Add application file

app.py

@@ -4,6 +4,7 @@ import streamlit as st
 from huggingface_hub import hf_hub_download
 import pandas as pd
 import tempfile
+import re
 
 HF_REPO_ID = "stcoats/temp-duckdb-upload"
 HF_FILENAME = "ycsep.duckdb"
@@ -19,32 +20,40 @@ if not os.path.exists(LOCAL_PATH):
         repo_id=HF_REPO_ID,
         repo_type="dataset",
         filename=HF_FILENAME,
-        local_dir="."
+        local_dir=".",
+        local_dir_use_symlinks=False
     )
     st.success("Download complete.")
 
-# Connect
+# Connect (only once)
+@st.cache_resource(show_spinner=False)
+def get_duckdb_connection():
+    return duckdb.connect(LOCAL_PATH, read_only=True)
+
 try:
-    con = duckdb.connect(LOCAL_PATH, read_only=True)
+    con = get_duckdb_connection()
     st.success("Connected to DuckDB.")
 except Exception as e:
     st.error(f"DuckDB connection failed: {e}")
     st.stop()
 
 # Search
-query = st.text_input("Search text (case-insensitive)", "")
-query = query.strip().lower()
+query = st.text_input("Search text (case-insensitive)", "").strip()
 
 if query:
     sql = """
         SELECT channel, file, speaker, start_time, end_time, text, pos_tags, audio
         FROM data
-        WHERE LOWER(text) LIKE ?
+        WHERE LOWER(text) LIKE LOWER(?)
         LIMIT 100
     """
     df = con.execute(sql, [f"%{query}%"]).df()
 else:
-    df = con.execute("SELECT channel, file, speaker, start_time, end_time, text, pos_tags, audio FROM data LIMIT 100").df()
+    df = con.execute("""
+        SELECT channel, file, speaker, start_time, end_time, text, pos_tags, audio 
+        FROM data 
+        LIMIT 100
+    """).df()
 
 st.markdown(f"### Showing {len(df)} results")
 
@@ -53,13 +62,19 @@ if len(df) == 0:
 
 # Show table with inline audio players
 for i, row in df.iterrows():
-    col1, col2, col3 = st.columns([2, 6, 2])
-    col1.markdown(f"**Channel:** {row['channel']}  ")
-    col1.markdown(f"**File:** {row['file']}  ")
-    col1.markdown(f"**Speaker:** {row['speaker']}  ")
-    col1.markdown(f"**Start:** {row['start_time']}  ")
-    col1.markdown(f"**End:** {row['end_time']}  ")
-    col2.markdown(f"**Text:** {row['text']}")
+    col1, col2, col3 = st.columns([3, 5, 2])
+
+    col1.markdown(f"**Channel:** {row['channel']}")
+    col1.markdown(f"**File:** {row['file']}")
+    col1.markdown(f"**Speaker:** {row['speaker']}")
+    col1.markdown(f"**Start Time:** {row['start_time']}")
+    col1.markdown(f"**End Time:** {row['end_time']}")
+
+    highlighted_text = row['text']
+    if query:
+        highlighted_text = re.sub(f'({re.escape(query)})', r'<mark>\1</mark>', highlighted_text, flags=re.IGNORECASE)
+
+    col2.markdown(f"**Text:** {highlighted_text}", unsafe_allow_html=True)
     col2.markdown(f"**POS tags:** {row['pos_tags']}")
 
     audio_data = row["audio"]