MosswartOverlord/main.py

"""
main.py - FastAPI-based telemetry server for Dereth Tracker.

This service ingests real-time position and event data from plugin clients via WebSockets,
stores telemetry and statistics in a TimescaleDB backend, and exposes HTTP and WebSocket
endpoints for browser clients to retrieve live and historical data, trails, and per-character stats.
"""

from datetime import datetime, timedelta, timezone
import json
import logging
import os
import sys
import time
from typing import Dict, List, Any
from pathlib import Path
import asyncio
import socket
import struct

from fastapi import (
    FastAPI,
    Header,
    HTTPException,
    Query,
    WebSocket,
    WebSocketDisconnect,
    Request,
)
from fastapi.responses import JSONResponse, Response, StreamingResponse
from fastapi.routing import APIRoute
from fastapi.staticfiles import StaticFiles
from fastapi.encoders import jsonable_encoder
from pydantic import BaseModel
from typing import Optional
import httpx

# Async database support
from sqlalchemy.dialects.postgresql import insert as pg_insert
from db_async import (
    database,
    telemetry_events,
    char_stats,
    rare_stats,
    rare_stats_sessions,
    spawn_events,
    rare_events,
    character_inventories,
    character_stats,
    portals,
    server_health_checks,
    server_status,
    init_db_async,
    cleanup_old_portals,
)
import asyncio

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
    handlers=[
        logging.StreamHandler(sys.stdout),
    ],
)
logger = logging.getLogger(__name__)

# Get log level from environment (DEBUG, INFO, WARNING, ERROR)
log_level = os.getenv("LOG_LEVEL", "INFO").upper()
logger.setLevel(getattr(logging, log_level, logging.INFO))

# Inventory service configuration
INVENTORY_SERVICE_URL = os.getenv(
    "INVENTORY_SERVICE_URL", "http://inventory-service:8000"
)
# In-memory caches for REST endpoints
_cached_live: dict = {"players": []}
_cached_trails: dict = {"trails": []}
_cached_total_rares: dict = {"all_time": 0, "today": 0, "last_updated": None}
_cached_total_kills: dict = {"total": 0, "last_updated": None}
_cache_task: asyncio.Task | None = None
_rares_cache_task: asyncio.Task | None = None
_cleanup_task: asyncio.Task | None = None
_broadcast_tasks: set[asyncio.Task] = set()

# Player tracking for debug purposes
_player_history: list = []  # List of player sets from last 10 refreshes
_player_events: list = []  # List of player enter/exit events
_max_history_size = 10  # Keep last 10 player sets
_max_events_size = 100  # Keep last 100 events

# Telemetry timing tracking for debug purposes
_player_telemetry_times: dict = {}  # character_name -> list of timestamps
_max_telemetry_history = 20  # Keep last 20 telemetry timestamps per player

# Simple WebSocket connection counters (Phase 1)
_plugin_connections = 0
_browser_connections = 0

# Simple database query performance counters (Phase 2)
_total_queries = 0
_total_query_time = 0.0

# Simple recent activity tracking (Phase 3)
_recent_telemetry_messages = []
_max_recent_messages = 50

# Server health monitoring
_server_health_task = None
_server_status_cache = {
    "status": "unknown",
    "latency_ms": None,
    "player_count": None,
    "last_check": None,
    "uptime_seconds": 0,
    "last_restart": None,
}

# Quest status cache - stores last received quest data per player
# Structure: {character_name: {quest_name: countdown_value}}
_quest_status_cache: Dict[str, Dict[str, str]] = {}


# AC Hash32 checksum algorithm (based on ThwargLauncher)
def calculate_hash32(data: bytes) -> int:
    """Calculate AC Hash32 checksum as used in ThwargLauncher."""
    length = len(data)
    checksum = (length << 16) & 0xFFFFFFFF

    # Process 4-byte chunks
    for i in range(0, length - 3, 4):
        chunk = struct.unpack("<I", data[i : i + 4])[0]
        checksum = (checksum + chunk) & 0xFFFFFFFF

    # Handle remaining bytes
    remaining_start = (length // 4) * 4
    shift = 24
    for i in range(remaining_start, length):
        byte_val = data[i] << shift
        checksum = (checksum + byte_val) & 0xFFFFFFFF
        shift -= 8

    return checksum


# Create AC EchoRequest packet for server health check (based on ThwargLauncher)
def create_echo_request_packet():
    """Create an AC EchoRequest packet for server health checking."""
    # AC packet header: sequence(4) + flags(4) + checksum(4) + id(2) + time(2) + size(2) + table(2) = 20 bytes + padding
    packet = bytearray(32)  # 32 bytes total (0x20)

    # Sequence (4 bytes) - can be 0
    struct.pack_into("<I", packet, 0, 0)

    # Flags (4 bytes) - EchoRequest = 0x02000000
    struct.pack_into("<I", packet, 4, 0x02000000)

    # Temporary checksum (4 bytes) - required for proper checksum calculation
    struct.pack_into("<I", packet, 8, 0x0BADD70D)

    # ID (2 bytes) - can be 0
    struct.pack_into("<H", packet, 12, 0)

    # Time (2 bytes) - can be 0
    struct.pack_into("<H", packet, 14, 0)

    # Size (2 bytes) - header size = 32 (0x20)
    struct.pack_into("<H", packet, 16, 32)

    # Table (2 bytes) - can be 0
    struct.pack_into("<H", packet, 18, 0)

    # Calculate proper AC Hash32 checksum
    # First, set checksum field to 0
    struct.pack_into("<I", packet, 8, 0)

    # Calculate checksum using Hash32 algorithm
    checksum = calculate_hash32(bytes(packet))
    struct.pack_into("<I", packet, 8, checksum)

    return bytes(packet)


AC_ECHO_PACKET = create_echo_request_packet()

# AC login packet for server health check (same as ThwargLauncher MakeLoginPacket)
AC_LOGIN_PACKET = bytes(
    [
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x01,
        0x00,
        0x93,
        0x00,
        0xD0,
        0x05,
        0x00,
        0x00,
        0x00,
        0x00,
        0x40,
        0x00,
        0x00,
        0x00,
        0x04,
        0x00,
        0x31,
        0x38,
        0x30,
        0x32,
        0x00,
        0x00,
        0x34,
        0x00,
        0x00,
        0x00,
        0x01,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x3E,
        0xB8,
        0xA8,
        0x58,
        0x1C,
        0x00,
        0x61,
        0x63,
        0x73,
        0x65,
        0x72,
        0x76,
        0x65,
        0x72,
        0x74,
        0x72,
        0x61,
        0x63,
        0x6B,
        0x65,
        0x72,
        0x3A,
        0x6A,
        0x6A,
        0x39,
        0x68,
        0x32,
        0x36,
        0x68,
        0x63,
        0x73,
        0x67,
        0x67,
        0x63,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
        0x00,
    ]
)


async def check_server_health(
    address: str, port: int, timeout: float = 3.0
) -> tuple[bool, float, int]:
    """Check AC server health via UDP packet with retry logic.

    Retries 6 times with 5-second delays before declaring server down.
    Returns: (is_up, latency_ms, player_count)
    """
    max_retries = 6
    retry_delay = 5.0

    for attempt in range(max_retries):
        logger.debug(
            f"🔍 Health check attempt {attempt + 1}/{max_retries} for {address}:{port}"
        )
        start_time = time.time()
        sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
        sock.setblocking(False)

        try:
            # Send login packet (same as ThwargLauncher)
            await asyncio.get_event_loop().sock_sendto(
                sock, AC_LOGIN_PACKET, (address, port)
            )

            # Wait for response with timeout
            try:
                data, addr = await asyncio.wait_for(
                    asyncio.get_event_loop().sock_recvfrom(sock, 1024), timeout=timeout
                )

                latency_ms = (time.time() - start_time) * 1000
                logger.debug(
                    f"📥 Received response from {addr}: {len(data)} bytes, latency: {latency_ms:.1f}ms"
                )

                # Check if valid response (support both TimeSynch 0x800000 and ConnectRequest 0x40000)
                if len(data) >= 24:
                    flags = struct.unpack("<I", data[4:8])[0]

                    # Accept both TimeSynch (0x800000) and ConnectRequest (0x40000) as valid responses
                    if (flags & 0x800000) or (flags & 0x40000):
                        # UDP health check is for server status and latency only
                        # Player count comes from TreeStats.net API (like ThwargLauncher)
                        logger.debug(
                            f"✅ Valid server response: latency: {latency_ms:.1f}ms"
                        )
                        return True, latency_ms, None

                # Any response indicates server is up, even if not the expected format
                logger.info(
                    f"✅ Server response (non-standard format): latency: {latency_ms:.1f}ms"
                )
                return True, latency_ms, None

            except asyncio.TimeoutError:
                logger.debug(
                    f"⏰ TIMEOUT: No response from {address}:{port} after {timeout}s"
                )
                if attempt < max_retries - 1:
                    logger.debug(f"Retrying in {retry_delay} seconds...")
                    await asyncio.sleep(retry_delay)
                continue

        except Exception as e:
            logger.error(f"Server health check error on attempt {attempt + 1}: {e}")
            if attempt < max_retries - 1:
                await asyncio.sleep(retry_delay)
            continue
        finally:
            sock.close()

    # Only declare down after all retries fail
    logger.warning(
        f"❌ Server {address}:{port} is DOWN after {max_retries} attempts over {max_retries * retry_delay} seconds"
    )
    return False, None, None


async def get_player_count_from_treestats(server_name: str) -> int:
    """Get player count from TreeStats.net API (same as ThwargLauncher)."""
    try:
        async with httpx.AsyncClient() as client:
            response = await client.get(
                "http://treestats.net/player_counts-latest.json", timeout=10
            )
            if response.status_code == 200:
                data = response.json()
                for server_data in data:
                    if server_data.get("server") == server_name:
                        return server_data.get("count", 0)
        return 0
    except Exception as e:
        logger.debug(f"Failed to get player count from TreeStats.net: {e}")
        return 0


async def monitor_server_health():
    """Background task to monitor server health every 30 seconds and cleanup old portals every minute."""
    server_name = "Coldeve"
    server_address = "play.coldeve.ac"
    server_port = 9000
    check_interval = 30  # seconds
    player_count_interval = (
        300  # 5 minutes (like ThwargLauncher's 10 minutes, but more frequent)
    )
    portal_cleanup_interval = 60  # 1 minute
    last_player_count_check = 0
    last_portal_cleanup = 0
    current_player_count = None

    # Initialize server status in database
    try:
        existing = await database.fetch_one(
            "SELECT * FROM server_status WHERE server_name = :name",
            {"name": server_name},
        )
        if not existing:
            await database.execute(
                server_status.insert().values(
                    server_name=server_name,
                    current_status="unknown",
                    total_uptime_seconds=0,
                )
            )
    except Exception as e:
        logger.error(f"Failed to initialize server status: {e}")

    while True:
        try:
            logger.debug(
                f"🏥 Running scheduled health check for {server_name} ({server_address}:{server_port})"
            )
            # Check server health via UDP (for status and latency)
            is_up, latency_ms, _ = await check_server_health(
                server_address, server_port
            )
            status = "up" if is_up else "down"
            now = datetime.now(timezone.utc)

            # Get player count from TreeStats.net API (like ThwargLauncher)
            current_time = time.time()
            if (
                current_time - last_player_count_check >= player_count_interval
                or current_player_count is None
            ):
                new_player_count = await get_player_count_from_treestats(server_name)
                if new_player_count > 0:  # Only update if we got a valid count
                    current_player_count = new_player_count
                    last_player_count_check = current_time
                    logger.info(
                        f"🏥 Updated player count from TreeStats.net: {current_player_count}"
                    )

            logger.debug(
                f"🏥 Health check result: {status}, latency: {latency_ms}, players: {current_player_count}"
            )

            # Record health check
            await database.execute(
                server_health_checks.insert().values(
                    server_name=server_name,
                    server_address=f"{server_address}:{server_port}",
                    timestamp=now,
                    status=status,
                    latency_ms=latency_ms,
                    player_count=current_player_count,
                )
            )

            # Get previous status
            prev_status = await database.fetch_one(
                "SELECT * FROM server_status WHERE server_name = :name",
                {"name": server_name},
            )

            # Calculate uptime and detect restarts
            last_restart = prev_status["last_restart"] if prev_status else None

            if (
                prev_status
                and prev_status["current_status"] == "down"
                and status == "up"
            ):
                # Server came back up - this is a restart
                last_restart = now
                logger.info(f"Server {server_name} came back online")
                # Broadcast to all browser clients
                await _broadcast_to_browser_clients(
                    {
                        "type": "server_status",
                        "server": server_name,
                        "status": "up",
                        "message": "Server is back online",
                    }
                )

            # Calculate uptime from last restart time (not accumulated)
            if last_restart and status == "up":
                uptime_seconds = int((now - last_restart).total_seconds())
            else:
                uptime_seconds = 0

            # Update server status (always include current_player_count if we have it)
            await database.execute(
                """
                INSERT INTO server_status (server_name, current_status, last_seen_up, last_restart, 
                                         total_uptime_seconds, last_check, last_latency_ms, last_player_count)
                VALUES (:name, :status, :last_seen, :restart, :uptime, :check, :latency, :players)
                ON CONFLICT (server_name) DO UPDATE SET
                    current_status = :status,
                    last_seen_up = CASE WHEN :status = 'up' THEN :last_seen ELSE server_status.last_seen_up END,
                    last_restart = :restart,
                    total_uptime_seconds = :uptime,
                    last_check = :check,
                    last_latency_ms = :latency,
                    last_player_count = CASE WHEN :players IS NOT NULL THEN :players ELSE server_status.last_player_count END
                """,
                {
                    "name": server_name,
                    "status": status,
                    "last_seen": now if status == "up" else None,
                    "restart": last_restart,
                    "uptime": uptime_seconds,
                    "check": now,
                    "latency": latency_ms,
                    "players": current_player_count,
                },
            )

            # Update cache
            global _server_status_cache
            _server_status_cache = {
                "status": status,
                "latency_ms": latency_ms,
                "player_count": current_player_count,
                "last_check": now.isoformat(),
                "uptime_seconds": uptime_seconds,
                "last_restart": last_restart.isoformat() if last_restart else None,
            }

            logger.debug(
                f"Server health check: {status}, latency={latency_ms}ms, players={current_player_count}"
            )

            # Portal cleanup (run every minute)
            current_time = time.time()
            if current_time - last_portal_cleanup >= portal_cleanup_interval:
                try:
                    deleted_count = await cleanup_old_portals()
                    logger.info(
                        f"Portal cleanup: removed {deleted_count} portals older than 1 hour"
                    )
                    last_portal_cleanup = current_time
                except Exception as cleanup_error:
                    logger.error(
                        f"Portal cleanup error: {cleanup_error}", exc_info=True
                    )

        except Exception as e:
            logger.error(f"Server health monitoring error: {e}", exc_info=True)

        await asyncio.sleep(check_interval)


async def cleanup_connections_loop():
    """Background task to clean up stale WebSocket connections every 5 minutes."""
    cleanup_interval = 300  # 5 minutes

    logger.info("🧹 Starting WebSocket connection cleanup task")

    while True:
        try:
            await asyncio.sleep(cleanup_interval)
            logger.debug("🧹 Running periodic WebSocket connection cleanup")
            await cleanup_stale_connections()
        except Exception as e:
            logger.error(f"WebSocket cleanup task error: {e}", exc_info=True)


def _track_player_changes(new_players: list) -> None:
    """Track player changes for debugging flapping issues."""
    from datetime import datetime, timezone

    # Get current player names
    current_players = {p["character_name"] for p in new_players}
    timestamp = datetime.now(timezone.utc)

    # Track telemetry timing for each player
    for player_data in new_players:
        player_name = player_data["character_name"]
        player_timestamp = player_data.get("timestamp")

        # Convert timestamp if it's a string
        if isinstance(player_timestamp, str):
            try:
                player_timestamp = datetime.fromisoformat(
                    player_timestamp.replace("Z", "+00:00")
                )
            except:
                player_timestamp = timestamp
        elif player_timestamp is None:
            player_timestamp = timestamp

