database.py

"""
Database Module - Turso (libSQL) Connection
Handles all database operations for the CBT Companion app
"""

import os
import libsql_experimental as libsql
from datetime import datetime, date, timezone, timedelta
import uuid
import json

# IST = UTC+5:30
_IST = timezone(timedelta(hours=5, minutes=30))

def ist_now() -> str:
    """Return current IST time as ISO string for storing in DB."""
    return datetime.now(_IST).strftime("%Y-%m-%d %H:%M:%S")


class Database:
    """
    Database handler for Turso (libSQL).
    
    Required environment variables:
    - TURSO_DATABASE_URL: Your Turso database URL
    - TURSO_AUTH_TOKEN: Your Turso auth token
    """
    
    def __init__(self):
        url = os.environ.get("TURSO_DATABASE_URL")
        token = os.environ.get("TURSO_AUTH_TOKEN")
        
        if not url or not token:
            raise ValueError("TURSO_DATABASE_URL and TURSO_AUTH_TOKEN are required")
        
        self.conn = libsql.connect(database=url, auth_token=token)
        self._init_tables()
        print("✅ Database connected to Turso")
    
    def _init_tables(self):
        """Create tables if they don't exist."""
        
        # Users table
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS users (
                id TEXT PRIMARY KEY,
                email TEXT UNIQUE NOT NULL,
                password_hash TEXT NOT NULL,
                name TEXT NOT NULL,
                context TEXT DEFAULT 'person',
                created_at TEXT DEFAULT CURRENT_TIMESTAMP,
                fcm_token TEXT
            )
        """)
        # Add fcm_token column to existing deployments that lack it
        try:
            self.conn.execute("ALTER TABLE users ADD COLUMN fcm_token TEXT")
            self.conn.commit()
        except Exception:
            pass  # Column already exists
        # Add last_fcm_sent column for cooldown tracking
        try:
            self.conn.execute("ALTER TABLE users ADD COLUMN last_fcm_sent TEXT")
            self.conn.commit()
        except Exception:
            pass  # Column already exists
        
        # Sessions table
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS sessions (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                started_at TEXT DEFAULT CURRENT_TIMESTAMP,
                ended_at TEXT,
                mood_start INTEGER,
                mood_end INTEGER,
                locked_group TEXT,
                stages_reached INTEGER DEFAULT 1,
                completed INTEGER DEFAULT 0,
                messages_in_current_stage INTEGER DEFAULT 0,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)
        
        # Messages table
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS messages (
                id TEXT PRIMARY KEY,
                session_id TEXT NOT NULL,
                user_id TEXT NOT NULL,
                role TEXT NOT NULL,
                content TEXT NOT NULL,
                timestamp TEXT DEFAULT CURRENT_TIMESTAMP,
                FOREIGN KEY (session_id) REFERENCES sessions(id),
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)
        
        # Exercises completed
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS exercises_completed (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                session_id TEXT,
                exercise_id TEXT NOT NULL,
                exercise_name TEXT,
                group_type TEXT,
                completed_at TEXT DEFAULT CURRENT_TIMESTAMP,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)
        
        # Crisis flags
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS crisis_flags (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                user_name TEXT NOT NULL,
                user_email TEXT NOT NULL,
                session_id TEXT NOT NULL,
                message_content TEXT NOT NULL,
                trigger_word TEXT NOT NULL,
                flagged_at TEXT DEFAULT CURRENT_TIMESTAMP,
                reviewed INTEGER DEFAULT 0,
                FOREIGN KEY (user_id) REFERENCES users(id),
                FOREIGN KEY (session_id) REFERENCES sessions(id)
            )
        """)
        
        # User stats
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS user_stats (
                user_id TEXT PRIMARY KEY,
                total_sessions INTEGER DEFAULT 0,
                total_exercises INTEGER DEFAULT 0,
                current_streak INTEGER DEFAULT 0,
                last_session_date TEXT,
                distortion_counts TEXT DEFAULT '{}',
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)

        # Wearable sensor data
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS wearable_data (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                ppg REAL NOT NULL,
                gsr REAL NOT NULL,
                acc_x REAL NOT NULL,
                acc_y REAL NOT NULL,
                acc_z REAL NOT NULL,
                device_timestamp TEXT,
                recorded_at TEXT DEFAULT CURRENT_TIMESTAMP,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)

        # Migrate: add columns that may be missing from older table versions
        for col, col_type, default in [
            ("dri_score", "REAL", None),
            ("condition", "TEXT", None),
            ("acknowledged", "INTEGER", "0"),
            ("ml_prediction", "TEXT", None),
            ("ml_confidence", "REAL", None),
            ("risk_level", "INTEGER", None),
        ]:
            try:
                default_clause = f" DEFAULT {default}" if default is not None else ""
                self.conn.execute(
                    f"ALTER TABLE wearable_data ADD COLUMN {col} {col_type}{default_clause}"
                )
            except Exception:
                pass  # Column already exists

        # Create index for faster queries on wearable data
        self.conn.execute("""
            CREATE INDEX IF NOT EXISTS idx_wearable_user_time
            ON wearable_data(user_id, recorded_at DESC)
        """)

        # Device API keys for wearables
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS device_keys (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                api_key TEXT UNIQUE NOT NULL,
                device_name TEXT DEFAULT 'ESP32 Wearable',
                created_at TEXT DEFAULT CURRENT_TIMESTAMP,
                last_used_at TEXT,
                is_active INTEGER DEFAULT 1,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)

        # Index for fast API key lookups
        self.conn.execute("""
            CREATE INDEX IF NOT EXISTS idx_device_api_key
            ON device_keys(api_key)
        """)

        # Beck sessions table for cognitive restructuring protocol
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS beck_sessions (
                session_id TEXT PRIMARY KEY,

                -- Current state in the protocol
                beck_state TEXT DEFAULT 'VALIDATE',

                -- Phase 1: Capture
                original_thought TEXT,
                initial_belief_rating INTEGER,
                emotion TEXT,
                initial_emotion_intensity INTEGER,

                -- Phase 2: Discovery (6 Questions)
                q1_evidence_for TEXT,
                q1_evidence_against TEXT,
                q2_alternative TEXT,
                q3_worst TEXT,
                q3_best TEXT,
                q3_realistic TEXT,
                q4_effect TEXT,
                q5_friend TEXT,
                q6_action TEXT,

