refine xinference (#2521)

### What problem does this PR solve?

#1588

### Type of change

- [x] Refactoring

rag/llm/embedding_model.py

@@ -268,6 +268,8 @@ class FastEmbed(Base):
 
 class XinferenceEmbed(Base):
     def __init__(self, key, model_name="", base_url=""):
+        if base_url.split("/")[-1] != "v1":
+            base_url = os.path.join(base_url, "v1")
         self.client = OpenAI(api_key="xxx", base_url=base_url)
         self.model_name = model_name