How can we use the Qwen3 and embedder of Alibaba?

[xiatiandegaga]

version:0.24.1

[cyyeh]

version:0.24.1

Maybe it's not supported yet

https://github.com/BerriAI/litellm/issues/9198

[xiatiandegaga]

version:0.24.1

Maybe it's not supported yet

BerriAI/litellm#9198

well,thanks

[cyyeh]

@xiatiandegaga could u still upload your config.yaml, I would like to take a look

[xiatiandegaga]

config.txt

@xiatiandegaga could u still upload your config.yaml, I would like to take a look

[cyyeh]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

[xiatiandegaga]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

can't work when I add openai/ for embedding ，I‘ll wait for the version of UI just like supersonic lol

[JinCheng666]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

can't work when I add openai/ for embedding ，I‘ll wait for the version of UI just like supersonic lol

@xiatiandegaga May I ask if you have tried using this configuration file? https://github.com/Canner/WrenAI/blob/main/wren-ai-service/docs/config_examples/config.qwen3.yaml

[liusaiweng]

Qwen3 is the most popular mode in china with a large amount of customer,we need to use it directly.

[we1s3]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

can't work when I add openai/ for embedding ，I‘ll wait for the version of UI just like supersonic lol

@xiatiandegaga May I ask if you have tried using this configuration file? https://github.com/Canner/WrenAI/blob/main/wren-ai-service/docs/config_examples/config.qwen3.yaml

config.qwen3.yaml 的中embedder 配置使用的是openai向量模型，并且qwen3在wrenai中是通过openrouter代理的方式使用，在配置文件的示例中并没有直接使用dashscope，所以wrenai在目前版本的配置文件稀里糊涂，我们甚至不知道到底怎么使用。。。然而wrenai文档中又明确说明支持dashscope方式的所有qwen大模型。。。实在难以理解

[cyyeh]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

can't work when I add openai/ for embedding ，I‘ll wait for the version of UI just like supersonic lol

@xiatiandegaga May I ask if you have tried using this configuration file? https://github.com/Canner/WrenAI/blob/main/wren-ai-service/docs/config_examples/config.qwen3.yaml

config.qwen3.yaml 的中embedder 配置使用的是openai向量模型，并且qwen3在wrenai中是通过openrouter代理的方式使用，在配置文件的示例中并没有直接使用dashscope，所以wrenai在目前版本的配置文件稀里糊涂，我们甚至不知道到底怎么使用。。。然而wrenai文档中又明确说明支持dashscope方式的所有qwen大模型。。。实在难以理解

For dashscope models, you could try following the docs here: https://docs.litellm.ai/docs/providers/dashscope

Basically we are simply using litellm underneath

[we1s3]

@xiatiandegaga since you declared qwen3-fast as alias in your llm config, you should put litellm_llm.qwen3-fast in pipe definitions instead of using litellm_llm.default. Also I think maybe you also need to put openai/ prefix for embedding model?

can't work when I add openai/ for embedding ，I‘ll wait for the version of UI just like supersonic lol

@xiatiandegaga May I ask if you have tried using this configuration file? https://github.com/Canner/WrenAI/blob/main/wren-ai-service/docs/config_examples/config.qwen3.yaml

config.qwen3.yaml 的中embedder 配置使用的是openai向量模型，并且qwen3在wrenai中是通过openrouter代理的方式使用，在配置文件的示例中并没有直接使用dashscope，所以wrenai在目前版本的配置文件稀里糊涂，我们甚至不知道到底怎么使用。。。然而wrenai文档中又明确说明支持dashscope方式的所有qwen大模型。。。实在难以理解

For dashscope models, you could try following the docs here: https://docs.litellm.ai/docs/providers/dashscope

Basically we are simply using litellm underneath

如果将wrenai默认的向量模型替换为qwen text-embedding-v3，在提问时候就会发生异常，因为qwen text-embedding-v3向量模型只支持1024 错误明细： PRQBT:job 2 have changes, returning question count: 0, updating

sendEvent home_generate_project_recommendation_questions_failed {

projectId: 2,

projectType: 'MSSQL',

status: 'FAILED',

questions: [],

error: {

code: 'OTHERS',

message: 'An error occurred during question recommendation generation: litellm.BadRequestError: DashscopeException - <400> InternalError.Algo.InvalidParameter: Range of input length should be [1, 30720]',

shortMessage: 'Internal server error'

}

} AI false

[heiniuhaha]

成功集成WrenAI与阿里云DashScope服务，使用最新的AI模型：

LLM模型: qwen-plus - 性能强大的大语言模型 Embedding模型: text-embedding-v4 - 最新的向量嵌入模型 详情见 https://github.com/lovrabet-ai/wrenai-qwen/blob/main/wrenai-dashscope-integration-guide.md

[JinCheng666]

成功集成WrenAI与阿里云DashScope服务，使用最新的AI模型：

LLM模型: qwen-plus - 性能强大的大语言模型 Embedding模型: text-embedding-v4 - 最新的向量嵌入模型 详情见 https://github.com/lovrabet-ai/wrenai-qwen/blob/main/wrenai-dashscope-integration-guide.md

thx!

[xiatiandegaga]

成功集成WrenAI与阿里云DashScope服务，使用最新的AI模型：

LLM模型: qwen-plus - 性能强大的大语言模型 Embedding模型: text-embedding-v4 - 最新的向量嵌入模型 详情见 https://github.com/lovrabet-ai/wrenai-qwen/blob/main/wrenai-dashscope-integration-guide.md

thk