                -- Phase 3: Reframe
                adaptive_thought TEXT,
                new_thought_belief INTEGER,

                -- Phase 4: Measure
                final_belief_rating INTEGER,
                final_emotion_intensity INTEGER,

                -- Phase 5: Action
                action_plan TEXT,

                -- Metadata
                started_at TEXT DEFAULT CURRENT_TIMESTAMP,
                completed_at TEXT,
                belief_improvement INTEGER,
                emotion_improvement INTEGER,

                FOREIGN KEY (session_id) REFERENCES sessions(id)
            )
        """)

        # ML window predictions — one row per inference call (every 25 readings)
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS ml_window_predictions (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                prediction TEXT NOT NULL,
                risk_level INTEGER NOT NULL,
                confidence REAL NOT NULL,
                readings_used INTEGER DEFAULT 25,
                predicted_at TEXT DEFAULT CURRENT_TIMESTAMP,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)

        self.conn.execute("""
            CREATE INDEX IF NOT EXISTS idx_ml_window_user_time
            ON ml_window_predictions(user_id, predicted_at DESC)
        """)

        # Depression episodes table (ML-detected high stress periods)
        self.conn.execute("""
            CREATE TABLE IF NOT EXISTS depression_episodes (
                id TEXT PRIMARY KEY,
                user_id TEXT NOT NULL,
                start_time TEXT NOT NULL,
                end_time TEXT,
                peak_risk_level INTEGER,
                total_readings INTEGER DEFAULT 0,
                avg_confidence REAL,
                is_active INTEGER DEFAULT 1,
                created_at TEXT DEFAULT CURRENT_TIMESTAMP,
                FOREIGN KEY (user_id) REFERENCES users(id)
            )
        """)

        # Index for fast episode lookups
        self.conn.execute("""
            CREATE INDEX IF NOT EXISTS idx_depression_episodes_user
            ON depression_episodes(user_id, is_active DESC)
        """)

        self.conn.commit()
    
    # ==================== USER OPERATIONS ====================
    
    def create_user(self, email: str, password_hash: str, name: str, context: str = "person") -> dict:
        """Create a new user."""
        user_id = str(uuid.uuid4())
        
        try:
            self.conn.execute(
                "INSERT INTO users (id, email, password_hash, name, context) VALUES (?, ?, ?, ?, ?)",
                (user_id, email.lower(), password_hash, name, context)
            )
            
            # Initialize user stats
            self.conn.execute(
                "INSERT INTO user_stats (user_id) VALUES (?)",
                (user_id,)
            )
            
            self.conn.commit()
            
            return {"id": user_id, "email": email, "name": name, "context": context}
        
        except Exception as e:
            if "UNIQUE constraint" in str(e):
                raise ValueError("Email already exists")
            raise e
    
    def get_user_by_email(self, email: str) -> dict:
        """Get user by email."""
        result = self.conn.execute(
            "SELECT id, email, password_hash, name, context FROM users WHERE email = ?",
            (email.lower(),)
        ).fetchone()
        
        if result:
            return {
                "id": result[0],
                "email": result[1],
                "password_hash": result[2],
                "name": result[3],
                "context": result[4]
            }
        return None
    
    def get_user_by_id(self, user_id: str) -> dict:
        """Get user by ID."""
        result = self.conn.execute(
            "SELECT id, email, name, context FROM users WHERE id = ?",
            (user_id,)
        ).fetchone()
        
        if result:
            return {
                "id": result[0],
                "email": result[1],
                "name": result[2],
                "context": result[3]
            }
        return None
    
    def save_fcm_token(self, user_id: str, token: str):
        """Save or update FCM device token for a user."""
        self.conn.execute(
            "UPDATE users SET fcm_token = ? WHERE id = ?",
            (token, user_id)
        )
        self.conn.commit()

    def get_fcm_token(self, user_id: str) -> str:
        """Get the FCM token for a user, or None."""
        result = self.conn.execute(
            "SELECT fcm_token FROM users WHERE id = ?",
            (user_id,)
        ).fetchone()
        return result[0] if result else None

    def fcm_cooldown_ok(self, user_id: str, cooldown_minutes: int = 30) -> bool:
        """Return True if enough time has passed since the last FCM push."""
        result = self.conn.execute(
            "SELECT last_fcm_sent FROM users WHERE id = ?", (user_id,)
        ).fetchone()
        if not result or not result[0]:
            return True
        try:
            last = datetime.fromisoformat(result[0])
            return (datetime.now(_IST) - last).total_seconds() >= cooldown_minutes * 60
        except Exception:
            return True

    def update_fcm_sent_time(self, user_id: str):
        """Record the current UTC time as the last FCM send for cooldown tracking."""
        self.conn.execute(
            "UPDATE users SET last_fcm_sent = ? WHERE id = ?",
            (ist_now(), user_id)
        )
        self.conn.commit()

    # ==================== SESSION OPERATIONS ====================

    def create_session(self, user_id: str) -> str:
        """Create a new chat session."""
        session_id = str(uuid.uuid4())
        
        self.conn.execute(
            "INSERT INTO sessions (id, user_id) VALUES (?, ?)",
            (session_id, user_id)
        )
        self.conn.commit()
        
        return session_id
    
    def get_session(self, session_id: str) -> dict:
        """Get session by ID."""
        result = self.conn.execute(
            """SELECT id, user_id, started_at, ended_at, mood_start, mood_end, 
                      locked_group, stages_reached, completed, messages_in_current_stage
               FROM sessions WHERE id = ?""",
            (session_id,)
        ).fetchone()
        
        if result:
            return {
                "id": result[0],
                "user_id": result[1],
                "started_at": result[2],
                "ended_at": result[3],
                "mood_start": result[4],
                "mood_end": result[5],
                "locked_group": result[6],
                "current_stage": result[7],
                "completed": bool(result[8]),
                "messages_in_current_stage": result[9] or 0,
                "state": "COMPLETED" if result[8] else ("IN_PROGRESS" if result[6] else "WAITING_FOR_PROBLEM")
            }
        return None
    
    def update_session(self, session_id: str, **kwargs):
        """Update session fields."""
        allowed_fields = {
            "mood_start", "mood_end", "locked_group", "stages_reached", 
            "completed", "ended_at", "messages_in_current_stage"
        }
        
        updates = {k: v for k, v in kwargs.items() if k in allowed_fields}
        
        if not updates:
            return
        
        # Map stages_reached to the column
        if "current_stage" in kwargs:
            updates["stages_reached"] = kwargs["current_stage"]
        
        set_clause = ", ".join(f"{k} = ?" for k in updates.keys())
        values = tuple(list(updates.values()) + [session_id])

        self.conn.execute(
            f"UPDATE sessions SET {set_clause} WHERE id = ?",
            values
        )
        self.conn.commit()
    
    def increment_stage_messages(self, session_id: str) -> int:
        """Increment message counter for current stage."""
        self.conn.execute(
            "UPDATE sessions SET messages_in_current_stage = messages_in_current_stage + 1 WHERE id = ?",
            (session_id,)
        )
        self.conn.commit()
        
        result = self.conn.execute(
            "SELECT messages_in_current_stage FROM sessions WHERE id = ?",
            (session_id,)
        ).fetchone()
        
        return result[0] if result else 0
    
    def reset_stage_messages(self, session_id: str):
        """Reset message counter (called when advancing stage)."""
        self.conn.execute(
            "UPDATE sessions SET messages_in_current_stage = 0 WHERE id = ?",
            (session_id,)
        )
        self.conn.commit()
    
    def get_user_sessions(self, user_id: str, limit: int = 20) -> list:
        """Get user's past sessions."""
        results = self.conn.execute(
            """SELECT id, started_at, ended_at, mood_start, mood_end, 
                      locked_group, stages_reached, completed
               FROM sessions WHERE user_id = ? 
               ORDER BY started_at DESC LIMIT ?""",
            (user_id, limit)
        ).fetchall()
        
        return [
            {
                "id": r[0],
                "started_at": r[1],
                "ended_at": r[2],
                "mood_start": r[3],
                "mood_end": r[4],
                "locked_group": r[5],
                "stages_reached": r[6],
                "completed": bool(r[7])
            }
            for r in results
        ]
    
    # ==================== MESSAGE OPERATIONS ====================
    
    def add_message(self, session_id: str, user_id: str, role: str, content: str):
        """Add a message to conversation history."""
        message_id = str(uuid.uuid4())
        
        self.conn.execute(
            "INSERT INTO messages (id, session_id, user_id, role, content) VALUES (?, ?, ?, ?, ?)",
            (message_id, session_id, user_id, role, content)
        )
        self.conn.commit()
    
    def get_session_messages(self, session_id: str, limit: int = None) -> list:
        """Get messages for a session."""
        query = """SELECT role, content, timestamp FROM messages 
                   WHERE session_id = ? ORDER BY timestamp ASC"""
        
        if limit:
            query += f" LIMIT {limit}"
        
        results = self.conn.execute(query, (session_id,)).fetchall()
        
        return [
            {"role": r[0], "content": r[1], "timestamp": r[2]}
            for r in results
        ]
    
    def get_recent_messages(self, session_id: str, n: int = 6) -> list:
        """Get last n messages for context window."""
        results = self.conn.execute(
            """SELECT role, content FROM messages 
               WHERE session_id = ? ORDER BY timestamp DESC LIMIT ?""",
            (session_id, n)
        ).fetchall()
        
        # Reverse to get chronological order
        return [{"role": r[0], "content": r[1]} for r in reversed(results)]
    
    # ==================== EXERCISE OPERATIONS ====================
    
    def log_exercise_completed(self, user_id: str, session_id: str, exercise_id: str, 
                                exercise_name: str, group_type: str):
        """Log a completed exercise."""
        entry_id = str(uuid.uuid4())
        
        self.conn.execute(
            """INSERT INTO exercises_completed 
               (id, user_id, session_id, exercise_id, exercise_name, group_type) 
               VALUES (?, ?, ?, ?, ?, ?)""",
            (entry_id, user_id, session_id, exercise_id, exercise_name, group_type)
        )
        
        # Update user stats
        self.conn.execute(
            "UPDATE user_stats SET total_exercises = total_exercises + 1 WHERE user_id = ?",
            (user_id,)
        )
        
        self.conn.commit()
    
    # ==================== CRISIS FLAG OPERATIONS ====================
    
    def flag_crisis(self, user_id: str, user_name: str, user_email: str,
                    session_id: str, message_content: str, trigger_word: str):
        """Flag a crisis message."""
        flag_id = str(uuid.uuid4())
        
        self.conn.execute(
            """INSERT INTO crisis_flags 
               (id, user_id, user_name, user_email, session_id, message_content, trigger_word) 
               VALUES (?, ?, ?, ?, ?, ?, ?)""",
            (flag_id, user_id, user_name, user_email, session_id, message_content, trigger_word)
        )
        self.conn.commit()
        
        print(f"🚨 CRISIS FLAGGED: User {user_name} ({user_email}) - Trigger: {trigger_word}")
    
    # ==================== STATS OPERATIONS ====================
    
    def get_user_stats(self, user_id: str) -> dict:
        """Get user statistics."""
        result = self.conn.execute(
            """SELECT total_sessions, total_exercises, current_streak, 
                      last_session_date, distortion_counts
               FROM user_stats WHERE user_id = ?""",
            (user_id,)
        ).fetchone()
        
        if result:
            return {
                "total_sessions": result[0],
                "total_exercises": result[1],
                "current_streak": result[2],
                "last_session_date": result[3],
                "distortion_counts": json.loads(result[4] or "{}")
            }
        
        return {
            "total_sessions": 0,
            "total_exercises": 0,
            "current_streak": 0,
            "last_session_date": None,
            "distortion_counts": {}
        }
    
    def update_user_stats_on_session_end(self, user_id: str, locked_group: str):
        """Update user stats when a session ends."""
        today = date.today().isoformat()
        
        # Get current stats
        stats = self.get_user_stats(user_id)
        
        # Update distortion counts
        distortion_counts = stats["distortion_counts"]
        if locked_group and locked_group != "G0":
            distortion_counts[locked_group] = distortion_counts.get(locked_group, 0) + 1
        
        # Calculate streak
        last_date = stats["last_session_date"]
        if last_date == today:
            new_streak = stats["current_streak"]
        elif last_date == (date.today().replace(day=date.today().day - 1)).isoformat():
            new_streak = stats["current_streak"] + 1
        else:
            new_streak = 1
        
        # Update stats
        self.conn.execute(
            """UPDATE user_stats SET 
               total_sessions = total_sessions + 1,
               current_streak = ?,
               last_session_date = ?,
               distortion_counts = ?
               WHERE user_id = ?""",
            (new_streak, today, json.dumps(distortion_counts), user_id)
        )
        self.conn.commit()

    # ==================== WEARABLE DATA OPERATIONS ====================

    def save_wearable_data(self, user_id: str, ppg: float, gsr: float,
                           acc_x: float, acc_y: float, acc_z: float,
                           device_timestamp: str = None) -> str:
        """Save wearable sensor data."""
        record_id = str(uuid.uuid4())

        self.conn.execute(
            """INSERT INTO wearable_data
               (id, user_id, ppg, gsr, acc_x, acc_y, acc_z, device_timestamp, recorded_at)
               VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)""",
            (record_id, user_id, ppg, gsr, acc_x, acc_y, acc_z, device_timestamp, ist_now())
        )
        self.conn.commit()

        return record_id

    def get_latest_wearable_data(self, user_id: str) -> dict:
        """Get the most recent wearable data for a user."""
        result = self.conn.execute(
            """SELECT id, ppg, gsr, acc_x, acc_y, acc_z, device_timestamp, recorded_at
               FROM wearable_data WHERE user_id = ?
               ORDER BY recorded_at DESC LIMIT 1""",
            (user_id,)
        ).fetchone()

        if result:
            return {
                "id": result[0],
                "ppg": result[1],
                "gsr": result[2],
                "acc_x": result[3],
                "acc_y": result[4],
                "acc_z": result[5],
                "device_timestamp": result[6],
                "recorded_at": result[7]
            }
        return None

    def get_wearable_history(self, user_id: str, limit: int = 100,
                             offset: int = 0, start_date: str = None,
                             end_date: str = None) -> list:
        """Get wearable data history for a user."""
        query = """SELECT id, ppg, gsr, acc_x, acc_y, acc_z, device_timestamp, recorded_at
                   FROM wearable_data WHERE user_id = ?"""
        params = [user_id]

        if start_date:
            query += " AND recorded_at >= ?"
            params.append(start_date)

        if end_date:
            query += " AND recorded_at <= ?"
            params.append(end_date)

        query += " ORDER BY recorded_at DESC LIMIT ? OFFSET ?"
        params.extend([limit, offset])

        results = self.conn.execute(query, tuple(params)).fetchall()

        return [
            {
                "id": r[0],
                "ppg": r[1],
                "gsr": r[2],
                "acc_x": r[3],
                "acc_y": r[4],
                "acc_z": r[5],
                "device_timestamp": r[6],
                "recorded_at": r[7]
            }
            for r in results
        ]

    def get_wearable_stats(self, user_id: str, period: str = "day") -> dict:
        """Get aggregated statistics for wearable data."""
        # Calculate date filter based on period
        if period == "day":
            date_filter = "datetime('now', '-1 day')"
        elif period == "week":
            date_filter = "datetime('now', '-7 days')"
        else:  # month
            date_filter = "datetime('now', '-30 days')"

        result = self.conn.execute(
            f"""SELECT
                COUNT(*) as count,
                AVG(ppg) as avg_ppg,
                MIN(ppg) as min_ppg,
                MAX(ppg) as max_ppg,
                AVG(gsr) as avg_gsr,
                MIN(gsr) as min_gsr,
                MAX(gsr) as max_gsr,
                AVG(acc_x) as avg_acc_x,
                AVG(acc_y) as avg_acc_y,
                AVG(acc_z) as avg_acc_z
               FROM wearable_data
               WHERE user_id = ? AND recorded_at >= {date_filter}""",
            (user_id,)
        ).fetchone()

        if result and result[0] > 0:
            return {
                "reading_count": result[0],
                "ppg": {
                    "avg": round(result[1], 2) if result[1] else None,
                    "min": round(result[2], 2) if result[2] else None,
                    "max": round(result[3], 2) if result[3] else None
                },
                "gsr": {
                    "avg": round(result[4], 4) if result[4] else None,
                    "min": round(result[5], 4) if result[5] else None,
                    "max": round(result[6], 4) if result[6] else None
                },
                "accelerometer": {
                    "avg_x": round(result[7], 4) if result[7] else None,
                    "avg_y": round(result[8], 4) if result[8] else None,
                    "avg_z": round(result[9], 4) if result[9] else None
                }
            }

        return {
            "reading_count": 0,
            "ppg": {"avg": None, "min": None, "max": None},
            "gsr": {"avg": None, "min": None, "max": None},
            "accelerometer": {"avg_x": None, "avg_y": None, "avg_z": None}
        }

    # ==================== DEVICE KEY OPERATIONS ====================

    def create_device_key(self, user_id: str, device_name: str = "ESP32 Wearable") -> dict:
        """Create a new device API key for a user."""
        import secrets

        key_id = str(uuid.uuid4())
        # Generate a secure 32-character API key
        api_key = secrets.token_hex(16)

        self.conn.execute(
            """INSERT INTO device_keys (id, user_id, api_key, device_name)
               VALUES (?, ?, ?, ?)""",
            (key_id, user_id, api_key, device_name)
        )
        self.conn.commit()

        return {
            "id": key_id,
            "api_key": api_key,
            "device_name": device_name,
            "created_at": ist_now()
        }

    def get_user_by_device_key(self, api_key: str) -> dict:
        """Get user associated with a device API key."""
        result = self.conn.execute(
            """SELECT u.id, u.email, u.name, u.context, dk.id as device_id
               FROM device_keys dk
               JOIN users u ON dk.user_id = u.id
               WHERE dk.api_key = ? AND dk.is_active = 1""",
            (api_key,)
        ).fetchone()

        if result:
            # Update last_used_at
            self.conn.execute(
                "UPDATE device_keys SET last_used_at = ? WHERE api_key = ?",
                (ist_now(), api_key)
            )
            self.conn.commit()

