feat: text_and_image blockbox

src/blackbox/blackbox_factory.py

@@ -10,6 +10,7 @@ from .text_to_audio import TextToAudio
 from .tesou import Tesou
 from .fastchat import Fastchat
 from .g2e import G2E
+from .text_and_image import TextAndImage
 
 class BlackboxFactory:
 
@@ -25,6 +26,7 @@ class BlackboxFactory:
         self.fastchat = Fastchat()
         self.audio_chat = AudioChat(self.asr, self.tesou, self.tts)
         self.g2e = G2E()
+        self.text_and_image = TextAndImage()
 
     def __call__(self, *args, **kwargs):
         return self.processing(*args, **kwargs)
@@ -52,4 +54,6 @@ class BlackboxFactory:
             return self.audio_chat
         if blackbox_name == "g2e":
             return self.g2e
+        if blackbox_name == 'text_and_image':
+            return self.text_and_image
         raise ValueError("Invalid blockbox type")

src/blackbox/g2e.py

@@ -22,7 +22,7 @@ class G2E(Blackbox):
         if context == None:
             context = []
         url = 'http://120.196.116.194:48890/v1'
-
+        #url = 'http://120.196.116.194:48892/v1'
 
         background_prompt = '''KOMBUKIKI是一款茶饮料，目标受众 年龄：20-35岁 性别：女性 地点：一线城市、二线城市 职业：精英中产、都市白领 收入水平：中高收入，有一定消费能力 兴趣和爱好：注重健康，有运动习惯
 
@@ -49,6 +49,9 @@ class G2E(Blackbox):
         prompt_template = [
            {"role": "system", "content": background_prompt + prompt1},
         ]
+        #prompt_template = [
+        #    {"role": "system", "content": ''},
+        #]
 
 
         messages = prompt_template + context + [
@@ -61,8 +64,8 @@ class G2E(Blackbox):
             api_key='YOUR_API_KEY',
             base_url=url
         )
-
         model_name = client.models.list().data[0].id
+        print(model_name)
         response = client.chat.completions.create(
         model=model_name,
         messages=messages,

src/blackbox/text_and_image.py

@@ -0,0 +1,28 @@
+from typing import Any, Coroutine
+
+from fastapi import Request, Response, status
+from fastapi.responses import JSONResponse
+from .blackbox import Blackbox
+
+
+class TextAndImage(Blackbox):
+    
+    def __call__(self, *args, **kwargs):
+        return self.processing(*args, **kwargs)
+
+    def processing(self, text: str, image: bytes):
+        # TODO: implement the processing logic
+        return "text and image processing..."
+    
+    def valid(self, *args, **kwargs) -> bool:
+        return True
+    
+    async def fast_api_handler(self, request: Request) -> Response:
+        # If the request is a form data, you can use the following code
+        data = (await request.form()).get("image")
+        if data is None:
+            return JSONResponse(content={"error": "image is required"}, status_code=status.HTTP_400_BAD_REQUEST)
+        imageBytes = await data.read()
+        text = (await request.form()).get("text")
+        result = self.processing(text, imageBytes)
+        return JSONResponse(content={"result": result }, status_code=status.HTTP_200_OK)