Update app.py

app.py

@@ -90,10 +90,10 @@ YOUTUBE_URL_2 = "https://www.youtube.com/watch?v=hdhZwyf24mE"
 ################################################
 #LLM Model mit dem gearbeitet wird
 #openai-------------------------------------
-#MODEL_NAME  = "gpt-3.5-turbo-16k"
+MODEL_NAME  = "gpt-3.5-turbo-16k"
 #MODEL_NAME = "gpt-3.5-turbo-1106"
 #MODEL_NAME= "gpt-4-1106-preview"
-MODEL_NAME = "gpt-4-vision-preview"
+MODEL_NAME_IMAGE = "gpt-4-vision-preview"
 
 
 #verfügbare Modelle anzeigen lassen
@@ -173,22 +173,33 @@ def process_image(image_path, prompt):
 
 
     # Prepare the data for the API request (specific to the API you're using)
-    data = {
-            "role": "user",
-            "content": [
-                        {
-                          "type": "text",
-                          "text": prompt
-                        },
-                        {
-                          "type": "image_url",
-                          "image_url": {
-                            "url": f"data:image/jpeg;base64,{encoded_string}"
-                           }
-                        }
-                       ]
-           }    
-    return data
+    headers = {
+              "Content-Type": "application/json",
+              "Authorization": f"Bearer {OAI_API_KEY}"
+            }
+
+    payload = {
+      "model": MODEL_NAME_IMAGE,
+      "messages": [
+        {
+          "role": "user",
+          "content": [
+            {
+              "type": "text",
+              "text": prompt
+            },
+            {
+              "type": "image_url",
+              "image_url": {
+                "url": f"data:image/jpeg;base64,{encoded_string}"
+              }
+            }
+          ]
+        }
+      ],
+      "max_tokens": 300
+    }   
+    return headers, payload
 
 
 def transfer_input(inputs):
@@ -412,7 +423,8 @@ def create_picture(history,  prompt):
 
 
 def generate_auswahl(prompt, file, chatbot, history, rag_option, model_option, openai_api_key, k=3, top_p=0.6, temperature=0.5, max_new_tokens=4048, max_context_length_tokens=2048, repetition_penalty=1.3,):
-     if (prompt.find('zeichnen') != -1):
+    #Bild nach Anweisung zeichnen und in History darstellen...
+    if (prompt.find('zeichnen') != -1):
         response = generate_bild(prompt)
         result = response.content 
         #Bild ausgeben
@@ -420,17 +432,24 @@ def generate_auswahl(prompt, file, chatbot, history, rag_option, model_option, o
         image_64 = umwandeln_fuer_anzeige(image)
         chatbot[-1][1] =  "<img src='data:image/png;base64,{0}'/>".format(b64encode(image_64).decode('utf-8')) 
         history = history + [(prompt, result)]  
-
         return chatbot, history, "Success"
      else:
-        result = generate_text(prompt, file, chatbot, history, rag_option, model_option, openai_api_key, k=3, top_p=0.6, temperature=0.5, max_new_tokens=4048, max_context_length_tokens=2048, repetition_penalty=1.3,)
-        #Antwort als Stream ausgeben... wenn Textantwort gefordert
-        chatbot[-1][1] = result
+        #kein Bild hochgeladen -> auf Text antworten...
         if (file == None):
+            result = generate_text(prompt, file, chatbot, history, rag_option, model_option, openai_api_key, k=3, top_p=0.6, temperature=0.5, max_new_tokens=4048, max_context_length_tokens=2048, repetition_penalty=1.3,)
             history = history + [(prompt, result)]
+            #Antwort als Stream ausgeben... wenn Textantwort gefordert
         else:
+            #Es wurde ein Bild angehängt -> wenn prompt dazu, das Bild analysieren
+            #geht nur über spezielle OpenAI-Schnittstelle...
             history = history + [((file,), None),(prompt, result)]
-
+            headers, payload = process_image
+            response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)
+            result = response.json()
+            history = history + [((file,), None),(prompt, result)]
+            
+        chatbot[-1][1] = result
+         
         return chatbot, history, "Success" 
         """ 
         for character in result: