DashScope Python SDK v2.1.0: Qwen Live Realtime Audio, devkit, and Automated Releases

[luk384090-cloud]

DashScope Python SDK v2.1.0: Qwen Live Realtime Audio, devkit, and Automated Releases

v2.1.0 of the DashScope Python SDK is out. This release brings realtime audio sessions, a new offline debugging toolkit, and a fully automated release pipeline — plus a broad pass on transport, parsing, and pagination across all resources.

🚀 Highlights

• Qwen Live realtime audio — realtime audio sessions built on a shared realtime transport, now with support for audio sources in voice enrollment and graceful Live/Realtime shutdown.
• New dashscope.devkit — offline self-diagnosis for failed calls, plus static / record / replay HTTP mocking at the transport layer. Reproduce and debug API issues without touching the network.
• Automated release pipeline — pushing a v* tag now builds, publishes to PyPI, and drafts the GitHub Release automatically (ci(release): add tag-triggered PyPI publishing workflow #234).

✨ Features

• Audio — Qwen Live sessions on the shared realtime transport (feat: add qwen live api #231); audio sources accepted for voice enrollment.
• Devkit — offline self-diagnosis for failed calls; static/record/replay dev mock at the HTTP transport.
• Transport & core — introduced APIResponse and a raw parameter for raw responses; per-request options across all resources; client request ID and response validation exception; fault-tolerant response model parsing with strict validation config.
• Pagination — new generic pagination base class and refactored resource list interfaces.
• AgentStudio — session mount fields for Bailian memory and OSS.
• v2 — typed parsing, per-request options, and pagination fixes extended to all resources; broader image model and input support, including wan2.7 image responses.

♻️ Refactoring

• CLI: migrated the agent runtime to dashscope-agenticx (refactor(cli): migrate agent runtime to dashscope-agenticx #230).
• Server: extracted client-disconnect response logic.

🐛 Bug fixes

• TTS — prevent KeyError for non-timestamp messages in speech parts.
• Reinforcement — preserve FaaS business errors and request context; allow local wheel paths in pip requirement tokens; address agentic RL review findings.
• Core — unify API key resolution and improve error handling.
• Logging — correlate SDK 2.0 response errors with request IDs.

⚠️ Breaking changes

None.

📦 Upgrade

pip install --upgrade dashscope==2.1.0

👥 Contributors

Thanks to everyone who made this release possible: @luk384090-cloud, @songguocola, @foleydang, @lzsweb

📝 Full changelog

See the complete diff: v2.0.4...v2.1.0

Give the new realtime audio APIs and dashscope.devkit a try — feedback and bug reports are very welcome here, and they directly shape the next iteration.