Update app.py

app.py

@@ -23,16 +23,11 @@ logging.basicConfig(level=logging.INFO,
 API_ENDPOINT = "https://api.deepseek.com/user/balance"
 TEST_MODEL_ENDPOINT = "https://api.deepseek.com/v1/chat/completions"
 MODELS_ENDPOINT = "https://api.deepseek.com/v1/models"
-EMBEDDINGS_ENDPOINT = "https://api.deepseek.com/v1/embeddings"
 
 app = Flask(__name__)
 
 text_models = []
 free_text_models = []
-embedding_models = []
-free_embedding_models = []
-image_models = []
-free_image_models = []
 
 invalid_keys_global = []
 free_keys_global = []
@@ -78,6 +73,7 @@ def get_credit_summary(api_key):
             exchange_rate = get_usd_to_cny_rate()
             if exchange_rate is not None:
                 total_balance_cny += usd_balance * exchange_rate
+                 logging.info(f"获取美元兑人民币汇率成功{total_balance_cny}")
             else:
                  logging.warning(f"获取美元兑人民币汇率失败，无法转换美元余额，API Key：{api_key}")
                  total_balance_cny += usd_balance * 7.2
@@ -731,8 +727,8 @@ def billing_subscription():
         "business_address": None
     })
 
-@app.route('/handsome/v1/embeddings', methods=['POST'])
-def handsome_embeddings():
+@app.route('/handsome/v1/chat/completions', methods=['POST'])
+def handsome_chat_completions():
     if not check_authorization(request):
         return jsonify({"error": "Unauthorized"}), 401
 
@@ -741,11 +737,13 @@ def handsome_embeddings():
         return jsonify({"error": "Invalid request data"}), 400
 
     model_name = data['model']
+    
     request_type = determine_request_type(
         model_name,
-        embedding_models,
-        free_embedding_models
+        text_models + image_models,
+        free_text_models + free_image_models
     )
+    
     api_key = select_key(request_type, model_name)
 
     if not api_key:
@@ -763,706 +761,86 @@ def handsome_embeddings():
         "Authorization": f"Bearer {api_key}",
         "Content-Type": "application/json"
     }
-
+    
     try:
         start_time = time.time()
         response = requests.post(
-            EMBEDDINGS_ENDPOINT,
+            TEST_MODEL_ENDPOINT,
             headers=headers,
             json=data,
-            timeout=120
+            stream=data.get("stream", False),
+            timeout=60
         )
 
         if response.status_code == 429:
             return jsonify(response.json()), 429
 
