Skip to content

Release Notes

The authoritative, per-release changelog is published on GitHub Releases. This page summarizes the major lines and links to recent tags. Current SDK version: 2.1.1.

Recent releases

Tag Notes
v2.1.0 See release page
v2.0.4 See release page
v2.0.3 See release page
v2.0.2 See release page
v2.0.1 See release page

2.0 highlights

Version 2.0 introduced the resource-style client modeled on the ergonomics of the OpenAI SDK:

  • Entry point — client.chat.completions.create(...) instead of module-level static classes; one-liner dashscope.chat("qwen-plus", "...").
  • Responses — typed objects (resp.output_text, resp.usage.total_tokens) instead of raw dict-like results.
  • Errors — typed DashScopeError exceptions raised by default.
  • Streaming — ChatStream with .text_stream, .reasoning_stream and .get_final_message().
  • Protocols — one client speaks native, openai or anthropic via protocol=.
  • Async — dashscope.AsyncClient() mirrors the sync surface.
  • Connections — one keep-alive pool per client, shared across resources.
  • Configuration — layered: arguments → dashscope.settings → environment → ~/.dashscope/config.toml.

1.x line

The 1.x API is preserved for backward compatibility under dashscope.legacy_1x and remains importable from the dashscope root. It is frozen: new capabilities land only in the 2.x surface. See Legacy 1.x.