ai: Enable stream/real-time responses.

* Revert 68ce31f, and integrated with
gradio docs.

Ref: https://www.gradio.app/docs/gradio/chatbot

jarvis.py

@@ -112,7 +112,8 @@ def chat_with_model(history, user_input, selected_model_display, sess):
 def respond(multi_input, history, selected_model_display, sess):
     message = {"text": multi_input.get("text", "").strip(), "files": multi_input.get("files", [])}
     if not message["text"] and not message["files"]:
-        return history, gr.MultimodalTextbox(value=None, interactive=True), sess
+        yield history, gr.MultimodalTextbox(value=None, interactive=True), sess
+        return
     combined_input = ""
     for file_item in message["files"]:
         if isinstance(file_item, dict) and "name" in file_item:
@@ -125,8 +126,11 @@ def respond(multi_input, history, selected_model_display, sess):
         combined_input += message["text"]
     history.append([combined_input, ""])
     ai_response = chat_with_model(history, combined_input, selected_model_display, sess)
-    history[-1][1] = ai_response
-    return history, gr.MultimodalTextbox(value=None, interactive=True), sess
+    history[-1][1] = ""
+    for character in ai_response:
+        history[-1][1] += character
+        time.sleep(0.0005)
+        yield history, gr.MultimodalTextbox(value=None, interactive=True), sess
 
 def change_model(new_model_display):
     return [], create_session(), new_model_display