-        response.raise_for_status()
-        end_time = time.time()
-        response_json = response.json()
-        total_time = end_time - start_time
-
-        try:
-            prompt_tokens = response_json["usage"]["prompt_tokens"]
-            embedding_data = response_json["data"]
-        except (KeyError, ValueError, IndexError) as e:
-            logging.error(
-                f"解析响应 JSON 失败: {e}, "
-                f"完整内容: {response_json}"
-            )
-            prompt_tokens = 0
-            embedding_data = []
-
-        logging.info(
-            f"使用的key: {api_key}, "
-            f"提示token: {prompt_tokens}, "
-            f"总共用时: {total_time:.4f}秒, "
-            f"使用的模型: {model_name}"
-        )
-
-        with data_lock:
-            request_timestamps.append(time.time())
-            token_counts.append(prompt_tokens)
-        
-        return jsonify({
-            "object": "list",
-            "data": embedding_data,
-            "model": model_name,
-            "usage": {
-                "prompt_tokens": prompt_tokens,
-                "total_tokens": prompt_tokens
-            }
-        })
-
-    except requests.exceptions.RequestException as e:
-        return jsonify({"error": str(e)}), 500
-
-@app.route('/handsome/v1/images/generations', methods=['POST'])
-def handsome_images_generations():
-    if not check_authorization(request):
-        return jsonify({"error": "Unauthorized"}), 401
-
-    data = request.get_json()
-    if not data or 'model' not in data:
-        return jsonify({"error": "Invalid request data"}), 400
-
-    model_name = data.get('model')
-    
-    request_type = determine_request_type(
-        model_name,
-        image_models,
-        free_image_models
-    )
-    
-    api_key = select_key(request_type, model_name)
-
-    if not api_key:
-         return jsonify(
-             {
-                 "error": (
-                     "No available API key for this "
-                     "request type or all keys have "
-                     "reached their limits"
-                 )
-             }
-         ), 429
-    
-    headers = {
-        "Authorization": f"Bearer {api_key}",
-        "Content-Type": "application/json"
-    }
-    
-    response_data = {}
-    
-    if "stable-diffusion" in model_name or model_name in ["black-forest-labs/FLUX.1-schnell", "Pro/black-forest-labs/FLUX.1-schnell","black-forest-labs/FLUX.1-dev", "black-forest-labs/FLUX.1-pro"]:
-        siliconflow_data = {
-            "model": model_name,
-            "prompt": data.get("prompt"),
-            
-        }
-        
-        if model_name == "black-forest-labs/FLUX.1-pro":
-            siliconflow_data["width"] = data.get("width", 1024)
-            siliconflow_data["height"] = data.get("height", 768)
-            siliconflow_data["prompt_upsampling"] = data.get("prompt_upsampling", False)
-            siliconflow_data["image_prompt"] = data.get("image_prompt")
-            siliconflow_data["steps"] = data.get("steps", 20)
-            siliconflow_data["guidance"] = data.get("guidance", 3)
-            siliconflow_data["safety_tolerance"] = data.get("safety_tolerance", 2)
-            siliconflow_data["interval"] = data.get("interval", 2)
-            siliconflow_data["output_format"] = data.get("output_format", "png")
-            seed = data.get("seed")
-            if isinstance(seed, int) and 0 < seed < 9999999999:
-                siliconflow_data["seed"] = seed
-
-            if siliconflow_data["width"] < 256 or siliconflow_data["width"] > 1440 or siliconflow_data["width"] % 32 != 0:
-                siliconflow_data["width"] = 1024
-            if siliconflow_data["height"] < 256 or siliconflow_data["height"] > 1440 or siliconflow_data["height"] % 32 != 0:
-                 siliconflow_data["height"] = 768
-
-            if siliconflow_data["steps"] < 1 or siliconflow_data["steps"] > 50:
-                siliconflow_data["steps"] = 20
-            if siliconflow_data["guidance"] < 1.5 or siliconflow_data["guidance"] > 5:
-                 siliconflow_data["guidance"] = 3
-            if siliconflow_data["safety_tolerance"] < 0 or siliconflow_data["safety_tolerance"] > 6:
-                siliconflow_data["safety_tolerance"] = 2
-            if siliconflow_data["interval"] < 1 or siliconflow_data["interval"] > 4 :
-                 siliconflow_data["interval"] = 2
-        else:    
-            siliconflow_data["image_size"] = data.get("image_size", "1024x1024") 
-            siliconflow_data["prompt_enhancement"] = data.get("prompt_enhancement", False)
-            seed = data.get("seed")
-            if isinstance(seed, int) and 0 < seed < 9999999999:
-                siliconflow_data["seed"] = seed
-        
-            if model_name not in ["black-forest-labs/FLUX.1-schnell", "Pro/black-forest-labs/FLUX.1-schnell"]:
-                siliconflow_data["batch_size"] = data.get("n", 1)
-                siliconflow_data["num_inference_steps"] = data.get("steps", 20)
-                siliconflow_data["guidance_scale"] = data.get("guidance_scale", 7.5)
-                siliconflow_data["negative_prompt"] = data.get("negative_prompt")
-                if siliconflow_data["batch_size"] < 1:
-                    siliconflow_data["batch_size"] = 1
-                if siliconflow_data["batch_size"] > 4:
-                    siliconflow_data["batch_size"] = 4
-
-                if siliconflow_data["num_inference_steps"] < 1:
-                    siliconflow_data["num_inference_steps"] = 1
-                if siliconflow_data["num_inference_steps"] > 50:
-                    siliconflow_data["num_inference_steps"] = 50
-                    
-                if siliconflow_data["guidance_scale"] < 0:
-                    siliconflow_data["guidance_scale"] = 0
-                if siliconflow_data["guidance_scale"] > 100:
-                    siliconflow_data["guidance_scale"] = 100
-
-        if "image_size" in siliconflow_data and siliconflow_data["image_size"] not in ["1024x1024", "512x1024", "768x512", "768x1024", "1024x576", "576x1024","960x1280", "720x1440", "720x1280"]:
-            siliconflow_data["image_size"] = "1024x1024"
-
-        try:
-            start_time = time.time()
-            response = requests.post(
-                "https://api.siliconflow.cn/v1/images/generations",
-                headers=headers,
-                json=siliconflow_data,
-                timeout=120
-            )
-
-            if response.status_code == 429:
-                return jsonify(response.json()), 429
-
-            response.raise_for_status()
-            end_time = time.time()
-            response_json = response.json()
-            total_time = end_time - start_time
-            
-            try:
-                images = response_json.get("images", [])
-                openai_images = []
-                for item in images:
-                    if isinstance(item, dict) and "url" in item:
-                        image_url = item["url"]
-                        print(f"image_url: {image_url}")
-                        if data.get("response_format") == "b64_json":
-                           try:
-                                image_data = requests.get(image_url, stream=True).raw
-                                image = Image.open(image_data)
