Update app.py

app.py

@@ -292,48 +292,63 @@ def bot(history, temperature, top_p, max_output_tokens):
     text_this_term = ""
     
     num_new_images = 0
-    previous_image = False
+    # previous_image = False
     for i, message in enumerate(history[:-1]):
         if type(message[0]) is tuple:
-            if previous_image:
-                gr.Warning("Only one image can be uploaded in a conversation. Please reduce the number of images and start a new conversation.")
-                our_chatbot.conversation = conv_templates[our_chatbot.conv_mode].copy()
-                return None
+            # if previous_image:
+            #     gr.Warning("Only one image can be uploaded in a conversation. Please reduce the number of images and start a new conversation.")
+            #     our_chatbot.conversation = conv_templates[our_chatbot.conv_mode].copy()
+            #     return None
            
             images_this_term.append(message[0][0])
             if is_valid_video_filename(message[0][0]):
-                raise ValueError("Video is not supported")
+                # raise ValueError("Video is not supported")
                 num_new_images += our_chatbot.num_frames
             elif is_valid_image_filename(message[0][0]):
                 print("#### Load image from local file",message[0][0])
                 num_new_images += 1
             else:
-                raise ValueError("Invalid image file")
-            previous_image = True
+                raise ValueError("Invalid file format")
+            # previous_image = True
         else:
             num_new_images = 0
-            previous_image = False
+            # previous_image = False
 
     all_image_hash = []
     all_image_path = []
-    for image_path in images_this_term:
-        with open(image_path, "rb") as image_file:
-            image_data = image_file.read()
-            image_hash = hashlib.md5(image_data).hexdigest()
-            all_image_hash.append(image_hash)
-            image = PIL.Image.open(image_path).convert("RGB")
+    for file_path in images_this_term:
+        with open(file_path, "rb") as file:
+            file_data = file.read()
+            file_hash = hashlib.md5(file_data).hexdigest()
+            all_file_hash.append(file_hash)
+            
             t = datetime.datetime.now()
-            filename = os.path.join(
+            output_dir = os.path.join(
                 LOGDIR,
-                "serve_images",
-                f"{t.year}-{t.month:02d}-{t.day:02d}",
-                f"{image_hash}.jpg",
+                "serve_files",
+                f"{t.year}-{t.month:02d}-{t.day:02d}"
             )
-            all_image_path.append(filename)
-            if not os.path.isfile(filename):
-                os.makedirs(os.path.dirname(filename), exist_ok=True)
-                print("image save to",filename)
-                image.save(filename)
+            os.makedirs(output_dir, exist_ok=True)
+            
+            if is_valid_image_filename(file_path):
+                # Process and save images
+                image = Image.open(file_path).convert("RGB")
+                filename = os.path.join(output_dir, f"{file_hash}.jpg")
+                all_file_path.append(filename)
+                if not os.path.isfile(filename):
+                    print("Image saved to", filename)
+                    image.save(filename)
+    
+            elif is_valid_video_filename(file_path):
+                # Simplified video saving
+                filename = os.path.join(output_dir, f"{file_hash}.mp4")
+                all_file_path.append(filename)
+                if not os.path.isfile(filename):
+                    print("Video saved to", filename)
+                    os.makedirs(os.path.dirname(filename), exist_ok=True)
+                    # Directly copy the video file
+                    with open(file_path, "rb") as src, open(filename, "wb") as dst:
+                        dst.write(src.read())
     
     image_list = []
     for f in images_this_term:
@@ -493,7 +508,7 @@ with gr.Blocks(
 
         chat_input = gr.MultimodalTextbox(
             interactive=True,
-            file_types=["image"],
+            file_types=["image", "video"],
             placeholder="Enter message or upload file...",
             show_label=False,
             submit_btn="🚀"
@@ -571,18 +586,25 @@ with gr.Blocks(
                             "text": "Based on the five species in the images, draw a food chain. Explain the role of each species in the food chain.",
                         }
                     ],
-                    [
-                        {
-                            "files": [
-                                f"{cur_dir}/examples/realcase_video.mp4",
-                            ],
-                            "text": "Please describe the video in detail.",
-                        },
-                    ]
                 ],
                 inputs=[chat_input],
-                label="RealWorld Cases",
+                label="Real World Image Cases",
             )
+        gr.Examples(
+            label="Video",
+            examples=[
+                [
+                    {
+                        "files": [
+                            f"{cur_dir}/examples/realcase_video.mp4",
+                        ],
+                        "text": "Please describe the video in detail.",
+                    },
+                ]
+            ]
+            inputs=[chat_input],
+            label="Real World Video Case"
+        )
 
         gr.Markdown(tos_markdown)
         gr.Markdown(learn_more_markdown)