Update api/utils.py

api/utils.py

@@ -1,6 +1,8 @@
+# api/utils.py
+
 from datetime import datetime
 import json
-from typing import Any, Dict, Optional, List
+from typing import Any, Dict, Optional
 
 import httpx
 from api.config import (
@@ -13,17 +15,66 @@ from api.config import (
     MODEL_REFERERS
 )
 from fastapi import HTTPException
-from api.models import ChatRequest, Message
+from api.models import ChatRequest
 
 from api.logger import setup_logger
 
-import uuid  # Ensure this is imported if needed elsewhere
+import uuid
 
 logger = setup_logger(__name__)
 
+def create_chat_completion_data(
+    content: str, model: str, timestamp: int, finish_reason: Optional[str] = None
+) -> Dict[str, Any]:
+    return {
+        "id": f"chatcmpl-{uuid.uuid4()}",
+        "object": "chat.completion.chunk",
+        "created": timestamp,
+        "model": model,
+        "choices": [
+            {
+                "index": 0,
+                "delta": {"content": content, "role": "assistant"},
+                "finish_reason": finish_reason,
+            }
+        ],
+        "usage": None,
+    }
+
+def message_to_dict(message, model_prefix: Optional[str] = None):
+    if isinstance(message.content, str):
+        content = message.content
+        if model_prefix:
+            content = f"{model_prefix} {content}"
+        return {"role": message.role, "content": content}
+    elif isinstance(message.content, list) and len(message.content) == 2:
+        content = message.content[0]["text"]
+        if model_prefix:
+            content = f"{model_prefix} {content}"
+        return {
+            "role": message.role,
+            "content": content,
+            "data": {
+                "imageBase64": message.content[1]["image_url"]["url"],
+                "fileText": "",
+                "title": "snapshot",
+            },
+        }
+    else:
+        return {"role": message.role, "content": message.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()
+    logger.debug("No prefix to strip from content.")
+    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, "")
     referer_path = MODEL_REFERERS.get(request.model, f"/?model={request.model}")
     referer_url = f"{BASE_URL}{referer_path}"
 
@@ -31,11 +82,8 @@ async def process_streaming_response(request: ChatRequest):
     dynamic_headers = headers.copy()
     dynamic_headers['Referer'] = referer_url
 
-    # Convert Message objects to dictionaries
-    messages = [msg.dict() for msg in request.messages]
-
     json_data = {
-        "messages": messages,  # Pass serialized messages
+        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages],
         "previewToken": None,
         "userId": None,
         "codeModelMode": True,
@@ -72,10 +120,12 @@ async def process_streaming_response(request: ChatRequest):
                         content = line
                         if content.startswith("$@$v=undefined-rv1$@$"):
                             content = content[21:]
-                        # Since we've removed custom processing, directly yield the content
-                        yield f"data: {content}\n\n"
+                        # Strip the model prefix from the response 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: [DONE]\n\n"
+                yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, 'stop'))}\n\n"
+                yield "data: [DONE]\n\n"
         except httpx.HTTPStatusError as e:
             logger.error(f"HTTP error occurred: {e}")
             raise HTTPException(status_code=e.response.status_code, detail=str(e))
@@ -83,10 +133,10 @@ async def process_streaming_response(request: ChatRequest):
             logger.error(f"Error occurred during request: {e}")
             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, "")
     referer_path = MODEL_REFERERS.get(request.model, f"/?model={request.model}")
     referer_url = f"{BASE_URL}{referer_path}"
 
@@ -94,11 +144,8 @@ async def process_non_streaming_response(request: ChatRequest):
     dynamic_headers = headers.copy()
     dynamic_headers['Referer'] = referer_url
 
-    # Convert Message objects to dictionaries
-    messages = [msg.dict() for msg in request.messages]
-
     json_data = {
-        "messages": messages,  # Pass serialized messages
+        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages],
         "previewToken": None,
         "userId": None,
         "codeModelMode": True,
@@ -121,25 +168,24 @@ async def process_non_streaming_response(request: ChatRequest):
     full_response = ""
     async with httpx.AsyncClient() as client:
         try:
-            response = await client.post(
-                f"{BASE_URL}/api/chat",
-                headers=dynamic_headers,
-                json=json_data,
-                timeout=100,
-            )
-            response.raise_for_status()
-            full_response = response.text  # Get the response text
+            async with client.stream(
+                method="POST", url=f"{BASE_URL}/api/chat", headers=dynamic_headers, json=json_data
+            ) as response:
+                response.raise_for_status()
+                async for chunk in response.aiter_text():
+                    full_response += chunk
         except httpx.HTTPStatusError as e:
             logger.error(f"HTTP error occurred: {e}")
             raise HTTPException(status_code=e.response.status_code, detail=str(e))
         except httpx.RequestError as e:
             logger.error(f"Error occurred during request: {e}")
             raise HTTPException(status_code=500, detail=str(e))
-    
     if full_response.startswith("$@$v=undefined-rv1$@$"):
         full_response = full_response[21:]
 
-    # Return the response as-is without custom processing
+    # Strip the model prefix from the full response
+    cleaned_full_response = strip_model_prefix(full_response, model_prefix)
+
     return {
         "id": f"chatcmpl-{uuid.uuid4()}",
         "object": "chat.completion",
@@ -148,7 +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",
             }
         ],