Update app.py

app.py

@@ -16,30 +16,34 @@ SD_API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-dif
 
 
 def query_zephyr(linkedin_text):
+    prompt = "Prepare a prompt for Stable Diffusion for the following LinkedIn post:"
+
+    # Attempting a simplified, possibly more correct structure
+    # Adjust this according to the actual requirements of the Zephyr model
+    payload = {
+        "inputs": f"{prompt}\n{linkedin_text}",
+    }
+
     headers = {
         "Authorization": f"Bearer {ZEPHYR_API_TOKEN}",
         "Content-Type": "application/json",
     }
 
-    # Assuming Zephyr supports a similar conversational structure
-    chat = [
-        {"role": "system", "content": "Prepare a prompt for Stable Diffusion for the following LinkedIn post:"},
-        {"role": "user", "content": linkedin_text},
-        # You can add more turns here if necessary
-    ]
-
-    payload = {
-        "inputs": chat,
-    }
-
     response = requests.post(ZEPHYR_API_URL, headers=headers, json=payload)
     if response.status_code == 200:
         return response.json()
     else:
-        print(f"Failed to query Zephyr model, status code: {response.status_code}")
-        print(response.text)  # Provides insight into what went wrong
-        return None
+        print(response.text)  # To get more insight into what went wrong
+        raise Exception(f"Failed to query Zephyr model, status code: {response.status_code}")
 
+# Replace this with the actual LinkedIn text you wish to process
+linkedin_text = "Your LinkedIn post content here."
+try:
+    zephyr_response = query_zephyr(linkedin_text)
+    print(zephyr_response)
+except Exception as e:
+    print(e)
+    
 
 def generate_image_from_prompt(prompt, negative_prompt, guidance_scale, width, height, num_inference_steps):
     headers = {"Authorization": f"Bearer {SD_API_TOKEN}"}