feat flask直接启动，解决部分机cpu占用过大问题

.github/workflows/ninja-image.yml

@@ -42,5 +42,6 @@ jobs:
         push: true
         tags: |
           yangclivia/pandora-to-api:${{ steps.tag_name.outputs.tag }}
+          yangclivia/pandora-to-api:0.7.7
         platforms: linux/amd64,linux/arm64
         build-args: TARGETPLATFORM=${{ matrix.platform }}

.github/workflows/xyhelper-deploy.yml

@@ -42,5 +42,6 @@ jobs:
         push: true
         tags: |
           yangclivia/pandora-to-api:${{ steps.tag_name.outputs.tag }}
+          yangclivia/pandora-to-api:0.7.8
         platforms: linux/amd64,linux/arm64
         build-args: TARGETPLATFORM=${{ matrix.platform }}

.gitignore

@@ -0,0 +1,5 @@
+
+*.json
+*.log
+*.json
+*.log

.idea/encodings.xml

@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<project version="4">
+  <component name="Encoding">
+    <file url="file://$PROJECT_DIR$/log/access.log" charset="GBK" />
+  </component>
+</project>

Dockerfile

@@ -10,15 +10,13 @@ COPY . /app
 # 设置环境变量
 ENV PYTHONUNBUFFERED=1
 
-RUN chmod +x /app/start.sh
-
-RUN apt update && apt install -y jq
+RUN chmod +x /app/main.py
 
 # # 设置 pip 源为清华大学镜像
 # RUN pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
 
 # 安装任何所需的依赖项
-RUN pip install --no-cache-dir flask flask_apscheduler gunicorn requests Pillow flask-cors tiktoken fake_useragent redis websocket-client pysocks requests[socks] websocket-client[optional]
+RUN pip install --no-cache-dir flask flask_apscheduler requests Pillow flask-cors tiktoken fake_useragent redis websocket-client pysocks requests[socks] websocket-client[optional]
 
 # 在容器启动时运行 Flask 应用
-CMD ["/app/start.sh"]
+CMD ["python3", "main.py"]

Readme.md

@@ -1,5 +1,5 @@
-## 项目简介
-
+## 0.7.8 xyhelper项目简介
+![Docker Image Size (tag)](https://img.shields.io/docker/image-size/yangclivia/pandora-to-api/0.7.8)![Docker Pulls](https://img.shields.io/docker/pulls/yangclivia/pandora-to-api)[![GitHub Repo stars](https://img.shields.io/github/stars/Yanyutin753/RefreshToV1Api?style=social)](https://github.com/Yanyutin753/refresh-gpt-chat/stargazers)
 > [!IMPORTANT]
 >
 > Respect `xyhelper` ,Respect `ninja` , Respect `Wizerd`！
@@ -14,7 +14,7 @@
 
 3. 支持直接把refresh_token作为请求key，方便接入one_api
 
-4. 支持 gpt-4-mobile 、gpt-4-s 、基本所有的GPTS
+4. 支持 gpt-4-mobile 、gpt-4-s 、动态支持所有的gpt-4-gizmo-XXX模型
 
 * **xyhelper 的 免费 backend-api 接口，无需打码**
 
@@ -99,9 +99,9 @@
 
 - `need_log_to_file`: 用于设置是否需要将日志输出到文件，可选值为：`true`、`false`，默认为 `true`，日志文件路径为：`./log/access.log`，默认每天会自动分割日志文件。
 
-- `process_workers`: 用于设置进程数，如果不需要设置，可以保持不变，如果需要设置，可以设置为需要设置的值，如果设置为 `1`，则会强制设置为单进程模式。
+- `process_workers`: 用于设置进程数，如果不需要设置，可以保持不变，如果需要设置，可以设置为需要设置的值，默认为 `2`。
 
-- `process_threads`: 用于设置线程数，如果不需要设置，可以保持不变，如果需要设置，可以设置为需要设置的值，如果设置为 `1`，则会强制设置为单线程模式。
+- `process_threads`: 用于设置线程数，如果不需要设置，可以保持不变，如果需要设置，可以设置为需要设置的值，默认为 `2`。
 
 - `upstream_base_url`: xyhelper 的免费接口地址，如：`https://demo.xyhelper.cn`，注意：不要以 `/` 结尾。
 
