Update api/utils.py

api/utils.py

@@ -8,7 +8,15 @@ from fastapi import Depends, HTTPException
 from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 
 from api import validate
-from api.config import APP_SECRET, BASE_URL, MODEL_MAPPING, headers
+from api.config import (
+    APP_SECRET,
+    BASE_URL,
+    MODEL_MAPPING,
+    AGENT_MODE,
+    TRENDING_AGENT_MODE,
+    MODEL_PREFIXES,
+    headers,
+)
 from api.models import ChatRequest
 from api.logger import setup_logger
 
@@ -38,30 +46,46 @@ def verify_app_secret(credentials: HTTPAuthorizationCredentials = Depends(securi
         raise HTTPException(status_code=403, detail="Invalid APP_SECRET")
     return credentials.credentials
 
-def message_to_dict(message):
-    if isinstance(message.content, str):
-        return {"role": message.role, "content": message.content}
-    elif isinstance(message.content, list) and len(message.content) == 2:
+def message_to_dict(message, model_prefix: Optional[str] = None):
+    content = message.content
+    if isinstance(content, list) and len(content) == 2 and "image_url" in content[1]:
+        # Handle messages with image data
+        content_text = content[0]["text"]
+        if model_prefix:
+            content_text = f"{model_prefix} {content_text}"
         return {
             "role": message.role,
-            "content": message.content[0]["text"],
+            "content": content_text,
             "data": {
-                "imageBase64": message.content[1]["image_url"]["url"],
+                "imageBase64": content[1]["image_url"]["url"],
                 "fileText": "",
                 "title": "snapshot",
             },
         }
     else:
-        return {"role": message.role, "content": message.content}
+        if model_prefix:
+            content = f"{model_prefix} {content}"
+        return {"role": message.role, "content": content}
+
+def strip_model_prefix(content: str, model_prefix: Optional[str] = None) -> str:
+    """Remove the model prefix from the response content if present."""
+    if model_prefix and content.startswith(model_prefix):
+        logger.debug(f"Stripping prefix '{model_prefix}' from content.")
+        return content[len(model_prefix):].strip()
+    return content
 
 async def process_streaming_response(request: ChatRequest):
+    agent_mode = AGENT_MODE.get(request.model, {})
+    trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {})
+    model_prefix = MODEL_PREFIXES.get(request.model, "")
+
     json_data = {
-        "messages": [message_to_dict(msg) for msg in request.messages],
+        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages],
         "previewToken": None,
         "userId": None,
         "codeModelMode": True,
-        "agentMode": {},
-        "trendingAgentMode": {},
+        "agentMode": agent_mode,
+        "trendingAgentMode": trending_agent_mode,
         "isMicMode": False,
         "userSystemPrompt": None,
         "maxTokens": request.max_tokens,
@@ -98,9 +122,10 @@ async def process_streaming_response(request: ChatRequest):
                             yield f"data: {json.dumps(create_chat_completion_data(content, request.model, timestamp))}\n\n"
                             break
                         if content.startswith("$@$v=undefined-rv1$@$"):
-                            yield f"data: {json.dumps(create_chat_completion_data(content[21:], request.model, timestamp))}\n\n"
-                        else:
-                            yield f"data: {json.dumps(create_chat_completion_data(content, request.model, timestamp))}\n\n"
+                            content = content[21:]
+                        # Strip model prefix from content
+                        cleaned_content = strip_model_prefix(content, model_prefix)
+                        yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp))}\n\n"
 
                 yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, 'stop'))}\n\n"
                 yield "data: [DONE]\n\n"
@@ -112,13 +137,17 @@ async def process_streaming_response(request: ChatRequest):
             raise HTTPException(status_code=500, detail=str(e))
 
 async def process_non_streaming_response(request: ChatRequest):
+    agent_mode = AGENT_MODE.get(request.model, {})
+    trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {})
+    model_prefix = MODEL_PREFIXES.get(request.model, "")
+
     json_data = {
-        "messages": [message_to_dict(msg) for msg in request.messages],
+        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages],
         "previewToken": None,
         "userId": None,
         "codeModelMode": True,
-        "agentMode": {},
-        "trendingAgentMode": {},
+        "agentMode": agent_mode,
+        "trendingAgentMode": trending_agent_mode,
         "isMicMode": False,
         "userSystemPrompt": None,
         "maxTokens": request.max_tokens,
@@ -155,6 +184,8 @@ async def process_non_streaming_response(request: ChatRequest):
         full_response = "hid已刷新，重新对话即可"
     if full_response.startswith("$@$v=undefined-rv1$@$"):
         full_response = full_response[21:]
+    # Strip model prefix from full_response
+    cleaned_full_response = strip_model_prefix(full_response, model_prefix)
     return {
         "id": f"chatcmpl-{uuid.uuid4()}",
         "object": "chat.completion",
@@ -163,7 +194,7 @@ async def process_non_streaming_response(request: ChatRequest):
         "choices": [
             {
                 "index": 0,
-                "message": {"role": "assistant", "content": full_response},
+                "message": {"role": "assistant", "content": cleaned_full_response},
                 "finish_reason": "stop",
             }
         ],