        # Initialize player telemetry tracking if needed
        if player_name not in _player_telemetry_times:
            _player_telemetry_times[player_name] = []

        # Add this telemetry timestamp
        _player_telemetry_times[player_name].append(player_timestamp)

        # Trim to max history
        if len(_player_telemetry_times[player_name]) > _max_telemetry_history:
            _player_telemetry_times[player_name].pop(0)

    # Get previous player names if we have history
    previous_players = set()
    if _player_history:
        previous_players = {p["character_name"] for p in _player_history[-1]["players"]}

    # Find players who entered and exited
    entered_players = current_players - previous_players
    exited_players = previous_players - current_players

    # Log events with telemetry timing analysis
    for player in entered_players:
        # Check if this is due to timing gap
        timing_gap = None
        if (
            player in _player_telemetry_times
            and len(_player_telemetry_times[player]) >= 2
        ):
            last_two = _player_telemetry_times[player][-2:]
            timing_gap = (last_two[1] - last_two[0]).total_seconds()

        event = {
            "timestamp": timestamp,
            "type": "enter",
            "character_name": player,
            "total_players": len(current_players),
            "timing_gap": timing_gap,
        }
        _player_events.append(event)
        gap_info = (
            f" (gap: {timing_gap:.1f}s)" if timing_gap and timing_gap > 25 else ""
        )
        logger.debug(
            f"Player entered: {player} (total: {len(current_players)}){gap_info}"
        )

    for player in exited_players:
        # Calculate time since last telemetry
        last_telemetry_age = None
        if player in _player_telemetry_times and _player_telemetry_times[player]:
            last_telemetry = _player_telemetry_times[player][-1]
            last_telemetry_age = (timestamp - last_telemetry).total_seconds()

        event = {
            "timestamp": timestamp,
            "type": "exit",
            "character_name": player,
            "total_players": len(current_players),
            "last_telemetry_age": last_telemetry_age,
        }
        _player_events.append(event)
        age_info = (
            f" (last telemetry: {last_telemetry_age:.1f}s ago)"
            if last_telemetry_age
            else ""
        )
        logger.debug(
            f"Player exited: {player} (total: {len(current_players)}){age_info}"
        )

    # Add current state to history
    history_entry = {
        "timestamp": timestamp,
        "players": new_players,
        "player_count": len(new_players),
        "player_names": list(current_players),
    }
    _player_history.append(history_entry)

    # Trim history to max size
    if len(_player_history) > _max_history_size:
        _player_history.pop(0)

    # Trim events to max size
    if len(_player_events) > _max_events_size:
        _player_events.pop(0)


def _analyze_flapping_patterns() -> dict:
    """Analyze player events to identify flapping patterns."""
    from collections import Counter, defaultdict

    if not _player_events:
        return {
            "flapping_players": [],
            "frequent_events": [],
            "analysis": "No events to analyze",
        }

    # Count events per player
    player_event_counts = Counter()
    player_flap_counts = defaultdict(int)

    # Track recent activity per player (last 10 events)
    recent_player_activity = defaultdict(list)

    for event in _player_events[-50:]:  # Analyze last 50 events
        player = event["character_name"]
        event_type = event["type"]
        player_event_counts[player] += 1
        recent_player_activity[player].append(event_type)

    # Identify flapping players (players with many enter/exit cycles)
    flapping_players = []
    for player, activity in recent_player_activity.items():
        if len(activity) >= 4:  # At least 4 events
            # Count alternating enter/exit patterns
            flap_score = 0
            for i in range(1, len(activity)):
                if activity[i] != activity[i - 1]:  # Different from previous
                    flap_score += 1

            if flap_score >= 3:  # At least 3 transitions
                flapping_players.append(
                    {
                        "character_name": player,
                        "events": len(activity),
                        "flap_score": flap_score,
                        "recent_activity": activity[-10:],  # Last 10 events
                    }
                )

    # Sort by flap score
    flapping_players.sort(key=lambda x: x["flap_score"], reverse=True)

    # Most active players
    frequent_events = [
        {"character_name": player, "event_count": count}
        for player, count in player_event_counts.most_common(10)
    ]