@@ -166,7 +166,9 @@ PS. 注意，arkose_urls中的地址需要支持PandoraNext的Arkose Token获取
 
 ## GPTS配置说明
 
-如果需要使用 GPTS，需要修改 `gpts.json` 文件，其中每个对象的key即为调用对应 GPTS 的时候使用的模型名称，而 `id` 则为对应的模型id，该 `id` 对应每个 GPTS 的链接的后缀。配置多个GPTS的时候用逗号隔开。
+### 使用 GPTS
+
+1. 可修改 `gpts.json` 文件，其中每个对象的key即为调用对应 GPTS 的时候使用的模型名称，而 `id` 则为对应的模型id，该 `id` 对应每个 GPTS 的链接的后缀。配置多个GPTS的时候用逗号隔开。
 
 例如：PandoraNext的官方 GPTS 的链接为：`https://chat.oaifree.com/g/g-CFsXuTRfy-pandoranextzhu-shou`，则该模型的 `id` 的值应为 `g-CFsXuTRfy-pandoranextzhu-shou`，而模型名可以自定义。
 
@@ -183,7 +185,16 @@ PS. 注意，arkose_urls中的地址需要支持PandoraNext的Arkose Token获取
 }
 ```
 
-注意：使用该配置的时候需要保证正确填写 `docker-compose.yml` 的环境变量 `KEY_FOR_GPTS_INFO`，同时该变量设置的 `key` 允许访问所有配置的 GPTS。
+2. 可直接请求的时候加上相应的gpt-4-gizmo-XXX
+```json
+{
+    "stream":true,
+    "model":"gpt-4-gizmo-XXXX",
+    "messages": [{"role": "user", "content": "你是什么模型"}]
+}
+```
+
+注意：使用该配置的时候需要保证正确填写 `config.json` 文件的环境变量 `KEY_FOR_GPTS_INFO`，同时该变量设置的 `key` 允许访问所有配置的 GPTS。
 
 ## 绘图接口使用说明
 
@@ -324,19 +335,20 @@ PS. 注意，arkose_urls中的地址需要支持PandoraNext的Arkose Token获取
 
 ## 示例
 
-以ChatGPT-Next-Web项目的docker-compose部署为例，这里提供一个简单的部署配置文件示例：
+以[ChatGPT-Next-Web](https://github.com/Yanyutin753/ChatGPT-Next-Web-LangChain-Gpt-4-All)项目插件版的docker-compose部署为例，支持完全适配项目，这里提供一个简单的部署配置文件示例：
 
 ```
 version: '3'
 services:
   chatgpt-next-web:
-    image: yidadaa/chatgpt-next-web
+    image: yangclivia/chatgpt-next-web-langchain
     ports:
       - "50013:3000"
     environment:
+      - CUSTOM_MODELS=-all,+gpt-3.5-turbo,+gpt-4-s,+gpt-4-mobile,+gpt-4-vision-preview,+gpt-4-gizmo-XXX
       - OPENAI_API_KEY=<正确的refresh_token>
       - BASE_URL=<backend-to-api容器地址>
-      - CUSTOM_MODELS=+gpt-4-s,+gpt-4-mobile,+<gpts.json 中的模型名>
+      - CUSTOM_MODELS=-gpt-4-0613,-gpt-4-32k,-gpt-4-32k-0613,-gpt-4-turbo-preview,-gpt-4-1106-preview,-gpt-4-0125-preview,-gpt-3.5-turbo-0125,-gpt-3.5-turbo-0613,-gpt-3.5-turbo-1106,-gpt-3.5-turbo-16k,-gpt-3.5-turbo-16k-0613,+gpt-3.5-turbo,+gpt-4,+gpt-4-mobile,+gpt-4-vision-preview,+gpt-4-mobile,+<gpts.json 中的模型名>
 
 ```
 
@@ -351,17 +363,20 @@ services:
 
 ![api-2](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/37c0381f-a70a-42bb-83f1-1491053240b7)
 
+### 读文件
+![api-3](https://github.com/Yanyutin753/xyhelperToV1Api_refresh/assets/132346501/e1f2e770-cd33-4af0-a0c7-5ddabf1c076d)
+
 ### 绘图
 
-![api-3](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/8eea9436-12ee-46b1-86c1-67e7e97da83a)
+![api-4](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/8eea9436-12ee-46b1-86c1-67e7e97da83a)
 
 ### GPT-4-Mobile
 
-![api-4](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/2eb4fd4f-7c66-4a1f-a54a-3c280a36e509)
+![api-5](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/2eb4fd4f-7c66-4a1f-a54a-3c280a36e509)
 
 ### GPTS
 
-![api-5](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/299df56a-d245-4920-8892-94e1a9cc644a)
+![api-6](https://github.com/Ink-Osier/PandoraToV1Api/assets/133617214/299df56a-d245-4920-8892-94e1a9cc644a)
 
 ### Bot 模式
 

data/config.json

@@ -26,7 +26,7 @@
     },
     "refresh_ToAccess": {
         "stream_sleep_time": 0,
-        "enableOai":"true",
+        "enableOai":"false",
         "xyhelper_refreshToAccess_Url": "https://demo.xyhelper.cn/applelogin"
     },
     "redis": {

main.py

@@ -202,7 +202,7 @@ def oaiGetAccessToken(refresh_token):
 def xyhelperGetAccessToken(getAccessTokenUrl, refresh_token):
     try:
         logger.info("将通过这个网址请求access_token：" + getAccessTokenUrl)
-        
+
         data = {
             'refresh_token': refresh_token,
         }
@@ -255,7 +255,7 @@ def fetch_gizmo_info(base_url, proxy_api_prefix, model_id):
 # 将配置添加到全局列表
 def add_config_to_global_list(base_url, proxy_api_prefix, gpts_data):
     global gpts_configurations
-    updateGptsKey()
+    updateGptsKey()  # cSpell:ignore Gpts
     # print(f"gpts_data: {gpts_data}")
     for model_name, model_info in gpts_data.items():
         # print(f"model_name: {model_name}")
@@ -271,22 +271,20 @@ def add_config_to_global_list(base_url, proxy_api_prefix, gpts_data):
         else:
             logger.info(f"Fetching gpts info for {model_name}, {model_id}")
             gizmo_info = fetch_gizmo_info(base_url, proxy_api_prefix, model_id)
-
             # 如果成功获取到数据，则将其存入 Redis
             if gizmo_info:
                 redis_client.set(model_id, str(gizmo_info))
                 logger.info(f"Cached gizmo info for {model_name}, {model_id}")
 
-        if gizmo_info:
-            # 检查模型名称是否已经在列表中
-            if not any(d['name'] == model_name for d in gpts_configurations):
-                gpts_configurations.append({
-                    'name': model_name,
-                    'id': model_id,
-                    'config': gizmo_info
-                })
-            else:
-                logger.info(f"Model already exists in the list, skipping...")
+        # 检查模型名称是否已经在列表中
+        if gizmo_info and not any(d['name'] == model_name for d in gpts_configurations):
+            gpts_configurations.append({
+                'name': model_name,
+                'id': model_id,
+                'config': gizmo_info
+            })
+        else:
+            logger.info(f"Model already exists in the list, skipping...")
 
 
 def generate_gpts_payload(model, messages):
@@ -324,9 +322,11 @@ scheduler.start()
 # PANDORA_UPLOAD_URL = 'files.pandoranext.com'
 
 
-VERSION = '0.7.8'
+VERSION = '0.7.8.3'
 # VERSION = 'test'
-UPDATE_INFO = '项目将脱离ninja，使用xyhelper，xyhelper_refreshToAccess_Url等配置需修改'
+UPDATE_INFO = 'flask直接启动，解决部分机cpu占用过大问题'
+
+
 # UPDATE_INFO = '【仅供临时测试使用】 '
 
 # 解析响应中的信息
@@ -864,10 +864,14 @@ def send_text_prompt_and_get_response(messages, api_key, stream, model):
 
     # 查找模型配置
     model_config = find_model_config(model)
-    if model_config:
+    if model_config or 'gpt-4-gizmo-' in model:
         # 检查是否有 ori_name
