Skip to content

feat(rerank): DashScope 提供商支持 qwen3-rerank - #494

Open
flyto168 wants to merge 1 commit into
EverMind-AI:mainfrom
flyto168:feature/rerank-qwen3-dashscope
Open

flyto168 wants to merge 1 commit into
EverMind-AI:mainfrom
flyto168:feature/rerank-qwen3-dashscope

Conversation

@flyto168

@flyto168 flyto168 commented Oct 13, 2026 •

Copy link
Copy Markdown
Contributor

问题

DashScope 提供商此前不支持 qwen3-rerank,使用 DashScope 的部署无法启用该模型。

改法

  • rerank 的 DashScope 提供商新增 qwen3-rerank 支持(模型接入与参数适配)。

验证

  • rerank 定向测试 87 passed;Ruff 通过。
  • 基于上游 main @ a345d08,仅 2 文件。

范围

仅含通用修复与测试。

The DashScope international region has no gte-rerank-v2, so the provider
could not be used there. qwen3-rerank takes the same native endpoint and
body; it also accepts the task instruction as parameters.instruct, which
is now forwarded (gte-rerank-v2 still sends none).

Measured on 55 agent-skill queries vs the DeepInfra Qwen3-Reranker-4B in
use: @1 38 vs 38, @3 50 vs 47, MRR 0.807 vs 0.791; latency p50 0.16s vs
0.52s, p90 0.18s vs 1.24s from ap-southeast-1.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@flyto168 flyto168 changed the title feat(rerank): support qwen3-rerank on the DashScope provider feat(rerank): DashScope 提供商支持 qwen3-rerank Oct 13, 2026

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant