added thinking support

main.py

@@ -65,7 +65,7 @@ MODEL_MAPPING = {
 BASE_URL = "https://pa002.abacus.ai"
 
 TIMEOUT = 30.0  # 请求超时时间（秒）
-MAX_RETRIES = 3  # 最大重试次数
+MAX_RETRIES = 1  # 最大重试次数
 RETRY_DELAY = 1  # 重试延迟（秒）
 
 @app.get("/hf/v1/models")
@@ -161,37 +161,48 @@ async def chat_completions(request: Request, chat_request: ChatRequest):
     # 创建会话ID
     # conversation_id = str(uuid.uuid4())[:-9]
     # conversation_id = "d166251ca"
-    conversation_id = "7f0f8d76d"
+    # conversation_id = "7f0f8d76d"
     # external_id = str(uuid.uuid4())[:-9]
     external_id = "f6ca7aa14"
     
     # 处理消息
     full_message = process_messages(chat_request.messages)
     
-    # 准备请求数据
-    request_data = {
-        "requestId": str(uuid.uuid4()),
-        "deploymentConversationId": conversation_id,
-        "message": full_message,
-        "isDesktop": True,
-        "chatConfig": {
-            "timezone": "America/Los_Angeles",
-            "language": "en-US"
-        },
-        "llmName": MODEL_MAPPING.get(chat_request.model, chat_request.model),
-        "externalApplicationId": external_id
-        # "externalApplicationId": "c9a23c6d8"
-    }
+    
+    print("start!!!!!")
+
+    new_chat_data = {"deploymentId":"47845a764","name":"New Chat","externalApplicationId":"f6ca7aa14"}
 
     # 流式请求处理
     async def generate_stream():
-        headers = get_headers(auth_token)
-        # headers["content-length"] = str(len(str(request_data)))
-        headers["referer"] = "https://apps.abacus.ai/chatllm/?appId=" + external_id + "&convoId=" + conversation_id
-        print(headers["referer"])
+        headers_new_request = get_headers(auth_token)
+        headers_new_request["referer"] = "https://apps.abacus.ai/chatllm/?appId=" + external_id
+
         
+        # headers["content-length"] = str(len(str(request_data)))
         for retry in range(MAX_RETRIES):
             try:
+                response = httpx.post("https://apps.abacus.ai/api/createDeploymentConversation",
+                    data=json.dumps(new_chat_data),
+                    headers=headers_new_request,
+                )
+                conversation_id = response.json()["result"]["deploymentConversationId"]
+                # 准备请求数据
+                request_data = {
+                    "requestId": str(uuid.uuid4()),
+                    "deploymentConversationId": conversation_id,
+                    "message": full_message,
+                    "isDesktop": True,
+                    "chatConfig": {
+                        "timezone": "America/Los_Angeles",
+                        "language": "en-US"
+                    },
+                    "llmName": MODEL_MAPPING.get(chat_request.model, chat_request.model),
+                    "externalApplicationId": external_id
+                    # "externalApplicationId": "c9a23c6d8"
+                }
+                headers = get_headers(auth_token)
+                headers["referer"] = "https://apps.abacus.ai/chatllm/?appId=" + external_id + "&convoId=" + conversation_id
                 async with httpx.AsyncClient() as client:
                     async with client.stream(
                         "POST",
@@ -200,14 +211,17 @@ async def chat_completions(request: Request, chat_request: ChatRequest):
                         content=json.dumps(request_data),
                         timeout=TIMEOUT
                     ) as response:
+                        thinking_started = False
                         async for line in response.aiter_lines():
                             if not line.strip():
                                 continue
                                 
                             try:
                                 data = json.loads(line)
-                                
-                                if data.get("type") == "text" and data.get("title") != "Thinking...":
+                                # print(line)
+                                if data.get("type") == "collapsible_component":
+                                    thinking_started = True
+                                    think_id = data.get("messageId", "")
                                     chunk = {
                                         "id": str(uuid.uuid4()),
                                         "object": "chat.completion.chunk",
@@ -216,12 +230,44 @@ async def chat_completions(request: Request, chat_request: ChatRequest):
                                         "choices": [{
                                             "delta": {
                                                 "role": "assistant",
-                                                "content": data.get("segment", "")
+                                                "content": "<think>"
                                             },
                                             "index": 0
                                         }]
                                     }
                                     yield f"data: {json.dumps(chunk)}\n\n"
+
+                                if data.get("type") == "text" and data.get("title") != "Thinking...":
+                                    if thinking_started == True and data.get("messageId", "") != think_id:
+                                        thinking_started = False
+                                        chunk = {
+                                            "id": str(uuid.uuid4()),
+                                            "object": "chat.completion.chunk",
+                                            "created": int(uuid.uuid1().time_low),
+                                            "model": chat_request.model,
+                                            "choices": [{
+                                                "delta": {
+                                                    "role": "assistant",
+                                                    "content": "</think>" + data.get("segment", "")
+                                                },
+                                                "index": 0
+                                            }]
+                                        }
+                                    else:
+                                        chunk = {
+                                            "id": str(uuid.uuid4()),
+                                            "object": "chat.completion.chunk",
+                                            "created": int(uuid.uuid1().time_low),
+                                            "model": chat_request.model,
+                                            "choices": [{
+                                                "delta": {
+                                                    "role": "assistant",
+                                                    "content": data.get("segment", "")
+                                                },
+                                                "index": 0
+                                            }]
+                                        }
+                                    yield f"data: {json.dumps(chunk)}\n\n"
                                     
                                 if data.get("end"):
                                     # 发送结束标记