Update app.py

app.py

@@ -33,16 +33,33 @@ def ovis_chat(message, history):
     response = ""
     text_input = message["text"]
     for msg in history:
-        if msg["role"] == "user" and "size" not in msg.keys():
+        # case history entry pair only has text
+        if isinstance(msg[0],str):
             conversations.append({
                 "from": "human",
-                "value": msg["content"]
+                "value": msg[0]
             })
-        elif msg["role"] == "assistant":
             conversations.append({
                 "from": "gpt",
-                "value": msg["content"]
+                "value": msg[1]
             })
+        # case history pair has an image
+        elif isinstance(msg[0],tuple): 
+            # case user did not pass a new image
+            # we override the none with the new image
+            if image_input is None : 
+                # if someone uploads a file and not an image this will break
+                image_input = Image.open(msg[0][0]).convert("RGB")
+            # we always process the text
+            conversations.append({
+                "from": "human",
+                "value": msg[1][0]
+            })
+            conversations.append({
+                "from": "gpt",
+                "value": msg[1][1]
+            })
+
     text_input = text_input.replace(image_placeholder, '')
     conversations.append({
         "from": "human",
@@ -133,5 +150,5 @@ latex_delimiters_set = [{
 #     submit_event = text_input.submit(submit_chat, [chatbot, text_input], [chatbot, text_input]).then(ovis_chat,[chatbot, image_input],chatbot)
 #     clear_btn.click(clear_chat, outputs=[chatbot, image_input, text_input])
 
-demo = gr.ChatInterface(fn=ovis_chat, type="messages", textbox=gr.MultimodalTextbox(),multimodal=True)
-demo.launch(debug=True)
+demo = gr.ChatInterface(fn=ovis_chat, textbox=gr.MultimodalTextbox(),multimodal=True)
+demo.launch(debug=True)