-        ori_model_name = model_config.get('ori_name', model)
-        logger.info(f"原模型名: {ori_model_name}")
+        if model_config:
+            ori_model_name = model_config.get('ori_name', model)
+            logger.info(f"原模型名: {ori_model_name}")
+        else:
+            ori_model_name = model
+            logger.info(f"请求模型名: {model}")
         if ori_model_name == 'gpt-4-s':
             payload = {
                 # 构建 payload
@@ -918,6 +922,54 @@ def send_text_prompt_and_get_response(messages, api_key, stream, model):
                 "force_paragen": False,
                 "force_rate_limit": False
             }
+        elif 'gpt-4-gizmo-' in model:
+            payload = generate_gpts_payload(model, formatted_messages)
+            if not payload:
+                global gpts_configurations
+                # 假设 model是 'gpt-4-gizmo-123'
+                split_name = model.split('gpt-4-gizmo-')
+                model_id = split_name[1] if len(split_name) > 1 else None
+                gizmo_info = fetch_gizmo_info(BASE_URL, PROXY_API_PREFIX, model_id)
+                logging.info(gizmo_info)
+
+                # 如果成功获取到数据，则将其存入 Redis
+                if gizmo_info:
+                    redis_client.set(model_id, str(gizmo_info))
+                    logger.info(f"Cached gizmo info for {model}, {model_id}")
+                    # 检查模型名称是否已经在列表中
+                    if not any(d['name'] == model for d in gpts_configurations):
+                        gpts_configurations.append({
+                            'name': model,
+                            'id': model_id,
+                            'config': gizmo_info
+                        })
+                    else:
+                        logger.info(f"Model already exists in the list, skipping...")
+                    payload = generate_gpts_payload(model, formatted_messages)
+                else:
+                    payload = {
+                        # 构建 payload
+                        "action": "next",
+                        "messages": formatted_messages,
+                        "parent_message_id": str(uuid.uuid4()),
+                        "model": "text-davinci-002-render-sha",
+                        "timezone_offset_min": -480,
+                        "suggestions": [
+                            "What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter.",
+                            "I want to cheer up my friend who's having a rough day. Can you suggest a couple short and sweet text messages to go with a kitten gif?",
+                            "Come up with 5 concepts for a retro-style arcade game.",
+                            "I have a photoshoot tomorrow. Can you recommend me some colors and outfit options that will look good on camera?"
+                        ],
+                        "history_and_training_disabled": False,
+                        "arkose_token": None,
+                        "conversation_mode": {
+                            "kind": "primary_assistant"
+                        },
+                        "force_paragen": False,
+                        "force_rate_limit": False
+                    }
+                    logger.debug('KEY_FOR_GPTS_INFO Or Request Model is not accessible')
+
         else:
             payload = generate_gpts_payload(model, formatted_messages)
             if not payload:
@@ -1093,7 +1145,7 @@ def replace_sandbox(text, conversation_id, message_id, api_key):
         sandbox_path = match.group(1)
         download_url = get_download_url(conversation_id, message_id, sandbox_path)
         if download_url == None:
-            return "\n```\nError: 沙箱文件下载失败，这可能是因为您启用了隐私模式\n```"
+            return "\n```\nError: 沙箱文件下载失败，这可能是因为您的帐号启用了隐私模式\n```"
         file_name = extract_filename(download_url)
         timestamped_file_name = timestamp_filename(file_name)
         if USE_OAIUSERCONTENT_URL == False:
@@ -1161,7 +1213,7 @@ def generate_actions_allow_payload(author_role, author_name, target_message_id,
             "action": "next",
             "messages": [
                 {
-                    "id": generate_custom_uuid_v4(),
+                    "id": str(uuid.uuid4()),
                     "author": {
                         "role": author_role,
                         "name": author_name
@@ -2012,7 +2064,7 @@ def old_data_fetcher(upstream_response, data_queue, stop_event, last_data_time,
                 "id": chat_message_id,
                 "object": "chat.completion.chunk",
                 "created": timestamp,
