Upon loading the page, start the indexing of the default PDF file. Add a clear button.

app.py

@@ -74,6 +74,11 @@ def generate_response(question: str) -> list[str, str, str]:
     return responses, references, snippets
 
 
+def on_load_build():
+    pdf_files = ["OECD_Dividend_tax_fraud_2023-en.pdf",]
+    return build_rag_model(pdf_files)
+
+
 with gr.Blocks() as demo:
     gr.Markdown("""
         # Retrieval (ColBERT) + Generation (DSPy & Mistral)
@@ -97,7 +102,7 @@ with gr.Blocks() as demo:
     # Question to answer
     question = gr.Textbox(
         label="Question",
-        placeholder="What is dividend stripping?"
+        placeholder=""
     )
     response = gr.Textbox(
         label="Response",
@@ -108,7 +113,9 @@ with gr.Blocks() as demo:
         snippets = gr.HTML(label="Snippets")
     
     # button
-    response_button = gr.Button("Submit", variant='primary')
+    with gr.Row():
+        response_button = gr.Button("Submit", variant='primary')
+        clear_question_button = gr.Button("Clear", variant='secondary')
     
     # Example questions given default provided PDF file
     with gr.Accordion("Sample questions", open=False):
@@ -154,6 +161,14 @@ with gr.Blocks() as demo:
         inputs=[question],
         outputs=[response, references, snippets]
     )
+    clear_question_button.click(
+        fn=lambda: ('', '', '', ''),
+        inputs=[],
+        outputs=[question, response, references, snippets]
+    )
+
+    # Upon loading, index a default PDF file
+    demo.load(on_load_build, outputs=[build_status])
 
 
 demo.launch(show_api=False)