Whisper support (#50)

* whisper model support correctly * Improve log file conversion performance * fix pytest
KenyonY · Jul 18, 2023 · e887bc6 · e887bc6
1 parent 69fd879
commit e887bc6
Show file tree

Hide file tree

Showing 14 changed files with 137 additions and 110 deletions.
diff --git a/.github/data/whisper.m4a b/.github/data/whisper.m4a
diff --git a/Examples/chat.py b/Examples/chat.py
@@ -0,0 +1,12 @@
+import openai
+
+openai.api_base = "https://api.openai-forward.com/v1"
+openai.api_key = "sk-******"
+
+resp = openai.ChatCompletion.create(
+    model="gpt-3.5-turbo",
+    messages=[
+        {"role": "user", "content": "Who won the world series in 2020?"},
+    ],
+)
+print(resp.choices)
diff --git a/Examples/embedding.py b/Examples/embedding.py
@@ -0,0 +1,9 @@
+import openai
+
+openai.api_base = "http://localhost:8000/v1"
+openai.api_key = "sk-******"
+response = openai.Embedding.create(
+    input="Your text string goes here", model="text-embedding-ada-002"
+)
+embeddings = response['data'][0]['embedding']
+print(embeddings)
diff --git a/Examples/whisper.py b/Examples/whisper.py
@@ -0,0 +1,10 @@
+# Note: you need to be using OpenAI Python v0.27.0 for the code below to work
+import openai
+from sparrow import relp
+
+openai.api_base = "https://api.openai-forward.com/v1"
+openai.api_key = "sk-******"
+
+audio_file = open(relp("../.github/data/whisper.m4a"), "rb")
+transcript = openai.Audio.transcribe("whisper-1", audio_file)
+print(transcript)
diff --git a/README.md b/README.md
@@ -61,16 +61,6 @@
 > https://render.openai-forward.com  
 > https://railway.openai-forward.com
 
-
-<details >
-<summary> 👉Tips  </summary>  
-
-🎉🎉🎉近期GPT-4 API 已经全面可用! 但它需要付费api账户，也就是需要先绑定信用卡。  
-目前比较推荐开源加密钱包[OneKey](https://github.com/OneKeyHQ)的VISA虚拟卡：[https://card.onekey.so](https://card.onekey.so/?i=O163GB) 
-
-</details>
-
-
 ## 功能
 
 **基础功能**
@@ -95,27 +85,26 @@
 
 👉 [部署文档](deploy.md)
 
-
 提供以下几种部署方式  
 **有海外vps方案**
 
-1. [pip 安装部署](deploy.md#pip部署) 
-2. [Docker部署](deploy.md#docker部署) 
+1. [pip 安装部署](deploy.md#pip部署)
+2. [Docker部署](deploy.md#docker部署)
    > https://api.openai-forward.com
 
 **无vps免费部署方案**
+
 1. [Railway部署](deploy.md#Railway-一键部署)
    > https://railway.openai-forward.com
 2. [Render一键部署](deploy.md#render-一键部署)
    > https://render.openai-forward.com
 
-
 ---
 下面的部署仅提供单一转发功能
 
 3. [一键Vercel部署](deploy.md#vercel-一键部署)
    > https://vercel.openai-forward.com
-4. [cloudflare部署](deploy.md#cloudflare-部署) 
+4. [cloudflare部署](deploy.md#cloudflare-部署)
    > https://cloudflare.page.openai-forward.com
 
 ## 应用
@@ -200,6 +189,7 @@ curl --location 'https://api.openai-forward.com/v1/images/generations' \
 另一种为读取环境变量的方式指定。
 
 ### 命令行参数
+
 可通过 `openai-forward run --help` 查看
 
 <details open>
@@ -217,10 +207,10 @@ curl --location 'https://api.openai-forward.com/v1/images/generations' \
 | --route_prefix | 同 ROUTE_PREFIX    |          `None`          |
 | --log_chat | 同 LOG_CHAT        |        `False`         |
 
-
 </details>
 
 ### 环境变量配置项
+
 支持从运行目录下的`.env`文件中读取
 
 | 环境变量            | 说明                                                                                                                                |           默认值            |
@@ -231,7 +221,6 @@ curl --location 'https://api.openai-forward.com/v1/images/generations' \
 | ROUTE_PREFIX    | 路由前缀                                                                                                                              |            无             |
 | LOG_CHAT        | 是否记录聊天内容                                                                                                                          |         `false`          |
 