            return {
                "id": result[0],
                "email": result[1],
                "name": result[2],
                "context": result[3],
                "device_id": result[4]
            }
        return None

    def get_user_device_keys(self, user_id: str) -> list:
        """Get all device keys for a user."""
        results = self.conn.execute(
            """SELECT id, api_key, device_name, created_at, last_used_at, is_active
               FROM device_keys
               WHERE user_id = ?
               ORDER BY created_at DESC""",
            (user_id,)
        ).fetchall()

        return [
            {
                "id": r[0],
                "api_key": r[1][:8] + "..." + r[1][-4:],  # Masked for security
                "device_name": r[2],
                "created_at": r[3],
                "last_used_at": r[4],
                "is_active": bool(r[5])
            }
            for r in results
        ]

    def revoke_device_key(self, key_id: str, user_id: str) -> bool:
        """Revoke a device API key."""
        # Pre-check existence — libsql rowcount is unreliable for UPDATE
        exists = self.conn.execute(
            "SELECT id FROM device_keys WHERE id = ? AND user_id = ?",
            (key_id, user_id)
        ).fetchone()
        if not exists:
            return False
        self.conn.execute(
            "UPDATE device_keys SET is_active = 0 WHERE id = ? AND user_id = ?",
            (key_id, user_id)
        )
        self.conn.commit()
        return True

    def delete_device_key(self, key_id: str, user_id: str) -> bool:
        """Permanently delete a device API key."""
        result = self.conn.execute(
            "DELETE FROM device_keys WHERE id = ? AND user_id = ?",
            (key_id, user_id)
        )
        self.conn.commit()
        return result.rowcount > 0

    # ==================== ADMIN OPERATIONS ====================

    def get_all_users(self) -> list:
        """Get all users with basic info for admin panel."""
        results = self.conn.execute(
            """SELECT u.id, u.email, u.name, u.context, u.created_at,
                      s.total_sessions, s.total_exercises, s.current_streak, s.last_session_date
               FROM users u
               LEFT JOIN user_stats s ON u.id = s.user_id
               ORDER BY u.created_at DESC"""
        ).fetchall()

        users = []
        for r in results:
            # Count crisis flags for this user
            flag_count = self.conn.execute(
                "SELECT COUNT(*) FROM crisis_flags WHERE user_id = ? AND reviewed = 0",
                (r[0],)
            ).fetchone()[0]

            # Check for active ML-detected risk episode
            active_ep = self.conn.execute(
                "SELECT COUNT(*) FROM depression_episodes WHERE user_id = ? AND is_active = 1",
                (r[0],)
            ).fetchone()[0]

            users.append({
                "id": r[0],
                "email": r[1],
                "name": r[2],
                "context": r[3],
                "created_at": r[4],
                "total_sessions": r[5] or 0,
                "total_exercises": r[6] or 0,
                "current_streak": r[7] or 0,
                "last_session_date": r[8],
                "unreviewed_alerts": flag_count,
                "has_active_episode": active_ep > 0,
            })

        return users

    def get_user_full_details(self, user_id: str) -> dict:
        """Get complete user data for admin patient detail view."""
        user = self.get_user_by_id(user_id)
        if not user:
            return None

        # Get user stats
        stats = self.get_user_stats(user_id)

        # Get created_at
        created_at = self.conn.execute(
            "SELECT created_at FROM users WHERE id = ?",
            (user_id,)
        ).fetchone()

        # Get sessions
        sessions = self.get_user_sessions(user_id, limit=50)

        # Get crisis history
        crisis_flags = self.conn.execute(
            """SELECT id, session_id, message_content, trigger_word, flagged_at, reviewed
               FROM crisis_flags WHERE user_id = ?
               ORDER BY flagged_at DESC""",
            (user_id,)
        ).fetchall()

        crisis_history = [
            {
                "id": r[0],
                "session_id": r[1],
                "message_content": r[2],
                "trigger_word": r[3],
                "flagged_at": r[4],
                "reviewed": bool(r[5])
            }
            for r in crisis_flags
        ]

        # Get latest wearable data
        latest_wearable = self.get_latest_wearable_data(user_id)

        return {
            **user,
            "created_at": created_at[0] if created_at else None,
            "stats": stats,
            "sessions": sessions,
            "crisis_history": crisis_history,
            "latest_wearable": latest_wearable
        }

    def get_all_crisis_flags(self, reviewed: bool = None) -> list:
        """Get all crisis flags, optionally filtered by reviewed status."""
        query = """SELECT cf.id, cf.user_id, cf.user_name, cf.user_email,
                          cf.session_id, cf.message_content, cf.trigger_word,
                          cf.flagged_at, cf.reviewed
                   FROM crisis_flags cf
                   ORDER BY cf.flagged_at DESC"""

        if reviewed is not None:
            query = """SELECT cf.id, cf.user_id, cf.user_name, cf.user_email,
                              cf.session_id, cf.message_content, cf.trigger_word,
                              cf.flagged_at, cf.reviewed
                       FROM crisis_flags cf
                       WHERE cf.reviewed = ?
                       ORDER BY cf.flagged_at DESC"""
            results = self.conn.execute(query, (1 if reviewed else 0,)).fetchall()
        else:
            results = self.conn.execute(query).fetchall()

        return [
            {
                "id": r[0],
                "user_id": r[1],
                "user_name": r[2],
                "user_email": r[3],
                "session_id": r[4],
                "message_content": r[5],
                "trigger_word": r[6],
                "flagged_at": r[7],
                "reviewed": bool(r[8])
            }
            for r in results
        ]

    def mark_crisis_reviewed(self, flag_id: str) -> bool:
        """Mark a crisis flag as reviewed."""
        self.conn.execute(
            "UPDATE crisis_flags SET reviewed = 1 WHERE id = ?",
            (flag_id,)
        )
        self.conn.commit()
        return True

    def get_dashboard_stats(self) -> dict:
        """Get overview statistics for admin dashboard."""
        # Total users
        total_users = self.conn.execute(
            "SELECT COUNT(*) FROM users"
        ).fetchone()[0]

        # Sessions today
        sessions_today = self.conn.execute(
            """SELECT COUNT(*) FROM sessions
               WHERE DATE(started_at) = DATE('now')"""
        ).fetchone()[0]

