Repository navigation
v2.23.7 Release Notes
🚀 Highlights
- Agent Studio usage analytics: new usage summary and trend APIs let you query consumption statistics directly from the SDK (#294).
- Much smaller distribution jars: the Qwen tokenizer vocabulary (
qwen.tiktoken) is no longer bundled in the jars. It is resolved at runtime fromDASHSCOPE_QWEN_VOCAB, the classpath, or a SHA256-verified ModelScope download cached under~/.cache/dashscope/tokenizers(#293).
✨ Features
- agentstudio: add usage summary and trend APIs (#294)
♻️ Refactoring
- tokenizers: stop shipping
qwen.tiktokenin distribution jars; resolve the vocab at runtime via environment variable, classpath, or verified ModelScope download with local cache, mirroring dashscope-sdk-python#237 (#293)
🔧 CI & Chore
- Add automated Maven Central release workflow triggered by
v*tags, gated by themaven-centralenvironment approval; upgrademaven-gpg-pluginto 3.2.7 with loopback pinentry for headless CI signing and add anautoPublishswitch (defaultfalse) to the central publishing plugin (#295) - Bump SDK version to 2.23.7 (#293)
⚠️ Breaking Changes
- No API breaks. Behavioral change for offline environments: since
qwen.tiktokenis no longer bundled,QwenTokenizerneeds network access on first use to download the vocab from ModelScope. In air-gapped/offline setups, provide the vocab via theDASHSCOPE_QWEN_VOCABenvironment variable or placeqwen.tiktokenon the classpath (#293).
📦 Installation
Maven:
<dependency>
<groupId>com.alibaba</groupId>
<artifactId>dashscope-sdk-java</artifactId>
<version>2.23.7</version>
</dependency>Gradle:
implementation 'com.alibaba:dashscope-sdk-java:2.23.7'👥 Contributors
Thanks to all the contributors who made this release possible: