app.py

import gradio as gr
import sqlite3
import chromadb
from chromadb.utils import embedding_functions
import os
import shutil
import zipfile

def convert_to_vector_db(sqlite_file):
    if sqlite_file is None:
        return None  # No file uploaded, return nothing

    # Copy uploaded file to a temporary path
    db_path = "temp.db"
    shutil.copy(sqlite_file, db_path)

    # Define vector DB path
    VECTOR_DB_PATH = "./legal_vector_db"

    # Clean existing directory if it exists
    if os.path.exists(VECTOR_DB_PATH):
        shutil.rmtree(VECTOR_DB_PATH)

    # Initialize ChromaDB persistent client
    chroma_client = chromadb.PersistentClient(path=VECTOR_DB_PATH)

    # Use multilingual embedding model suitable for Nepali
    sentence_transformer_ef = embedding_functions.SentenceTransformerEmbeddingFunction(
        model_name="paraphrase-multilingual-MiniLM-L12-v2"
    )

    # Create collection
    collection = chroma_client.create_collection(
        name="legal_cases_collection",
        embedding_function=sentence_transformer_ef
    )

    # Load data from SQLite
    with sqlite3.connect(db_path) as conn:
        cursor = conn.cursor()
        cursor.execute("""
            SELECT लिङ्क, निर्णय_नं, साल, मुद्दाको_किसिम, विषय, निवेदक, विपक्षी, प्रकरण, ठहर
            FROM cases
        """)
        rows = cursor.fetchall()

    documents = []
    metadatas = []
    ids = []

    for i, row in enumerate(rows):
        link, decision_no, year, mudda_type, subject, nibedak, vipakshi, prakaran, thahar = row
        
        # Combine text in Nepali format
        case_text = f"""
        मुद्दाको किसिम: {mudda_type}
        विषय: {subject}
        निवेदक: {nibedak}
        विपक्षी: {vipakshi}
        प्रकरण: {prakaran}
        ठहर: {thahar}
        """
        
        documents.append(case_text.strip())
        metadatas.append({
            "link": link,
            "decision_no": decision_no,
            "year": year,
            "mudda_type": mudda_type,
            "subject": subject,
            "nibedak": nibedak,
            "vipakshi": vipakshi,
            "prakaran": prakaran,
            "thahar": thahar
        })
        ids.append(f"case_{i}")

    # Add to collection in batches
    batch_size = 100
    for i in range(0, len(documents), batch_size):
        batch_docs = documents[i:i+batch_size]
        batch_meta = metadatas[i:i+batch_size]
        batch_ids = ids[i:i+batch_size]
        
        collection.add(
            documents=batch_docs,
            metadatas=batch_meta,
            ids=batch_ids
        )

    # Zip the vector DB directory
    zip_path = "legal_vector_db.zip"
    if os.path.exists(zip_path):
        os.remove(zip_path)

    with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zipf:
        for root, dirs, files in os.walk(VECTOR_DB_PATH):
            for file in files:
                zipf.write(os.path.join(root, file),
                           os.path.relpath(os.path.join(root, file),
                                           os.path.join(VECTOR_DB_PATH, '..')))

    # Clean up temp DB
    os.remove(db_path)

    return zip_path

# Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("SQLite DB to Vector DB Converter (Nepali Legal Cases Supported)")
    
    sqlite_upload = gr.File(label="Upload SQLite DB File (e.g., after_2061.db)")
    convert_btn = gr.Button("Convert to Vector DB")
    download_file = gr.File(label="Download Vector DB (ZIP File)")

    convert_btn.click(convert_to_vector_db, inputs=sqlite_upload, outputs=download_file)

if __name__ == "__main__":
    demo.launch()