        # Unreviewed crisis flags
        unreviewed_alerts = self.conn.execute(
            "SELECT COUNT(*) FROM crisis_flags WHERE reviewed = 0"
        ).fetchone()[0]

        # Average mood improvement (mood_end - mood_start for completed sessions)
        mood_result = self.conn.execute(
            """SELECT AVG(mood_end - mood_start)
               FROM sessions
               WHERE mood_start IS NOT NULL
               AND mood_end IS NOT NULL
               AND completed = 1"""
        ).fetchone()
        avg_mood_change = round(mood_result[0], 2) if mood_result[0] else 0

        # Total sessions
        total_sessions = self.conn.execute(
            "SELECT COUNT(*) FROM sessions"
        ).fetchone()[0]

        # Completed sessions
        completed_sessions = self.conn.execute(
            "SELECT COUNT(*) FROM sessions WHERE completed = 1"
        ).fetchone()[0]

        # Active ML-detected risk episodes
        active_episodes = self.conn.execute(
            "SELECT COUNT(*) FROM depression_episodes WHERE is_active = 1"
        ).fetchone()[0]

        # Wearable readings in last 24 hours
        wearable_today = self.conn.execute(
            """SELECT COUNT(*) FROM wearable_data
               WHERE recorded_at >= datetime('now', '-24 hours')"""
        ).fetchone()[0]

        return {
            "total_users": total_users,
            "sessions_today": sessions_today,
            "unreviewed_alerts": unreviewed_alerts,
            "avg_mood_change": avg_mood_change,
            "total_sessions": total_sessions,
            "completed_sessions": completed_sessions,
            "active_episodes": active_episodes,
            "wearable_readings_today": wearable_today,
        }

    def get_user_wearable_summary(self, user_id: str) -> dict:
        """Get vitals summary for a patient."""
        # Get stats for day, week, month
        day_stats = self.get_wearable_stats(user_id, "day")
        week_stats = self.get_wearable_stats(user_id, "week")
        month_stats = self.get_wearable_stats(user_id, "month")

        # Get latest reading
        latest = self.get_latest_wearable_data(user_id)

        return {
            "latest": latest,
            "day": day_stats,
            "week": week_stats,
            "month": month_stats
        }

    def get_wearable_timeseries(self, user_id: str, hours: int = 24) -> list:
        """Get time-series wearable data for charts."""
        results = self.conn.execute(
            """SELECT ppg, gsr, acc_x, acc_y, acc_z, recorded_at
               FROM wearable_data
               WHERE user_id = ?
               AND recorded_at >= datetime('now', ? || ' hours')
               ORDER BY recorded_at ASC""",
            (user_id, -hours)
        ).fetchall()

        return [
            {
                "ppg": r[0],
                "gsr": r[1],
                "acc_x": r[2],
                "acc_y": r[3],
                "acc_z": r[4],
                "recorded_at": r[5]
            }
            for r in results
        ]

    def get_daily_session_counts(self, days: int = 30) -> list:
        """Get daily session counts for trend chart."""
        results = self.conn.execute(
            """SELECT DATE(started_at) as day, COUNT(*) as count
               FROM sessions
               WHERE started_at >= datetime('now', ? || ' days')
               GROUP BY DATE(started_at)
               ORDER BY day ASC""",
            (-days,)
        ).fetchall()

        return [{"date": r[0], "count": r[1]} for r in results]

    def get_distortion_distribution(self) -> dict:
        """Get aggregate distortion group distribution."""
        results = self.conn.execute(
            """SELECT locked_group, COUNT(*) as count
               FROM sessions
               WHERE locked_group IS NOT NULL AND locked_group != 'G0'
               GROUP BY locked_group"""
        ).fetchall()

        distribution = {"G1": 0, "G2": 0, "G3": 0, "G4": 0}
        for r in results:
            if r[0] in distribution:
                distribution[r[0]] = r[1]

        return distribution

    def get_user_mood_history(self, user_id: str, limit: int = 20) -> list:
        """Get mood history for a user's sessions."""
        results = self.conn.execute(
            """SELECT started_at, mood_start, mood_end, locked_group, completed
               FROM sessions
               WHERE user_id = ?
               AND mood_start IS NOT NULL
               ORDER BY started_at DESC
               LIMIT ?""",
            (user_id, limit)
        ).fetchall()

        return [
            {
                "date": r[0],
                "mood_start": r[1],
                "mood_end": r[2],
                "locked_group": r[3],
                "completed": bool(r[4])
            }
            for r in reversed(results)  # Chronological order
        ]

    def get_user_distortion_pattern(self, user_id: str) -> dict:
        """Get distortion pattern for radar chart."""
        stats = self.get_user_stats(user_id)
        counts = stats.get("distortion_counts", {})

        return {
            "G1": counts.get("G1", 0),
            "G2": counts.get("G2", 0),
            "G3": counts.get("G3", 0),
            "G4": counts.get("G4", 0)
        }

    # ==================== BECK SESSION OPERATIONS ====================

    def create_beck_session(self, session_id: str) -> dict:
        """Initialize a Beck session when distortion is detected."""
        try:
            self.conn.execute(
                """INSERT INTO beck_sessions (session_id, beck_state)
                   VALUES (?, 'VALIDATE')""",
                (session_id,)
            )
            self.conn.commit()
            return {"session_id": session_id, "beck_state": "VALIDATE"}
        except Exception as e:
            print(f"Error creating Beck session: {e}")
            return None

    def get_beck_session(self, session_id: str) -> dict:
        """Get current Beck session state and all data."""
        result = self.conn.execute(
            """SELECT session_id, beck_state, original_thought, initial_belief_rating,
                      emotion, initial_emotion_intensity, q1_evidence_for, q1_evidence_against,
                      q2_alternative, q3_worst, q3_best, q3_realistic, q4_effect, q5_friend,
                      q6_action, adaptive_thought, new_thought_belief, final_belief_rating,
                      final_emotion_intensity, action_plan, started_at, completed_at,
                      belief_improvement, emotion_improvement, full_protocol_state,
                      quality_scores, agenda_items, bdi_score
               FROM beck_sessions WHERE session_id = ?""",
            (session_id,)
        ).fetchone()

