app.py

import os
import duckdb
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"
LOCAL_PATH = "./ycsep.duckdb"

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 from HF Hub...")
    hf_hub_download(
        repo_id=HF_REPO_ID,
        repo_type="dataset",
        filename=HF_FILENAME,
        local_dir=".",
        local_dir_use_symlinks=False
    )
    st.success("Download complete.")

# Connect (only once)
@st.cache_resource(show_spinner=False)
def get_duckdb_connection():
    return duckdb.connect(LOCAL_PATH, read_only=True)

try:
    con = get_duckdb_connection()
    st.success("Connected to DuckDB.")
except Exception as e:
    st.error(f"DuckDB connection failed: {e}")
    st.stop()

# Search input
query = st.text_input("Search text (case-insensitive, exact substring match)", "").strip()

# Build query
if query:
    sql = """
        SELECT id, channel, video_id, speaker, start_time, end_time, upload_date, text, pos_tags, audio
        FROM data
        WHERE LOWER(text) LIKE LOWER(?)
        LIMIT 100
    """
    df = con.execute(sql, [f"%{query}%"]).df()
else:
    df = con.execute("""
        SELECT id, channel, video_id, speaker, start_time, end_time, upload_date, text, pos_tags, audio
        FROM data
        LIMIT 100
    """).df()

st.markdown(f"### Showing {len(df)} results")

if len(df) == 0:
    st.warning("No matches found.")
else:
    def render_audio_tag(audio_bytes, index):
        try:
            if isinstance(audio_bytes, (bytes, bytearray, memoryview)):
                data = bytes(audio_bytes)
            elif isinstance(audio_bytes, list):
                data = bytes(audio_bytes)
            else:
                return ""
            with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp:
                tmp.write(data)
                tmp.flush()
                return f'<audio controls preload="metadata" style="height:20px;width:100px;"><source src="file://{tmp.name}" type="audio/mpeg"></audio>'
        except Exception:
            return ""

    df["audio_player"] = [render_audio_tag(b, i) for i, b in enumerate(df["audio"])]
    df_display = df.drop(columns=["audio"])

    from streamlit.components.v1 import html

    def display_html_table(df):
        table_html = """
        <table border='1' style='border-collapse:collapse;width:100%;font-size:13px;'>
            <thead>
                <tr>
                    <th style='width:8em;'>id</th>
                    <th>channel</th>
                    <th>video_id</th>
                    <th>speaker</th>
                    <th>start_time</th>
                    <th>end_time</th>
                    <th>upload_date</th>
                    <th>text</th>
                    <th>pos_tags</th>
                    <th>audio</th>
                </tr>
            </thead>
            <tbody>
        """
        for _, row in df.iterrows():
            table_html += "<tr>"
            for col in ["id", "channel", "video_id", "speaker", "start_time", "end_time", "upload_date", "text", "pos_tags"]:
                table_html += f"<td>{row[col]}</td>"
            table_html += f"<td>{row['audio_player']}</td>"
            table_html += "</tr>"
        table_html += "</tbody></table>"
        return table_html

    st.markdown("### Results Table (Sortable with Audio Column)")
    html(display_html_table(df_display), height=800, scrolling=True)