litellm - 💡(How to fix) Fix [Feature]: Support DashScope embedding (text-embedding-v4) and rerank (qwen3-rerank) [1 comments, 2 participants]

Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

The Feature

Support DashScope embedding (text-embedding-v4) and rerank (qwen3-rerank)

Currently, LiteLLM does not support DashScope embedding and rerank APIs properly. Even when using the OpenAI-compatible endpoint, there are compatibility issues (response parsing, encoding_format, etc.).

It would be helpful to have native support for: 1. DashScope embedding: • text-embedding-v4 2. DashScope rerank: • qwen3-rerank

Motivation, pitch

I’m using LiteLLM as a unified interface for multiple providers (OpenAI, DeepSeek, DashScope, etc.) in a RAG pipeline.

Current issues

1. Embedding (text-embedding-v4) Using DashScope via OpenAI-compatible endpoint works partially, but: • encoding_format must be manually set to "float" • response is sometimes returned as dict instead of OpenAI-style object • requires custom parsing instead of standard response.data[0].embedding

This breaks compatibility with frameworks expecting OpenAI-like responses.

1. Rerank (qwen3-rerank)

DashScope rerank is not supported at all:

ValueError: Unsupported provider: dashscope

From:

litellm.rerank(...)

So currently there’s no way to use DashScope rerank through LiteLLM.

Expected behavior

1. Native provider support for dashscope:

• embedding
• erank

1. Proper model mapping, e.g.:

• model="dashscope/text-embedding-v4"
• model="dashscope/qwen3-rerank"

1. Standardized response format (OpenAI-compatible object), so downstream code can safely use:

response.data[0].embedding

1. No need for manual fixes like:

• forcing encoding_format
• custom response parsing

Why this matters

DashScope is widely used (especially in China), and having first-class support in LiteLLM would make it much easier to:

• build RAG systems
• switch between providers
• void writing custom adapters

What part of LiteLLM is this about?

SDK (litellm Python package)

LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

No

Twitter / LinkedIn details

No response