-                "model": message.get("metadata", {}).get("model_slug"),
+                "model": model,
                 "choices": [
                     {
                         "index": 0,
@@ -2248,8 +2300,10 @@ def chat_completions():
     messages = data.get('messages')
     model = data.get('model')
     accessible_model_list = get_accessible_model_list()
-    if model not in accessible_model_list:
+    if model not in accessible_model_list and not 'gpt-4-gizmo-' in model:
         return jsonify({"error": "model is not accessible"}), 401
+    elif 'gpt-4-gizmo-' in model and not KEY_FOR_GPTS_INFO:
+        return jsonify({"error": "key_for_gpts_info is not accessible"}), 400
 
     stream = data.get('stream', False)
 
@@ -2368,32 +2422,36 @@ def chat_completions():
             ori_model_name = model_config.get('ori_name', model)
         input_tokens = count_total_input_words(messages, ori_model_name)
         comp_tokens = count_tokens(all_new_text, ori_model_name)
-        response_json = {
-            "id": generate_unique_id("chatcmpl"),
-            "object": "chat.completion",
-            "created": int(time.time()),  # 使用当前时间戳
-            "model": model,  # 使用请求中指定的模型
-            "choices": [
-                {
-                    "index": 0,
-                    "message": {
-                        "role": "assistant",
-                        "content": all_new_text  # 使用累积的文本
-                    },
-                    "finish_reason": "stop"
-                }
-            ],
-            "usage": {
-                # 这里的 token 计数需要根据实际情况计算
-                "prompt_tokens": input_tokens,
-                "completion_tokens": comp_tokens,
-                "total_tokens": input_tokens + comp_tokens
-            },
-            "system_fingerprint": None
-        }
-
-        # 返回 JSON 响应
-        return jsonify(response_json)
+        if input_tokens >= 100 and comp_tokens <= 0:
+            # 返回错误消息和状态码429
+            error_response = {"error": "空回复"}
+            return jsonify(error_response), 429
+        else:
+            response_json = {
+                "id": generate_unique_id("chatcmpl"),
+                "object": "chat.completion",
+                "created": int(time.time()),  # 使用当前时间戳
+                "model": model,  # 使用请求中指定的模型
+                "choices": [
+                    {
+                        "index": 0,
+                        "message": {
+                            "role": "assistant",
+                            "content": all_new_text  # 使用累积的文本
+                        },
+                        "finish_reason": "stop"
+                    }
+                ],
+                "usage": {
+                    # 这里的 token 计数需要根据实际情况计算
+                    "prompt_tokens": input_tokens,
+                    "completion_tokens": comp_tokens,
+                    "total_tokens": input_tokens + comp_tokens
+                },
+                "system_fingerprint": None
+            }
+            # 返回 JSON 响应
+            return jsonify(response_json)
     else:
         return Response(generate(), mimetype='text/event-stream')
 
@@ -2406,8 +2464,10 @@ def images_generations():
     # messages = data.get('messages')
     model = data.get('model')
     accessible_model_list = get_accessible_model_list()
-    if model not in accessible_model_list:
+    if model not in accessible_model_list and not 'gpt-4-gizmo-' in model:
         return jsonify({"error": "model is not accessible"}), 401
+    elif 'gpt-4-gizmo-' in model and not KEY_FOR_GPTS_INFO:
+        return jsonify({"error": "key_for_gpts_info is not accessible"}), 400
 
     prompt = data.get('prompt', '')
 
@@ -2639,4 +2699,4 @@ scheduler.add_job(id='updateRefresh_run', func=updateRefresh_dict, trigger='cron
 
 # 运行 Flask 应用
 if __name__ == '__main__':
-    app.run(host='0.0.0.0')
+    app.run(host='0.0.0.0', port=33333, threaded=True)

start.sh

@@ -11,7 +11,7 @@ if [ -z "$PROCESS_WORKERS" ]; then
     export PROCESS_WORKERS
 
     if [ -z "$PROCESS_WORKERS" ]; then
-        PROCESS_WORKERS=1
+        PROCESS_WORKERS=2
     fi
 fi
 
@@ -31,4 +31,5 @@ echo "PROCESS_WORKERS: ${PROCESS_WORKERS}"
 echo "PROCESS_THREADS: ${PROCESS_THREADS}"
 
 # 启动 Gunicorn 并使用 tee 命令同时输出日志到文件和控制台
-exec gunicorn -w ${PROCESS_WORKERS} --threads ${PROCESS_THREADS} --bind 0.0.0.0:33333 main:app --access-logfile - --error-logfile -
+exec gunicorn -w ${PROCESS_WORKERS} --threads ${PROCESS_THREADS} --bind 0.0.0.0:33333 main:app --access-logfile - --error-logfile - --timeout 60
+