-
 ## 高级配置
 
 **设置openai api_key为自定义的forward key**  
@@ -245,8 +234,8 @@ FORWARD_KEY=fk-****** # 这里fk-token由我们自己定义
 ```
 
 这里我们配置了FORWARD_KEY为`fk-******`, 那么后面客户端在调用时只需设置OPENAI_API_KEY为我们自定义的`fk-******` 即可。  
-这样的好处是在使用一些需要输入OPENAI_API_KEY的第三方应用时，我们可以使用自定义的api-key`fk-******`, 
-无需担心真正的OPENAI_API_KEY被泄露。并且可以对外分发`fk-******`。  
+这样的好处是在使用一些需要输入OPENAI_API_KEY的第三方应用时，我们可以使用自定义的api-key`fk-******`,
+无需担心真正的OPENAI_API_KEY被泄露。并且可以对外分发`fk-******`。
 
 **用例:**
 

diff --git a/openai_forward/__init__.py b/openai_forward/__init__.py
@@ -1,4 +1,4 @@
-__version__ = "0.2.4"
+__version__ = "0.3.0-alpha"
 
 from dotenv import load_dotenv
 

diff --git a/openai_forward/__main__.py b/openai_forward/__main__.py
@@ -60,12 +60,12 @@ def run(
         )
 
     @staticmethod
-    def convert(log_path: str = "./Log/chat.log", target_path: str = "./Log/chat.json"):
-        """Convert log file  to jsonl file"""
-        from openai_forward.tool import convert_chatlog_to_jsonl
+    def convert(log_folder: str = "./Log/chat", target_path: str = "./Log/chat.json"):
+        """Convert log folder to jsonl file"""
+        from openai_forward.tool import convert_folder_to_jsonl
 
-        print(f"Convert {log_path} to {target_path}")
-        convert_chatlog_to_jsonl(log_path, target_path)
+        print(f"Convert {log_folder}/*.log to {target_path}")
+        convert_folder_to_jsonl(log_folder, target_path)
 
 
 def main():

diff --git a/openai_forward/base.py b/openai_forward/base.py
@@ -9,6 +9,7 @@
 
 from .config import print_startup_info, setting_log
 from .content.chat import ChatSaver
+from .content.whisper import WhisperSaver
 from .tool import env2list
 
 
@@ -36,6 +37,7 @@ class OpenaiBase:
     if _LOG_CHAT:
         setting_log(save_file=False)
         chatsaver = ChatSaver()
+        whispersaver = WhisperSaver()
 
     def validate_request_host(self, ip):
         if self.IP_WHITELIST and ip not in self.IP_WHITELIST:
@@ -56,10 +58,13 @@ async def aiter_bytes(cls, r: httpx.Response, route_path: str, uid: str):
             bytes_ += chunk
             yield chunk
         try:
-            target_info = cls.chatsaver.parse_bytes_to_content(bytes_, route_path)
-            cls.chatsaver.add_chat(
-                {target_info["role"]: target_info["content"], "uid": uid}
-            )
+            if route_path == "/v1/chat/completions":
+                target_info = cls.chatsaver.parse_bytes_to_content(bytes_, route_path)
+                cls.chatsaver.add_chat(
+                    {target_info["role"]: target_info["content"], "uid": uid}
+                )
+            elif route_path.startswith("/v1/audio/"):
+                cls.whispersaver.add_log(bytes_)
         except Exception as e:
             logger.debug(f"log chat (not) error:\n{e=}")
 
@@ -71,23 +76,27 @@ async def _reverse_proxy(cls, request: Request):
         url = httpx.URL(path=url_path, query=request.url.query.encode("utf-8"))
         headers = dict(request.headers)
         auth = headers.pop("authorization", "")
-        auth_headers_dict = {"Content-Type": "application/json", "Authorization": auth}
+        content_type = headers.pop("content-type", "application/json")
+        auth_headers_dict = {"Content-Type": content_type, "Authorization": auth}
         auth_prefix = "Bearer "
         if cls._no_auth_mode or auth and auth[len(auth_prefix) :] in cls._FWD_KEYS:
             auth = auth_prefix + next(cls._cycle_api_key)
             auth_headers_dict["Authorization"] = auth
 
-        log_chat_completions = False
+        if_log = False
         uid = None
         if cls._LOG_CHAT and request.method == "POST":
             try:
-                chat_info = await cls.chatsaver.parse_payload_to_content(
-                    request, route_path=url_path
-                )
-                if chat_info:
-                    cls.chatsaver.add_chat(chat_info)
-                    uid = chat_info.get("uid")
-                    log_chat_completions = True
+                if url_path.startswith("/v1/audio/"):
+                    if_log = True
+                else:
+                    chat_info = await cls.chatsaver.parse_payload_to_content(
+                        request, route_path=url_path
+                    )
+                    if chat_info:
+                        cls.chatsaver.add_chat(chat_info)
+                        uid = chat_info.get("uid")
+                        if_log = True
             except Exception as e:
                 logger.debug(
                     f"log chat error:\n{request.client.host=} {request.method=}: {e}"
@@ -117,11 +126,7 @@ async def _reverse_proxy(cls, request: Request):
                 status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=e
             )
 
-        aiter_bytes = (
-            cls.aiter_bytes(r, url_path, uid)
-            if log_chat_completions
-            else r.aiter_bytes()
-        )
+        aiter_bytes = cls.aiter_bytes(r, url_path, uid) if if_log else r.aiter_bytes()
         return StreamingResponse(
             aiter_bytes,
             status_code=r.status_code,

diff --git a/openai_forward/config.py b/openai_forward/config.py
@@ -69,12 +69,19 @@ def setting_log(save_file=False, log_name="openai_forward", multi_process=True):
     config_handlers = [
         {"sink": sys.stdout, "level": "DEBUG"},
         {
-            "sink": f"./Log/chat.log",
+            "sink": f"./Log/chat/chat.log",
             "enqueue": multi_process,
-            "rotation": "20 MB",
+            "rotation": "50 MB",
             "filter": lambda record: "chat" in record["extra"],
             "format": "{message}",
         },
+        {
+            "sink": f"./Log/whisper/whisper.log",
+            "enqueue": multi_process,
+            "rotation": "30 MB",
+            "filter": lambda record: "whisper" in record["extra"],
+            "format": "{message}",
+        },
     ]
     if save_file:
         config_handlers += [

diff --git a/openai_forward/content/image.py b/openai_forward/content/image.py
diff --git a/openai_forward/content/whisper.py b/openai_forward/content/whisper.py
@@ -0,0 +1,10 @@
+from loguru import logger
+
+
+class WhisperSaver:
+    def __init__(self):
+        self.logger = logger.bind(whisper=True)
+
+    def add_log(self, bytes_: bytes):
+        text_content = bytes_.decode("utf-8")
+        self.logger.debug(text_content)