-                                buffered = io.BytesIO()
-                                image.save(buffered, format="PNG")
-                                img_str = base64.b64encode(buffered.getvalue()).decode()
-                                openai_images.append({"b64_json": img_str})
-                           except Exception as e:
-                                logging.error(f"图片转base64失败: {e}")
-                                openai_images.append({"url": image_url})
-                        else:
-                            openai_images.append({"url": image_url})
-                    else:
-                        logging.error(f"无效的图片数据: {item}")
-                        openai_images.append({"url": item})
-
-
-                response_data = {
-                    "created": int(time.time()),
-                    "data": openai_images
-                }
-            except (KeyError, ValueError, IndexError) as e:
-                logging.error(
-                    f"解析响应 JSON 失败: {e}, "
-                    f"完整内容: {response_json}"
-                )
-                response_data = {
-                    "created": int(time.time()),
-                    "data": []
-                }
-
-            logging.info(
-                f"使用的key: {api_key}, "
-                f"总共用时: {total_time:.4f}秒, "
-                f"使用的模型: {model_name}"
-            )
-
-            with data_lock:
-                request_timestamps.append(time.time())
-                token_counts.append(0)
-
-            return jsonify(response_data)
-
-        except requests.exceptions.RequestException as e:
-            logging.error(f"请求转发异常: {e}")
-            return jsonify({"error": str(e)}), 500
-    else:
-        return jsonify({"error": "Unsupported model"}), 400
-
-@app.route('/handsome/v1/chat/completions', methods=['POST'])
-def handsome_chat_completions():
-    if not check_authorization(request):
-        return jsonify({"error": "Unauthorized"}), 401
-
-    data = request.get_json()
-    if not data or 'model' not in data:
-        return jsonify({"error": "Invalid request data"}), 400
-
-    model_name = data['model']
-    
-    request_type = determine_request_type(
-        model_name,
-        text_models + image_models,
-        free_text_models + free_image_models
-    )
-    
-    api_key = select_key(request_type, model_name)
-
-    if not api_key:
-        return jsonify(
-            {
-                "error": (
-                    "No available API key for this "
-                    "request type or all keys have "
-                    "reached their limits"
-                )
-            }
-        ), 429
-
-    headers = {
-        "Authorization": f"Bearer {api_key}",
-        "Content-Type": "application/json"
-    }
-    
-    if model_name in image_models:
-        user_content = ""
-        messages = data.get("messages", [])
-        for message in messages:
-            if message["role"] == "user":
-                if isinstance(message["content"], str):
-                    user_content += message["content"] + " "
-                elif isinstance(message["content"], list):
-                    for item in message["content"]:
-                        if (
-                            isinstance(item, dict) and
-                            item.get("type") == "text"
-                        ):
-                            user_content += (
-                                item.get("text", "") +
-                                " "
-                            )
-        user_content = user_content.strip()
-
-        siliconflow_data = {
-            "model": model_name,
-            "prompt": user_content,
-            
-        }
-        if model_name == "black-forest-labs/FLUX.1-pro":
-            siliconflow_data["width"] = data.get("width", 1024)
-            siliconflow_data["height"] = data.get("height", 768)
-            siliconflow_data["prompt_upsampling"] = data.get("prompt_upsampling", False)
-            siliconflow_data["image_prompt"] = data.get("image_prompt")
-            siliconflow_data["steps"] = data.get("steps", 20)
-            siliconflow_data["guidance"] = data.get("guidance", 3)
-            siliconflow_data["safety_tolerance"] = data.get("safety_tolerance", 2)
-            siliconflow_data["interval"] = data.get("interval", 2)
-            siliconflow_data["output_format"] = data.get("output_format", "png")
-            seed = data.get("seed")
-            if isinstance(seed, int) and 0 < seed < 9999999999:
-                siliconflow_data["seed"] = seed
-            if siliconflow_data["width"] < 256 or siliconflow_data["width"] > 1440 or siliconflow_data["width"] % 32 != 0:
-                siliconflow_data["width"] = 1024
-            if siliconflow_data["height"] < 256 or siliconflow_data["height"] > 1440 or siliconflow_data["height"] % 32 != 0:
-                 siliconflow_data["height"] = 768
-
-            if siliconflow_data["steps"] < 1 or siliconflow_data["steps"] > 50:
-                siliconflow_data["steps"] = 20
-            if siliconflow_data["guidance"] < 1.5 or siliconflow_data["guidance"] > 5:
-                 siliconflow_data["guidance"] = 3
-            if siliconflow_data["safety_tolerance"] < 0 or siliconflow_data["safety_tolerance"] > 6:
-                siliconflow_data["safety_tolerance"] = 2
-            if siliconflow_data["interval"] < 1 or siliconflow_data["interval"] > 4 :
-                 siliconflow_data["interval"] = 2
-        else:
-            siliconflow_data["image_size"] = "1024x1024"
-            siliconflow_data["batch_size"] = 1
-            siliconflow_data["num_inference_steps"] = 20
-            siliconflow_data["guidance_scale"] = 7.5
-            siliconflow_data["prompt_enhancement"] = False
-        
-            if data.get("size"):
-                siliconflow_data["image_size"] = data.get("size")
-            if data.get("n"):
-                siliconflow_data["batch_size"] = data.get("n")
-            if data.get("steps"):
-                siliconflow_data["num_inference_steps"] = data.get("steps")
-            if data.get("guidance_scale"):
-               siliconflow_data["guidance_scale"] = data.get("guidance_scale")
-            if data.get("negative_prompt"):
-               siliconflow_data["negative_prompt"] = data.get("negative_prompt")
-            if data.get("seed"):
-               siliconflow_data["seed"] = data.get("seed")
-            if data.get("prompt_enhancement"):
-               siliconflow_data["prompt_enhancement"] = data.get("prompt_enhancement")
-
-            if siliconflow_data["batch_size"] < 1:
-                siliconflow_data["batch_size"] = 1
-            if siliconflow_data["batch_size"] > 4:
-                siliconflow_data["batch_size"] = 4
-
-            if siliconflow_data["num_inference_steps"] < 1:
-                siliconflow_data["num_inference_steps"] = 1
-            if siliconflow_data["num_inference_steps"] > 50:
-                siliconflow_data["num_inference_steps"] = 50
-                
-            if siliconflow_data["guidance_scale"] < 0:
-                siliconflow_data["guidance_scale"] = 0
-            if siliconflow_data["guidance_scale"] > 100:
-                siliconflow_data["guidance_scale"] = 100
