fix: add support for multimodal embeddings in AliyunBaiLianEmbedding class

Author: wxg0103

apps/models_provider/impl/aliyun_bai_lian_model_provider/model/embedding.py

@@ -6,6 +6,7 @@
     @date：2024/10/16 16:34
     @desc:
 """
+from http import HTTPStatus
 from typing import Dict, List
 
 from openai import OpenAI
@@ -16,11 +17,15 @@
 class AliyunBaiLianEmbedding(MaxKBBaseModel):
     model_name: str
     optional_params: dict
+    api_base: str
+    api_key: str
 
     def __init__(self, api_key, model_name: str, api_base: str, optional_params: dict):
         self.client = OpenAI(api_key=api_key, base_url=api_base).embeddings
         self.model_name = model_name
         self.optional_params = optional_params
+        self.api_key = api_key
+        self.api_base = api_base
 
     def is_cache_model(self):
         return False
@@ -42,6 +47,24 @@ def embed_query(self, text: str):
     def embed_documents(
             self, texts: List[str], chunk_size: int | None = None
     ) -> List[List[float]]:
+        # 处理多模态的向量化
+        if 'vl-embedding' in self.model_name or 'embedding-vision' in self.model_name or 'multimodal' in self.model_name:
+            import dashscope
+            dashscope.api_key = self.api_key
+            dashscope.base_http_api_url = self.api_base
+            multimodal_input = [{"text": text} for text in texts]
+            resp = dashscope.MultiModalEmbedding.call(
+                model="tongyi-embedding-vision-plus",
+                input=multimodal_input,  # type: ignore
+                **self.optional_params
+            )
+
+            if resp.status_code == HTTPStatus.OK:
+                embeddings_data = resp.output.get('embeddings', [])
+                return [item.get('embedding', []) for item in embeddings_data]
+            else:
+                raise Exception(f'MultiModalEmbedding call failed: status={resp.status_code}, message={resp.message}')
+
         if len(self.optional_params) > 0:
             res = self.client.create(
                 input=texts, model=self.model_name, encoding_format="float",