add support for PerfXCloud (#1883)

### What problem does this PR solve?

#1853  add support for PerfXCloud

### Type of change


- [x] New Feature (non-breaking change which adds functionality)

Co-authored-by: Zhedong Cen <cenzhedong2@126.com>

rag/llm/__init__.py

@@ -38,7 +38,8 @@ EmbeddingModel = {
     "NVIDIA": NvidiaEmbed,
     "LM-Studio": LmStudioEmbed,
     "OpenAI-API-Compatible": OpenAI_APIEmbed,
-    "cohere": CoHereEmbed
+    "cohere": CoHereEmbed,
+    "PerfXCloud": PerfXCloudEmbed,
 }
 
 
@@ -84,7 +85,8 @@ ChatModel = {
     "LM-Studio": LmStudioChat,
     "OpenAI-API-Compatible": OpenAI_APIChat,
     "cohere": CoHereChat,
-    "LeptonAI": LeptonAIChat
+    "LeptonAI": LeptonAIChat,
+    "PerfXCloud": PerfXCloudChat
 }
 
 

rag/llm/chat_model.py

@@ -987,4 +987,11 @@ class LeptonAIChat(Base):
     def __init__(self, key, model_name, base_url=None):
         if not base_url:
             base_url = os.path.join("https://"+model_name+".lepton.run","api","v1")
-        super().__init__(key, model_name, base_url)
+        super().__init__(key, model_name, base_url)
+
+
+class PerfXCloudChat(Base):
+    def __init__(self, key, model_name, base_url="https://cloud.perfxlab.cn/v1"):
+        if not base_url:
+            base_url = "https://cloud.perfxlab.cn/v1"
+        super().__init__(key, model_name, base_url)

rag/llm/embedding_model.py

@@ -553,3 +553,10 @@ class CoHereEmbed(Base):
         return np.array([d for d in res.embeddings.float]), int(
             res.meta.billed_units.input_tokens
         )
+
+
+class PerfXCloudEmbed(OpenAIEmbed):
+    def __init__(self, key, model_name, base_url="https://cloud.perfxlab.cn/v1"):
+        if not base_url:
+            base_url = "https://cloud.perfxlab.cn/v1"
+        super().__init__(key, model_name, base_url)