Adding cached examples, need to find out how to change the order.

app.py

@@ -28,7 +28,7 @@ tokenizer = AutoTokenizer.from_pretrained('derek-thomas/jais-13b-chat-hf')
 # Examples
 examples = ['من كان طرفي معركة اكتيوم البحرية؟',
             'لم السماء زرقاء؟',
-            "من فاز بكأس العالم للرجال في عام 2014؟",]
+            "من فاز بكأس العالم للرجال في عام 2014؟", ]
 
 
 def add_text(history, text):
@@ -45,7 +45,8 @@ def bot(history, hyde=False):
     # Retrieve documents relevant to query
     document_start = perf_counter()
     if hyde:
-        hyde_document = generate(f"Write a wikipedia article intro paragraph to answer this query: {query}").split('### Response: [|AI|]')[-1]
+        hyde_document = generate(f"Write a wikipedia article intro paragraph to answer this query: {query}").split(
+            '### Response: [|AI|]')[-1]
 
         logger.warning(hyde_document)
         documents = retriever(hyde_document, top_k=top_k)
@@ -76,6 +77,7 @@ def bot(history, hyde=False):
     history[-1][1] = response.split('### Response: [|AI|]')[-1]
     return history, prompt_html
 
+
 intro_md = """
 # Arabic RAG
 This is a project to demonstrate Retreiver Augmented Generation (RAG) in Arabic and English. It uses 
@@ -91,6 +93,14 @@ I'm using Inference Endpoint's Scale to Zero to save money on GPUs. If the staus
 chat to wake it up. You will get a `500 error` and it will take ~7 min to wake up.
 """
 
+
+def process_example(text, history=[]):
+    history = history + [[text, None]]
+    return bot(history)
+
+
+# hyde_prompt_html = gr.HTML()
+
 with gr.Blocks() as demo:
     gr.Markdown(intro_md)
     endpoint_status = gr.Textbox(check_endpoint_status, label="Inference Endpoint Status", every=1)
@@ -114,8 +124,15 @@ with gr.Blocks() as demo:
                     )
             txt_btn = gr.Button(value="Submit text", scale=1)
 
-        gr.Examples(examples, txt)
+        # gr.Examples(examples, txt)
         prompt_html = gr.HTML()
+        gr.Examples(
+                examples=examples,
+                inputs=txt,
+                outputs=[chatbot, prompt_html],
+                fn=process_example,
+                cache_examples=True, )
+        # prompt_html.render()
         # Turn off interactivity while generating if you click
         txt_msg = txt_btn.click(add_text, [chatbot, txt], [chatbot, txt], queue=False).then(
                 bot, chatbot, [chatbot, prompt_html])
@@ -150,8 +167,14 @@ with gr.Blocks() as demo:
                     )
             hyde_txt_btn = gr.Button(value="Submit text", scale=1)
 
-        gr.Examples(examples, hyde_txt)
         hyde_prompt_html = gr.HTML()
+        gr.Examples(
+                examples=examples,
+                inputs=hyde_txt,
+                outputs=[hyde_chatbot, hyde_prompt_html],
+                fn=process_example,
+                cache_examples=True, )
+        # prompt_html.render()
         # Turn off interactivity while generating if you click
         hyde_txt_msg = hyde_txt_btn.click(add_text, [hyde_chatbot, hyde_txt], [hyde_chatbot, hyde_txt],
                                           queue=False).then(