Update app.py

app.py

@@ -122,11 +122,43 @@ def extract_full_paper_with_labels(pdf_path, progress=None):
         "content": content
     }
 
-# NEW: Function to clear file-related inputs/outputs only.
+def process_pdf_file(pdf_file, api_key, repo_address):
+    if pdf_file is None:
+        return None, "No PDF file uploaded."
+    # Extract content from PDF.
+    # pdf_file can be a file-like object or a dict depending on how Gradio returns it.
+    file_path = pdf_file.name if hasattr(pdf_file, "name") else pdf_file['name']
+    result = extract_full_paper_with_labels(file_path)
+    
+    # Convert the result dictionary into a DataFrame and write it to a parquet file.
+    df = pd.DataFrame([result])
+    base = os.path.splitext(result['filename'])[0]
+    parquet_filename = f"{base}.parquet"
+    df.to_parquet(parquet_filename, index=False)
+    
+    repo_status = ""
+    # If API key and repo address are provided, attempt to upload the parquet file.
+    if api_key and repo_address:
+        api = HfApi()
+        try:
+            api.upload_file(
+                path_or_fileobj=parquet_filename,
+                path_in_repo=parquet_filename,
+                repo_id=repo_address,
+                token=api_key
+            )
+            repo_status = f"File uploaded to repo {repo_address} successfully."
+        except Exception as e:
+            repo_status = f"Failed to upload to repo: {str(e)}"
+    else:
+        repo_status = "API key or repo address not provided, skipping repo upload."
+    
+    # Return the parquet file for local download and the status message.
+    return parquet_filename, repo_status
+
+# Function to clear only file-related inputs/outputs, preserving the API key and repo address.
 def clear_files():
-    # Return empty values for file input and output display.
-    # Notice that we do NOT return anything for the API key or repo address.
-    return None, "", None
+    return None, None, ""
 
 # Gradio interface setup
 with gr.Blocks() as demo:
@@ -137,24 +169,20 @@ with gr.Blocks() as demo:
         pdf_file_input = gr.File(label="Upload PDF")
         convert_button = gr.Button("Convert to Parquet")
         clear_button = gr.Button("Clear Files")
-    output_display = gr.Textbox(label="Output")
-    # (Optional) A hidden textbox for parquet data, if used later.
-    parquet_output = gr.Textbox(label="Parquet Data", visible=False)
-
+    with gr.Row():
+        download_file_output = gr.File(label="Download Parquet File")
+        repo_status_output = gr.Textbox(label="Repo Upload Status")
+    
     convert_button.click(
-        extract_full_paper_with_labels,
-        inputs=pdf_file_input,
-        outputs=output_display
+        process_pdf_file,
+        inputs=[pdf_file_input, api_key_input, repo_address_input],
+        outputs=[download_file_output, repo_status_output]
     )
-    # The clear button now only clears file-related components;
-    # API key and Repo Address remain untouched.
+    # The clear button now only clears file-related components; API key and Repo Address remain unchanged.
     clear_button.click(
         clear_files,
         inputs=None,
-        outputs=[pdf_file_input, output_display, parquet_output]
+        outputs=[pdf_file_input, download_file_output, repo_status_output]
     )
 
 demo.launch()
-
-
-