-            
-            if siliconflow_data["image_size"] not in ["1024x1024", "512x1024", "768x512", "768x1024", "1024x576", "576x1024", "960x1280", "720x1440", "720x1280"]:
-                siliconflow_data["image_size"] = "1024x1024"
-
-        try:
-            start_time = time.time()
-            response = requests.post(
-                "https://api.siliconflow.cn/v1/images/generations",
-                headers=headers,
-                json=siliconflow_data,
-                timeout=120,
-                stream=data.get("stream", False)
-            )
-           
-            if response.status_code == 429:
-                return jsonify(response.json()), 429
-
-            if data.get("stream", False):
-                def generate():
-                    first_chunk_time = None
-                    full_response_content = ""
-                    try:
-                        response.raise_for_status()
-                        end_time = time.time()
-                        response_json = response.json()
-                        total_time = end_time - start_time
-                    
-                        images = response_json.get("images", [])
-                        
-                        image_url = ""
-                        if images and isinstance(images[0], dict) and "url" in images[0]:
-                            image_url = images[0]["url"]
-                            logging.info(f"Extracted image URL: {image_url}")
-                        elif images and isinstance(images[0], str):
-                            image_url = images[0]
-                            logging.info(f"Extracted image URL: {image_url}")
-                        
-                        markdown_image_link = f"![image]({image_url})"
-                        if image_url:
-                            chunk_data = {
-                                "id": f"chatcmpl-{uuid.uuid4()}",
-                                "object": "chat.completion.chunk",
-                                "created": int(time.time()),
-                                "model": model_name,
-                                "choices": [
-                                    {
-                                        "index": 0,
-                                        "delta": {
-                                            "role": "assistant",
-                                            "content": markdown_image_link
-                                        },
-                                        "finish_reason": None
-                                    }
-                                ]
-                            }
-                            yield f"data: {json.dumps(chunk_data)}\n\n".encode('utf-8')
-                            full_response_content = markdown_image_link
-                        else:
-                            chunk_data = {
-                                "id": f"chatcmpl-{uuid.uuid4()}",
-                                "object": "chat.completion.chunk",
-                                "created": int(time.time()),
-                                "model": model_name,
-                                "choices": [
-                                    {
-                                        "index": 0,
-                                        "delta": {
-                                            "role": "assistant",
-                                            "content": "Failed to generate image"
-                                        },
-                                        "finish_reason": None
-                                    }
-                                ]
-                            }
-                            yield f"data: {json.dumps(chunk_data)}\n\n".encode('utf-8')
-                            full_response_content = "Failed to generate image"
-                        
-                        end_chunk_data = {
-                            "id": f"chatcmpl-{uuid.uuid4()}",
-                            "object": "chat.completion.chunk",
-                            "created": int(time.time()),
-                            "model": model_name,
-                            "choices": [
-                                {
-                                    "index": 0,
-                                    "delta": {},
-                                    "finish_reason": "stop"
-                                }
-                            ]
-                        }
-                        yield f"data: {json.dumps(end_chunk_data)}\n\n".encode('utf-8')
-                        with data_lock:
-                            request_timestamps.append(time.time())
-                            token_counts.append(0)
-                    except requests.exceptions.RequestException as e:
-                        logging.error(f"请求转发异常: {e}")
-                        error_chunk_data = {
-                            "id": f"chatcmpl-{uuid.uuid4()}",
-                            "object": "chat.completion.chunk",
-                            "created": int(time.time()),
-                            "model": model_name,
-                            "choices": [
-                                {
-                                    "index": 0,
-                                    "delta": {
-                                        "role": "assistant",
-                                        "content": f"Error: {str(e)}"
-                                    },
-                                    "finish_reason": None
-                                }
-                            ]
-                        }
-                        yield f"data: {json.dumps(error_chunk_data)}\n\n".encode('utf-8')
-                        end_chunk_data = {
-                                "id": f"chatcmpl-{uuid.uuid4()}",
-                                "object": "chat.completion.chunk",
-                                "created": int(time.time()),
-                                "model": model_name,
-                                "choices": [
-                                    {
-                                        "index": 0,
-                                        "delta": {},
-                                        "finish_reason": "stop"
-                                    }
-                                ]
-                            }
-                        yield f"data: {json.dumps(end_chunk_data)}\n\n".encode('utf-8')
-                    logging.info(
-                        f"使用的key: {api_key}, "
-                        f"使用的模型: {model_name}"
-                    )
-                    yield "data: [DONE]\n\n".encode('utf-8')
-                return Response(stream_with_context(generate()), content_type='text/event-stream')
+        if data.get("stream", False):
+            def generate():
+                first_chunk_time = None
+                full_response_content = ""
+                for chunk in response.iter_content(chunk_size=1024):
+                    if chunk:
+                        if first_chunk_time is None:
+                            first_chunk_time = time.time()
+                        full_response_content += chunk.decode("utf-8")
+                        yield chunk
 