        if result:
            return {
                "session_id": result[0],
                "beck_state": result[1],
                "original_thought": result[2],
                "initial_belief_rating": result[3],
                "emotion": result[4],
                "initial_emotion_intensity": result[5],
                "q1_evidence_for": result[6],
                "q1_evidence_against": result[7],
                "q2_alternative": result[8],
                "q3_worst": result[9],
                "q3_best": result[10],
                "q3_realistic": result[11],
                "q4_effect": result[12],
                "q5_friend": result[13],
                "q6_action": result[14],
                "adaptive_thought": result[15],
                "new_thought_belief": result[16],
                "final_belief_rating": result[17],
                "final_emotion_intensity": result[18],
                "action_plan": result[19],
                "started_at": result[20],
                "completed_at": result[21],
                "belief_improvement": result[22],
                "emotion_improvement": result[23],
                "full_protocol_state": result[24],
                "quality_scores": result[25],
                "agenda_items": result[26],
                "bdi_score": result[27],
            }
        return None

    def update_beck_state(self, session_id: str, new_state: str, **fields):
        """Update state and save any new field values."""
        # Start with state update
        updates = {"beck_state": new_state}
        updates.update(fields)

        set_clause = ", ".join(f"{k} = ?" for k in updates.keys())
        values = tuple(list(updates.values()) + [session_id])

        self.conn.execute(
            f"UPDATE beck_sessions SET {set_clause} WHERE session_id = ?",
            values
        )
        self.conn.commit()

    def complete_beck_session(self, session_id: str):
        """Mark session complete, calculate improvements."""
        # Get current data
        beck_data = self.get_beck_session(session_id)
        if not beck_data:
            return

        # Calculate improvements
        belief_improvement = None
        emotion_improvement = None

        if beck_data.get('initial_belief_rating') and beck_data.get('final_belief_rating'):
            belief_improvement = beck_data['initial_belief_rating'] - beck_data['final_belief_rating']

        if beck_data.get('initial_emotion_intensity') and beck_data.get('final_emotion_intensity'):
            emotion_improvement = beck_data['initial_emotion_intensity'] - beck_data['final_emotion_intensity']

        # Update completion status
        self.conn.execute(
            """UPDATE beck_sessions SET
               completed_at = CURRENT_TIMESTAMP,
               belief_improvement = ?,
               emotion_improvement = ?
               WHERE session_id = ?""",
            (belief_improvement, emotion_improvement, session_id)
        )
        self.conn.commit()

    # ==================== ML INFERENCE & DEPRESSION TRACKING ====================

    def get_recent_readings_for_ml(self, user_id: str, limit: int = 50) -> list:
        """
        Get most recent sensor readings for ML inference.
        Returns data ordered from OLDEST to NEWEST (required for time-series analysis).
        """
        results = self.conn.execute(
            """SELECT ppg, gsr, acc_x, acc_y, acc_z, recorded_at
               FROM wearable_data
               WHERE user_id = ?
               ORDER BY recorded_at DESC
               LIMIT ?""",
            (user_id, limit)
        ).fetchall()

        # Reverse to get oldest-to-newest order
        readings = [
            {
                "ppg": r[0],
                "gsr": r[1],
                "acc_x": r[2],
                "acc_y": r[3],
                "acc_z": r[4],
                "timestamp": r[5]
            }
            for r in reversed(results)
        ]

        return readings

    def update_ml_prediction(self, record_id: str, prediction: str, confidence: float, risk_level: int):
        """Update a wearable data record with ML prediction results."""
        condition = prediction  # prediction is already "NORMAL", "MILD_STRESS", or "HIGH_STRESS"
        self.conn.execute(
            """UPDATE wearable_data
               SET ml_prediction = ?, ml_confidence = ?, risk_level = ?,
                   condition = ?, dri_score = ?
               WHERE id = ?""",
            (prediction, confidence, risk_level, condition, confidence, record_id)
        )
        self.conn.commit()

    def get_active_depression_episode(self, user_id: str) -> dict:
        """Get the currently active depression episode for a user, if any."""
        result = self.conn.execute(
            """SELECT id, user_id, start_time, peak_risk_level, total_readings, avg_confidence
               FROM depression_episodes
               WHERE user_id = ? AND is_active = 1
               ORDER BY start_time DESC LIMIT 1""",
            (user_id,)
        ).fetchone()

        if result:
            return {
                "id": result[0],
                "user_id": result[1],
                "start_time": result[2],
                "peak_risk_level": result[3],
                "total_readings": result[4],
                "avg_confidence": result[5]
            }
        return None

    def start_depression_episode(self, user_id: str, risk_level: int, confidence: float) -> str:
        """Start a new depression episode."""
        episode_id = str(uuid.uuid4())

        self.conn.execute(
            """INSERT INTO depression_episodes
               (id, user_id, start_time, peak_risk_level, total_readings, avg_confidence, is_active)
               VALUES (?, ?, CURRENT_TIMESTAMP, ?, 1, ?, 1)""",
            (episode_id, user_id, risk_level, confidence)
        )
        self.conn.commit()

        return episode_id

    def update_depression_episode(self, episode_id: str, risk_level: int, confidence: float):
        """Update an active depression episode with new reading."""
        # Get current stats
        result = self.conn.execute(
            """SELECT total_readings, avg_confidence, peak_risk_level
               FROM depression_episodes WHERE id = ?""",
            (episode_id,)
        ).fetchone()

        if result:
            total_readings = result[0]
            avg_confidence = result[1]
            peak_risk = result[2]

            # Update rolling average confidence
            new_total = total_readings + 1
            new_avg_confidence = ((avg_confidence * total_readings) + confidence) / new_total

            # Update peak risk level
            new_peak_risk = max(peak_risk, risk_level)

            self.conn.execute(
                """UPDATE depression_episodes
                   SET total_readings = ?,
                       avg_confidence = ?,
                       peak_risk_level = ?
                   WHERE id = ?""",
                (new_total, new_avg_confidence, new_peak_risk, episode_id)
            )
            self.conn.commit()

    def end_depression_episode(self, episode_id: str):
        """Mark a depression episode as ended."""
        self.conn.execute(
            """UPDATE depression_episodes
               SET end_time = CURRENT_TIMESTAMP,
                   is_active = 0
               WHERE id = ?""",
            (episode_id,)
        )
        self.conn.commit()

    def get_user_depression_stats(self, user_id: str) -> dict:
        """Get depression episode statistics for a user."""
        # Total episodes
        total_result = self.conn.execute(
            """SELECT COUNT(*) FROM depression_episodes WHERE user_id = ?""",
            (user_id,)
        ).fetchone()

