.22.23

What's Changed

• funasr support payload.input and bug fix in transcription by @songguocola in #232
• Add parentNodeId field to WorkflowMessage by @lzsweb in #233
• Add latest model constants (#229) by @lzsweb in #234

Full Changelog: v2.22.22...v2.22.23

Additional Notes

New Features

• Add latest model constants (PR #234)
• Add parentNodeId field to WorkflowMessage (PR #233)

Bug Fixes

• Fix funasr transcription bugs and support payload.input (PR #232)

v2.22.22

What's Changed

• Add doc_url content type support in MessageAdapter (#123)
  • New classes: DocURL, MessageContentDocURL
  • Serialization/deserialization support in MessageAdapter
• Update version to 2.22.22

Full Changelog: v2.22.21...v2.22.22

v2.22.20

Release Notes — v2.22.20

New Features

Plugins usage information in API responses

Added plugins field to GenerationUsage and MultiModalConversationUsage, exposing search plugin details (count and strategy) returned by the DashScope API.

• GenerationUsage.getPlugins() → returns Plugins with nested Search object containing count (Integer) and strategy (String)
• MultiModalConversationUsage.getPlugins() → same structure

Full Changelog: v2.22.19...v2.22.20

v2.22.19

Highlights

• New: headers: Map<String, Object> field exposed on all concrete *Result classes (GenerationResult, MultiModalConversationResult, ConversationResult, CodeGenerationResult, ImageGenerationResult, ImageSynthesisResult, VideoSynthesisResult) — lets callers read response headers (e.g. x-request-id) for tracing. (#215)
• Fix: response headers are now also populated when isFlattenResult=true (previously skipped).
• Breaking: Result#headers field type widened from Map<String, String> to Map<String, Object>. Source-compatible for .get(key) usage; explicit Map<String, String> declarations need to be updated.

What's Changed

• Add a header to all Result classes. by @mose-zm in #215
• DEV/Extension/Plugin by @luk384090-cloud in #216

New Contributors

• @luk384090-cloud made their first contribution in #216

Full Changelog: v2.22.18...v2.22.19

v2.22.18

What's Changed

• (model/cosyvoice): support flush api by @songguocola in #213

Full Changelog: v2.22.17...v2.22.18

v2.22.17

What's Changed

• param build fixes for audio models by @songguocola in #208
• Add hh model usage parameters by @mose-zm in #209

Full Changelog: v2.22.16...v2.22.17