-            else:
-                response.raise_for_status()
                 end_time = time.time()
-                response_json = response.json()
-                total_time = end_time - start_time
-            
-                try:
-                    images = response_json.get("images", [])
-                    
-                    image_url = ""
-                    if images and isinstance(images[0], dict) and "url" in images[0]:
-                        image_url = images[0]["url"]
-                        logging.info(f"Extracted image URL: {image_url}")
-                    elif images and isinstance(images[0], str):
-                        image_url = images[0]
-                        logging.info(f"Extracted image URL: {image_url}")
-                    
-                    markdown_image_link = f"![image]({image_url})"
-                    response_data = {
-                        "id": f"chatcmpl-{uuid.uuid4()}",
-                        "object": "chat.completion",
-                        "created": int(time.time()),
-                        "model": model_name,
-                        "choices": [
-                            {
-                            "index": 0,
-                            "message": {
-                                "role": "assistant",
-                                "content": markdown_image_link if image_url else "Failed to generate image",
-                            },
-                            "finish_reason": "stop",
-                            }
-                        ],
-                    }
-                except (KeyError, ValueError, IndexError) as e:
-                    logging.error(
-                        f"解析响应 JSON 失败: {e}, "
-                        f"完整内容: {response_json}"
-                    )
-                    response_data = {
-                        "id": f"chatcmpl-{uuid.uuid4()}",
-                        "object": "chat.completion",
-                        "created": int(time.time()),
-                        "model": model_name,
-                        "choices": [
-                            {
-                            "index": 0,
-                            "message": {
-                                "role": "assistant",
-                                "content": "Failed to process image data",
-                            },
-                            "finish_reason": "stop",
-                            }
-                        ],
-                    }
-
-                logging.info(
-                    f"使用的key: {api_key}, "
-                    f"总共用时: {total_time:.4f}秒, "
-                    f"使用的模型: {model_name}"
-                )
-                with data_lock:
-                    request_timestamps.append(time.time())
-                    token_counts.append(0)
-                return jsonify(response_data)
-
-        except requests.exceptions.RequestException as e:
-            logging.error(f"请求转发异常: {e}")
-            return jsonify({"error": str(e)}), 500
-    else:
-        try:
-            start_time = time.time()
-            response = requests.post(
-                TEST_MODEL_ENDPOINT,
-                headers=headers,
-                json=data,
-                stream=data.get("stream", False),
-                timeout=60
-            )
-
-            if response.status_code == 429:
-                return jsonify(response.json()), 429
-
-            if data.get("stream", False):
-                def generate():
-                    first_chunk_time = None
-                    full_response_content = ""
-                    for chunk in response.iter_content(chunk_size=1024):
-                        if chunk:
-                            if first_chunk_time is None:
-                                first_chunk_time = time.time()
-                            full_response_content += chunk.decode("utf-8")
-                            yield chunk
-
-                    end_time = time.time()
-                    first_token_time = (
-                        first_chunk_time - start_time
-                        if first_chunk_time else 0
-                    )
-                    total_time = end_time - start_time
-
-                    prompt_tokens = 0
-                    completion_tokens = 0
-                    response_content = ""
-                    for line in full_response_content.splitlines():
-                        if line.startswith("data:"):
-                            line = line[5:].strip()
-                            if line == "[DONE]":
-                                continue
-                            try:
-                                response_json = json.loads(line)
-
-                                if (
-                                    "usage" in response_json and
-                                    "completion_tokens" in response_json["usage"]
-                                ):
-                                    completion_tokens = response_json[
-                                        "usage"
-                                    ]["completion_tokens"]
-
-                                if (
-                                    "choices" in response_json and
-                                    len(response_json["choices"]) > 0 and
-                                    "delta" in response_json["choices"][0] and
-                                    "content" in response_json[
-                                        "choices"
-                                    ][0]["delta"]
-                                ):
-                                    response_content += response_json[
-                                        "choices"
-                                    ][0]["delta"]["content"]
-
-                                if (
-                                    "usage" in response_json and
-                                    "prompt_tokens" in response_json["usage"]
-                                ):
-                                    prompt_tokens = response_json[
-                                        "usage"
-                                    ]["prompt_tokens"]
-
-                            except (
-                                KeyError,
-                                ValueError,
-                                IndexError
-                            ) as e:
-                                logging.error(
-                                    f"解析流式响应单行 JSON 失败: {e}, "
-                                    f"行内容: {line}"
-                                )
-
-                    user_content = ""
-                    messages = data.get("messages", [])
-                    for message in messages:
-                        if message["role"] == "user":
-                            if isinstance(message["content"], str):
-                                user_content += message["content"] + " "
-                            elif isinstance(message["content"], list):
-                                for item in message["content"]:
-                                    if (
-                                        isinstance(item, dict) and
-                                        item.get("type") == "text"
-                                    ):
-                                        user_content += (
-                                            item.get("text", "") +
-                                            " "
-                                        )
-
-                    user_content = user_content.strip()
-
-                    user_content_replaced = user_content.replace(
-                        '\n', '\\n'
-                    ).replace('\r', '\\n')
-                    response_content_replaced = response_content.replace(
-                        '\n', '\\n'
-                    ).replace('\r', '\\n')
-
-                    logging.info(
-                        f"使用的key: {api_key}, "
-                        f"提示token: {prompt_tokens}, "
-                        f"输出token: {completion_tokens}, "
-                        f"首字用时: {first_token_time:.4f}秒, "
-                        f"总共用时: {total_time:.4f}秒, "
-                        f"使用的模型: {model_name}, "
-                        f"用户的内容: {user_content_replaced}, "
-                        f"输出的内容: {response_content_replaced}"
-                    )
-
-                    with data_lock:
-                        request_timestamps.append(time.time())
-                        token_counts.append(prompt_tokens+completion_tokens)
-
-                return Response(
-                    stream_with_context(generate()),
-                    content_type=response.headers['Content-Type']
+                first_token_time = (
+                    first_chunk_time - start_time
+                    if first_chunk_time else 0
                 )
-            else:
-                response.raise_for_status()
-                end_time = time.time()
-                response_json = response.json()
                 total_time = end_time - start_time
 
