Upload app.py

app.py

@@ -24,13 +24,6 @@ checker_pipe = pipeline('text-classification', 'ksg-dfci/TrialChecker', tokenize
 
 
 
-import gradio as gr
-import pandas as pd
-import torch
-import torch.nn.functional as F
-import tempfile
-
-
 import gradio as gr
 import pandas as pd
 import torch
@@ -43,7 +36,7 @@ import tempfile
 #
 # For example:
 # trial_spaces = pd.read_csv("some_file.csv")
-# embedding_model = SentenceTransformer("model-name")
+# embedding_model = SentenceTransformer("model-name", device="cuda")
 # trial_space_embeddings = torch.load("trial_space_embeddings.pt")
 # checker_pipe = pipeline(...)
 # etc.
@@ -51,7 +44,7 @@ import tempfile
 def match_clinical_trials_dropdown(patient_summary: str):
     """
     1) Runs the trial matching logic.
-    2) Returns a dict for the dropdown ({"choices": [...], "value": None}),
+    2) Returns a gr.update(...) for the dropdown (setting its choices),
        plus a DataFrame for further use.
     """
     # 1. Encode user input
@@ -104,34 +97,31 @@ def match_clinical_trials_dropdown(patient_summary: str):
         'trial_checker_score'
     ]]
 
-    # Build the dropdown choices, e.g. "NCT001 - Some Title"
+    # Build the dropdown choices, e.g., "NCT001 - Some Title"
     dropdown_options = []
     for _, row in out_df.iterrows():
         option_str = f"{row['nct_id']} - {row['trial_title']}"
         dropdown_options.append(option_str)
 
-    # Return a dict that sets "choices" and resets "value" to None
-    dropdown_dict = {
-        "choices": dropdown_options,
-        "value": None
-    }
+    # Return an update for the dropdown (choices + clear any initial value)
+    dropdown_update = gr.Dropdown.update(
+        choices=dropdown_options,
+        value=None  # ensures we don't default to any item not in the list
+    )
 
-    # Returning (dropdown_dict, out_df) means the first output updates the dropdown,
-    # the second output is stored in state.
-    return dropdown_dict, out_df
+    return dropdown_update, out_df
 
 def show_selected_trial(selected_option: str, df: pd.DataFrame):
     """
-    Given the selected dropdown option (e.g. "NCT001 - Some Title")
-    and the DataFrame, find the row and build a text summary.
+    1) Given the selected dropdown option, e.g. "NCT001 - Some Title"
+    2) Find the row in df and build a summary string.
     """
     if not selected_option:
         return ""
 
-    # Extract NCT ID from "NCT001 - Some Title"
+    # Parse NCT ID from "NCT001 - Some Title"
     nct_id = selected_option.split(" - ")[0].strip()
 
-    # Find the matching row
     row = df[df['nct_id'] == nct_id]
     if row.empty:
         return "No data found for the selected trial."
@@ -151,12 +141,13 @@ def show_selected_trial(selected_option: str, df: pd.DataFrame):
 
 def export_results(df: pd.DataFrame):
     """
-    Saves the DataFrame to a temporary CSV file so Gradio can provide a download link.
+    Saves the DataFrame to a temporary CSV file so Gradio can provide a downloadable link.
     """
     temp = tempfile.NamedTemporaryFile(delete=False, suffix=".csv")
     df.to_csv(temp.name, index=False)
     return temp.name
 
+# A little CSS for the input box
 custom_css = """
 #input_box textarea {
     width: 600px !important;
@@ -185,7 +176,7 @@ with gr.Blocks(css=custom_css) as demo:
     # We'll store the DataFrame in a State for CSV export + reference
     results_state = gr.State()
 
-    # Dropdown: no initial choices, no value
+    # Dropdown (initially empty)
     trial_dropdown = gr.Dropdown(
         label="Select a Trial",
         choices=[],
@@ -204,14 +195,14 @@ with gr.Blocks(css=custom_css) as demo:
     export_btn = gr.Button("Export Results")
     download_file = gr.File()
 
-    # 1) "Find Matches" => set dropdown's choices and store the DataFrame
+    # 1) "Find Matches" => updates the dropdown choices and the state
     submit_btn.click(
         fn=match_clinical_trials_dropdown,
         inputs=patient_summary_input,
         outputs=[trial_dropdown, results_state]
     )
 
-    # 2) On dropdown change => show details
+    # 2) Selecting from the dropdown => shows more info
     trial_dropdown.change(
         fn=show_selected_trial,
         inputs=[trial_dropdown, results_state],
@@ -225,7 +216,7 @@ with gr.Blocks(css=custom_css) as demo:
         outputs=download_file
     )
 
-    # Enable queue so "Processing..." shows if logic is slow
+    # Enable queue so "Processing..." is shown if logic is slow
     demo.queue()
 
 if __name__ == "__main__":