        # Active episode
        active_result = self.conn.execute(
            """SELECT COUNT(*) FROM depression_episodes
               WHERE user_id = ? AND is_active = 1""",
            (user_id,)
        ).fetchone()

        # Last 7 days episodes
        recent_result = self.conn.execute(
            """SELECT COUNT(*) FROM depression_episodes
               WHERE user_id = ? AND start_time >= datetime('now', '-7 days')""",
            (user_id,)
        ).fetchone()

        # Peak risk level in last 7 days
        peak_result = self.conn.execute(
            """SELECT MAX(peak_risk_level) FROM depression_episodes
               WHERE user_id = ? AND start_time >= datetime('now', '-7 days')""",
            (user_id,)
        ).fetchone()

        # Latest ML prediction
        latest_pred = self.conn.execute(
            """SELECT ml_prediction, ml_confidence, risk_level, recorded_at
               FROM wearable_data
               WHERE user_id = ? AND ml_prediction IS NOT NULL
               ORDER BY recorded_at DESC LIMIT 1""",
            (user_id,)
        ).fetchone()

        return {
            "total_episodes": total_result[0] if total_result else 0,
            "has_active_episode": (active_result[0] > 0) if active_result else False,
            "episodes_last_7_days": recent_result[0] if recent_result else 0,
            "peak_risk_last_7_days": peak_result[0] if (peak_result and peak_result[0]) else 0,
            "latest_prediction": {
                "prediction": latest_pred[0],
                "confidence": latest_pred[1],
                "risk_level": latest_pred[2],
                "timestamp": latest_pred[3]
            } if latest_pred else None
        }

    def get_all_depression_episodes(self, user_id: str, limit: int = 50) -> list:
        """Get all depression episodes for a user."""
        results = self.conn.execute(
            """SELECT id, start_time, end_time, peak_risk_level, total_readings, avg_confidence, is_active
               FROM depression_episodes
               WHERE user_id = ?
               ORDER BY start_time DESC
               LIMIT ?""",
            (user_id, limit)
        ).fetchall()

        return [
            {
                "id": r[0],
                "start_time": r[1],
                "end_time": r[2],
                "peak_risk_level": r[3],
                "total_readings": r[4],
                "avg_confidence": r[5],
                "is_active": bool(r[6])
            }
            for r in results
        ]

    def get_ml_prediction_history(self, user_id: str, limit: int = 100) -> list:
        """Get history of ML predictions for a user."""
        results = self.conn.execute(
            """SELECT ml_prediction, ml_confidence, risk_level, recorded_at
               FROM wearable_data
               WHERE user_id = ? AND ml_prediction IS NOT NULL
               ORDER BY recorded_at DESC
               LIMIT ?""",
            (user_id, limit)
        ).fetchall()

        return [
            {
                "prediction": r[0],
                "confidence": r[1],
                "risk_level": r[2],
                "timestamp": r[3]
            }
            for r in results
        ]


    # ==================== ML WINDOW PREDICTION OPERATIONS ====================

    def save_window_prediction(self, user_id: str, prediction: str,
                               confidence: float, risk_level: int,
                               readings_used: int = 25) -> str:
        """Save one ML inference result (covers a window of readings)."""
        record_id = str(uuid.uuid4())
        self.conn.execute(
            """INSERT INTO ml_window_predictions
               (id, user_id, prediction, risk_level, confidence, readings_used)
               VALUES (?, ?, ?, ?, ?, ?)""",
            (record_id, user_id, prediction, risk_level,
             round(confidence, 4), readings_used)
        )
        self.conn.commit()
        return record_id

    def get_window_predictions(self, user_id: str, limit: int = 200) -> list:
        """Get ML window prediction history for a user (newest first)."""
        results = self.conn.execute(
            """SELECT id, prediction, risk_level, confidence, readings_used, predicted_at
               FROM ml_window_predictions
               WHERE user_id = ?
               ORDER BY predicted_at DESC
               LIMIT ?""",
            (user_id, limit)
        ).fetchall()

        return [
            {
                "id": r[0],
                "prediction": r[1],
                "risk_level": r[2],
                "confidence": r[3],
                "readings_used": r[4],
                "predicted_at": r[5]
            }
            for r in reversed(results)  # chronological for charting
        ]

    def delete_user(self, user_id: str) -> bool:
        """Delete a user and all their associated data (cascading)."""
        try:
            # Delete child records first, then the user row
            child_tables = [
                "messages",
                "beck_sessions",
                "exercises_completed",
                "crisis_flags",
                "user_stats",
                "wearable_data",
                "device_keys",
                "depression_episodes",
                "window_predictions",
            ]
            for table in child_tables:
                try:
                    self.conn.execute(f"DELETE FROM {table} WHERE user_id = ?", (user_id,))
                except Exception:
                    pass  # table may not exist or may use a different FK

            # messages / beck_sessions may be linked via session_id
            try:
                session_ids = self.conn.execute(
                    "SELECT id FROM sessions WHERE user_id = ?", (user_id,)
                ).fetchall()
                for (sid,) in session_ids:
                    self.conn.execute("DELETE FROM messages     WHERE session_id = ?", (sid,))
                    self.conn.execute("DELETE FROM beck_sessions WHERE session_id = ?", (sid,))
            except Exception:
                pass

            self.conn.execute("DELETE FROM sessions WHERE user_id = ?", (user_id,))
            self.conn.execute("DELETE FROM users    WHERE id = ?",      (user_id,))
            self.conn.commit()
            return True
        except Exception as e:
            print(f"Error deleting user {user_id}: {e}")
            return False


# Singleton instance
_db_instance = None

def get_db() -> Database:
    """Get database singleton instance."""
    global _db_instance
    if _db_instance is None:
        _db_instance = Database()
    return _db_instance