Add application file

app.py

@@ -2,37 +2,77 @@ import os
 import duckdb
 import streamlit as st
 from huggingface_hub import hf_hub_download
+import pandas as pd
+import tempfile
 
 HF_REPO_ID = "stcoats/temp-duckdb-upload"
 HF_FILENAME = "ycsep.duckdb"
 LOCAL_PATH = "./ycsep.duckdb"
 
-st.title("YCSEP Audio Debug")
+st.set_page_config(layout="wide")
+st.title("YCSEP Audio Dataset Viewer")
 
+# Download database if missing
 if not os.path.exists(LOCAL_PATH):
-    st.write("Downloading DB...")
+    st.write("Downloading from HF Hub...")
     hf_hub_download(
         repo_id=HF_REPO_ID,
         repo_type="dataset",
         filename=HF_FILENAME,
         local_dir="."
     )
-    st.success("Downloaded.")
-
-con = duckdb.connect(LOCAL_PATH, read_only=True)
-
-st.markdown("### Raw Preview")
-df = con.execute("SELECT speaker, text, audio FROM data LIMIT 5").fetchall()
-
-for row in df:
-    speaker, text, audio = row
-    st.write(f"Speaker: {speaker}")
-    st.write(f"Text: {text}")
-    st.write(f"AUDIO TYPE: {type(audio)}")
-    st.write(f"AUDIO LENGTH: {len(audio) if hasattr(audio, '__len__') else 'n/a'}")
-    if isinstance(audio, (bytes, bytearray, memoryview)):
-        st.audio(audio, format="audio/mp3")
-    else:
-        st.warning("Audio not valid binary")
-    st.markdown("---")
+    st.success("Download complete.")
+
+# Connect
+try:
+    con = duckdb.connect(LOCAL_PATH, read_only=True)
+    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 or speaker", "")
+query = query.strip().lower()
+
+if query:
+    sql = """
+        SELECT speaker, text, audio
+        FROM data
+        WHERE LOWER(CAST(speaker AS VARCHAR)) LIKE ? OR LOWER(CAST(text AS VARCHAR)) LIKE ?
+        LIMIT 100
+    """
+    df = con.execute(sql, [f"%{query}%", f"%{query}%"]).df()
+else:
+    df = con.execute("SELECT speaker, text, audio FROM data LIMIT 100").df()
+
+st.markdown(f"### Showing {len(df)} results")
+
+if len(df) == 0:
+    st.warning("No matches found.")
+
+# Show table with inline audio players
+for i, row in df.iterrows():
+    col1, col2, col3 = st.columns([2, 5, 3])
+    col1.markdown(f"**{row['speaker']}**")
+    col2.markdown(row['text'])
+
+    audio_data = row["audio"]
+    try:
+        if isinstance(audio_data, (bytes, bytearray, memoryview)):
+            audio_bytes = bytes(audio_data)
+        elif isinstance(audio_data, list):  # DuckDB sometimes gives list[int]
+            audio_bytes = bytes(audio_data)
+        else:
+            audio_bytes = None
+
+        if audio_bytes:
+            with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmpfile:
+                tmpfile.write(audio_bytes)
+                tmpfile.flush()
+                col3.audio(tmpfile.name, format="audio/mp3")
+        else:
+            col3.warning("Audio missing or invalid format.")
+    except Exception as e:
+        col3.error(f"Audio error: {e}")