-                try:
-                    prompt_tokens = response_json["usage"]["prompt_tokens"]
-                    completion_tokens = response_json[
-                        "usage"
-                    ]["completion_tokens"]
-                    response_content = response_json[
-                        "choices"
-                    ][0]["message"]["content"]
-                except (KeyError, ValueError, IndexError) as e:
-                    logging.error(
-                        f"解析非流式响应 JSON 失败: {e}, "
-                        f"完整内容: {response_json}"
-                    )
-                    prompt_tokens = 0
-                    completion_tokens = 0
-                    response_content = ""
+                prompt_tokens = 0
+                completion_tokens = 0
+                response_content = ""
+                for line in full_response_content.splitlines():
+                    if line.startswith("data:"):
+                        line = line[5:].strip()
+                        if line == "[DONE]":
+                            continue
+                        try:
+                            response_json = json.loads(line)
+
+                            if (
+                                "usage" in response_json and
+                                "completion_tokens" in response_json["usage"]
+                            ):
+                                completion_tokens = response_json[
+                                    "usage"
+                                ]["completion_tokens"]
+
+                            if (
+                                "choices" in response_json and
+                                len(response_json["choices"]) > 0 and
+                                "delta" in response_json["choices"][0] and
+                                "content" in response_json[
+                                    "choices"
+                                ][0]["delta"]
+                            ):
+                                response_content += response_json[
+                                    "choices"
+                                ][0]["delta"]["content"]
+
+                            if (
+                                "usage" in response_json and
+                                "prompt_tokens" in response_json["usage"]
+                            ):
+                                prompt_tokens = response_json[
+                                    "usage"
+                                ]["prompt_tokens"]
+
+                        except (
+                            KeyError,
+                            ValueError,
+                            IndexError
+                        ) as e:
+                            logging.error(
+                                f"解析流式响应单行 JSON 失败: {e}, "
+                                f"行内容: {line}"
+                            )
 
