import os
from pathlib import Path

# Pastikan Streamlit & HF cache menulis ke /tmp (selalu writable di Spaces)
os.environ.setdefault("HOME", "/tmp")
os.environ.setdefault("STREAMLIT_USER_SETTINGS_DIR", "/tmp/.streamlit")
os.environ.setdefault("HF_HOME", "/tmp/.cache/huggingface")
os.environ.setdefault("SENTENCE_TRANSFORMERS_HOME", "/tmp/.cache/sentence-transformers")
# TRANSFORMERS_CACHE deprecated; HF memakainya dari HF_HOME -> boleh dihapus
# os.environ.setdefault("TRANSFORMERS_CACHE", "/tmp/.cache/huggingface/transformers")

os.environ["TOKENIZERS_PARALLELISM"] = "false"
os.environ["STREAMLIT_BROWSER_GATHERUSAGESTATS"] = "false"
os.environ["STREAMLIT_SERVER_ADDRESS"] = "0.0.0.0"
os.environ["STREAMLIT_SERVER_PORT"] = os.environ.get("PORT", "7860")

# Buat folder-foldernya
for p in ["/tmp/.streamlit", "/tmp/.cache/huggingface", "/tmp/.cache/sentence-transformers"]:
    Path(p).mkdir(parents=True, exist_ok=True)
# ---- END PATCH ----

import streamlit as st


# Sekarang baru import streamlit
import streamlit as st
import faiss
import pickle
from sentence_transformers import SentenceTransformer
import pandas as pd

# Konfigurasi
MODEL_NAME = "Qwen/Qwen3-Embedding-0.6B"

# Get absolute path for data directory (independent from maintenance_web)
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
INDEX_DIR = os.path.join(SCRIPT_DIR, "data")

@st.cache_resource(show_spinner=True)
def load_model():
    """Load embedding model"""
    # Model akan di-cache otomatis
    model = SentenceTransformer(MODEL_NAME)
    return model

@st.cache_resource
def load_index():
    """Load FAISS index and metadata"""
    index_path = os.path.join(INDEX_DIR, "skripsi.faiss")
    metadata_path = os.path.join(INDEX_DIR, "metadata.pkl")
    
    if not os.path.exists(index_path):
        st.error(f"Index not found: {index_path}")
        return None, None
    
    if not os.path.exists(metadata_path):
        st.error(f"Metadata not found: {metadata_path}")
        return None, None
    
    index = faiss.read_index(index_path)
    
    with open(metadata_path, 'rb') as f:
        metadata = pickle.load(f)
    
    return index, metadata

def search(query, model, index, metadata, top_k=10):
    """Perform semantic search"""
    # Generate query embedding
    query_embedding = model.encode([query])
    
    # Search
    distances, indices = index.search(query_embedding, top_k)
    
    # Get data list from metadata
    data_list = metadata.get('data', [])
    
    # Format results
    results = []
    for i, (dist, idx) in enumerate(zip(distances[0], indices[0])):
        if idx < len(data_list):
            meta = data_list[idx]
            # Combine pembimbing info
            pembimbing = meta.get('nama_pembimbing', 'N/A')
            gelar_depan = meta.get('gelar_depan_pembimbing', '')
            gelar_belakang = meta.get('gelar_belakang_pembimbing', '')
            if gelar_depan or gelar_belakang:
                pembimbing = f"{gelar_depan} {pembimbing}, {gelar_belakang}".strip(', ')
            
            results.append({
                'Rank': i + 1,
                'Score': f"{dist:.4f}",
                'Judul': meta.get('judul', 'N/A'),
                'NIM': meta.get('nim', 'N/A'),
                'Nama': meta.get('nama', 'N/A'),
                'Pembimbing': pembimbing,
                'Tahun': meta.get('tahun', 'N/A'),
                'Semester': meta.get('semester', 'N/A')
            })
    
    return results

# Streamlit UI
st.set_page_config(page_title="Semantic Search - Skripsi UNIKOM", layout="wide")

st.title("🔍 Semantic Search - Database Skripsi Prodi Teknik Informatika UNIKOM")
st.markdown("*Pencarian semantik berdasarkan kemiripan makna judul skripsi*")
st.markdown("---")

# Sidebar
with st.sidebar:
    st.header("⚙️ Settings")
    top_k = st.slider("Number of results", min_value=5, max_value=50, value=10, step=5)
    
    st.markdown("---")
    st.markdown("### 📊 Model Info")
    st.info(f"""
    **Model**: {MODEL_NAME}
    **Index**: {INDEX_DIR}
    """)

# Load resources
try:
    model = load_model()
    index, metadata = load_index()
    
    if index is None or metadata is None:
        st.error("Failed to load index or metadata")
        st.stop()
    
    st.success(f"✅ Model loaded | Index: {index.ntotal} vectors | Dimension: {index.d}")
    
except Exception as e:
    st.error(f"Error loading resources: {e}")
    st.stop()

# Search interface
st.markdown("### 💬 Enter your search query")
query = st.text_input("Search Query", placeholder="e.g., machine learning, web application, sistem informasi...", label_visibility="collapsed")

if st.button("🔍 Search", type="primary") or query:
    if query.strip():
        with st.spinner("Searching..."):
            results = search(query, model, index, metadata, top_k)
        
        st.markdown(f"### 📋 Found {len(results)} results")
        
        # Display as dataframe
        if results:
            df = pd.DataFrame(results)
            st.dataframe(df, width="stretch", hide_index=True)
            
            # Detailed view
            st.markdown("---")
            st.markdown("### 📖 Detailed Results")
            for result in results:
                with st.expander(f"#{result['Rank']} - {result['Judul'][:100]}... (Score: {result['Score']})"):
                    col1, col2 = st.columns(2)
                    with col1:
                        st.markdown(f"**NIM**: {result['NIM']}")
                        st.markdown(f"**Nama**: {result['Nama']}")
                        st.markdown(f"**Pembimbing**: {result['Pembimbing']}")
                    with col2:
                        st.markdown(f"**Tahun**: {result['Tahun']}")
                        st.markdown(f"**Semester**: {result['Semester']}")
                    st.markdown(f"**Judul Lengkap**: {result['Judul']}")
        else:
            st.warning("No results found")
    else:
        st.warning("Please enter a search query")

# Footer
st.markdown("---")
st.markdown("""
<div style='text-align: center;'>
    <p><a href='https://galih.eu'>Galih Hermawan</a> | Akabot Research Group</p>
    <p>Prodi Teknik Informatika | Universitas Komputer Indonesia</p>
    <p>Powered by Qwen3 Embedding Model</p>
</div>
""", unsafe_allow_html=True)