Update app.py

app.py

@@ -9,10 +9,16 @@ def query(payload):
     return response.json()
 
 def main():
-    st.title("Audio Upload and Prediction")
+    st.title("Audio and Image Upload for Prediction")
 
     # File uploader for audio files
     audio_file = st.file_uploader("Upload an audio file (e.g., mp3, wav, ogg, webm)", type=["mp3", "wav", "ogg", "webm", "m4a", "aac"])
+    
+    # File uploader for image files
+    image_file = st.file_uploader("Upload an image file (e.g., png, jpg, jpeg)", type=["png", "jpg", "jpeg"])
+
+    # Create payload for the query
+    payload = {}
 
     if audio_file is not None:
         # Convert the audio file to a Base64 string
@@ -20,25 +26,45 @@ def main():
         audio_base64 = base64.b64encode(audio_bytes).decode('utf-8')
         audio_mime_type = audio_file.type  # Get the mime type
 
-        # Prepare the payload
-        payload = {
-            "uploads": [
-                {
-                    "data": f'data:{audio_mime_type};base64,{audio_base64}',
-                    "type": 'audio',
-                    "name": audio_file.name,
-                    "mime": audio_mime_type
-                }
-            ],
-            "content": "This is the audio file being processed."  # Modify as needed
-        }
-
-        # Send the request to the API and get the output
-        if st.button("Submit"):
+        # Add audio data to the payload
+        payload["uploads"] = [
+            {
+                "data": f'data:{audio_mime_type};base64,{audio_base64}',
+                "type": 'audio',
+                "name": audio_file.name,
+                "mime": audio_mime_type
+            }
+        ]
+
+    if image_file is not None:
+        # Convert the image file to a Base64 string
+        image_bytes = image_file.read()
+        image_base64 = base64.b64encode(image_bytes).decode('utf-8')
+        image_mime_type = image_file.type  # Get the mime type
+
+        # Add image data to the payload
+        payload.setdefault("uploads", []).append(
+            {
+                "data": f'data:{image_mime_type};base64,{image_base64}',
+                "type": 'file',
+                "name": image_file.name,
+                "mime": image_mime_type
+            }
+        )
+    
+    # Add a question to the payload if needed
+    if st.text_input("Enter your question", "Can you describe the image?"):
+        payload["question"] = st.text_input("Enter your question", "Can you describe the image?")
+
+    # Send the request to the API and get the output
+    if st.button("Submit"):
+        if payload:
             with st.spinner("Sending request..."):
                 output = query(payload)
                 st.success("Request sent successfully!")
                 st.json(output)  # Display the API response
+        else:
+            st.error("Please upload at least one audio or image file.")
 
 if __name__ == "__main__":
     main()