                 user_content = ""
                 messages = data.get("messages", [])
@@ -1477,7 +855,8 @@ def handsome_chat_completions():
                                     item.get("type") == "text"
                                 ):
                                     user_content += (
-                                        item.get("text", "") + " "
+                                        item.get("text", "") +
+                                        " "
                                     )
 
                 user_content = user_content.strip()
@@ -1493,24 +872,91 @@ def handsome_chat_completions():
                     f"使用的key: {api_key}, "
                     f"提示token: {prompt_tokens}, "
                     f"输出token: {completion_tokens}, "
-                    f"首字用时: 0, "
+                    f"首字用时: {first_token_time:.4f}秒, "
                     f"总共用时: {total_time:.4f}秒, "
                     f"使用的模型: {model_name}, "
                     f"用户的内容: {user_content_replaced}, "
                     f"输出的内容: {response_content_replaced}"
                 )
+
                 with data_lock:
                     request_timestamps.append(time.time())
-                    if "prompt_tokens" in response_json["usage"] and "completion_tokens" in response_json["usage"]:
-                        token_counts.append(response_json["usage"]["prompt_tokens"] + response_json["usage"]["completion_tokens"])
-                    else:
-                        token_counts.append(0)
+                    token_counts.append(prompt_tokens+completion_tokens)
+
+            return Response(
+                stream_with_context(generate()),
+                content_type=response.headers['Content-Type']
+            )
+        else:
+            response.raise_for_status()
+            end_time = time.time()
+            response_json = response.json()
+            total_time = end_time - start_time
+
+            try:
+                prompt_tokens = response_json["usage"]["prompt_tokens"]
+                completion_tokens = response_json[
+                    "usage"
+                ]["completion_tokens"]
+                response_content = response_json[
+                    "choices"
+                ][0]["message"]["content"]
+            except (KeyError, ValueError, IndexError) as e:
+                logging.error(
+                    f"解析非流式响应 JSON 失败: {e}, "
+                    f"完整内容: {response_json}"
+                )
+                prompt_tokens = 0
+                completion_tokens = 0
+                response_content = ""
+
+            user_content = ""
+            messages = data.get("messages", [])
+            for message in messages:
+                if message["role"] == "user":
+                    if isinstance(message["content"], str):
+                        user_content += message["content"] + " "
+                    elif isinstance(message["content"], list):
+                        for item in message["content"]:
+                            if (
+                                isinstance(item, dict) and
+                                item.get("type") == "text"
+                            ):
+                                user_content += (
+                                    item.get("text", "") + " "
+                                )
 
-                return jsonify(response_json)
+            user_content = user_content.strip()
 
-        except requests.exceptions.RequestException as e:
-            logging.error(f"请求转发异常: {e}")
-            return jsonify({"error": str(e)}), 500
+            user_content_replaced = user_content.replace(
+                '\n', '\\n'
+            ).replace('\r', '\\n')
+            response_content_replaced = response_content.replace(
+                '\n', '\\n'
+            ).replace('\r', '\\n')
+
+            logging.info(
+                f"使用的key: {api_key}, "
+                f"提示token: {prompt_tokens}, "
+                f"输出token: {completion_tokens}, "
+                f"首字用时: 0, "
+                f"总共用时: {total_time:.4f}秒, "
+                f"使用的模型: {model_name}, "
+                f"用户的内容: {user_content_replaced}, "
+                f"输出的内容: {response_content_replaced}"
+            )
+            with data_lock:
+                request_timestamps.append(time.time())
+                if "prompt_tokens" in response_json["usage"] and "completion_tokens" in response_json["usage"]:
+                    token_counts.append(response_json["usage"]["prompt_tokens"] + response_json["usage"]["completion_tokens"])
+                else:
+                    token_counts.append(0)
+
+            return jsonify(response_json)
+
+    except requests.exceptions.RequestException as e:
+        logging.error(f"请求转发异常: {e}")
+        return jsonify({"error": str(e)}), 500
 
 if __name__ == '__main__':
     import json