    # Recent activity summary
    recent_enters = sum(1 for e in _player_events[-20:] if e["type"] == "enter")
    recent_exits = sum(1 for e in _player_events[-20:] if e["type"] == "exit")

    return {
        "flapping_players": flapping_players,
        "frequent_events": frequent_events,
        "recent_activity": {
            "enters": recent_enters,
            "exits": recent_exits,
            "net_change": recent_enters - recent_exits,
        },
        "analysis": f"Found {len(flapping_players)} potentially flapping players",
    }


def _analyze_telemetry_timing() -> dict:
    """Analyze telemetry timing patterns for all players."""
    from datetime import datetime, timezone

    timing_analysis = {}
    problem_players = []

    for player_name, timestamps in _player_telemetry_times.items():
        if len(timestamps) < 2:
            continue

        # Calculate intervals between telemetry messages
        intervals = []
        for i in range(1, len(timestamps)):
            interval = (timestamps[i] - timestamps[i - 1]).total_seconds()
            intervals.append(interval)

        if not intervals:
            continue

        # Calculate timing statistics
        avg_interval = sum(intervals) / len(intervals)
        min_interval = min(intervals)
        max_interval = max(intervals)

        # Count problematic intervals (>30s)
        long_gaps = [i for i in intervals if i > 30]
        recent_long_gaps = [i for i in intervals[-5:] if i > 30]  # Last 5 intervals

        # Determine if this player has timing issues
        has_timing_issues = len(long_gaps) > 0 or max_interval > 35

        timing_stats = {
            "character_name": player_name,
            "total_messages": len(timestamps),
            "avg_interval": round(avg_interval, 1),
            "min_interval": round(min_interval, 1),
            "max_interval": round(max_interval, 1),
            "long_gaps_count": len(long_gaps),
            "recent_long_gaps": len(recent_long_gaps),
            "last_message_age": (
                datetime.now(timezone.utc) - timestamps[-1]
            ).total_seconds()
            if timestamps
            else 0,
            "has_timing_issues": has_timing_issues,
            "recent_intervals": [
                round(i, 1) for i in intervals[-5:]
            ],  # Last 5 intervals
        }

        timing_analysis[player_name] = timing_stats

        if has_timing_issues:
            problem_players.append(timing_stats)

    # Sort problem players by severity (max interval)
    problem_players.sort(key=lambda x: x["max_interval"], reverse=True)

    return {
        "all_players": timing_analysis,
        "problem_players": problem_players,
        "summary": {
            "total_tracked_players": len(timing_analysis),
            "players_with_issues": len(problem_players),
            "avg_intervals": [
                stats["avg_interval"] for stats in timing_analysis.values()
            ],
        },
    }


async def _refresh_cache_loop() -> None:
    """Background task: refresh `/live` and `/trails` caches every 5 seconds."""
    consecutive_failures = 0
    max_consecutive_failures = 5

    while True:
        try:
            # Recompute live players (last 30s)
            cutoff = datetime.now(timezone.utc) - ACTIVE_WINDOW
            sql_live = """
                SELECT sub.*,
                       COALESCE(rs.total_rares, 0)      AS total_rares,
                       COALESCE(rss.session_rares, 0)  AS session_rares,
                       COALESCE(cs.total_kills, 0)     AS total_kills
                FROM (
                    SELECT DISTINCT ON (character_name) *
                    FROM telemetry_events
                    WHERE timestamp > :cutoff
                    ORDER BY character_name, timestamp DESC
                ) sub
                LEFT JOIN rare_stats rs
                  ON sub.character_name = rs.character_name
                LEFT JOIN rare_stats_sessions rss
                  ON sub.character_name = rss.character_name
                 AND sub.session_id    = rss.session_id
                LEFT JOIN char_stats cs
                  ON sub.character_name = cs.character_name
            """

            # Use a single connection for both queries to reduce connection churn
            async with database.connection() as conn:
                rows = await conn.fetch_all(sql_live, {"cutoff": cutoff})
                new_players = [dict(r) for r in rows]

                # Track player changes for debugging
                _track_player_changes(new_players)

                _cached_live["players"] = new_players

                # Recompute trails (last 600s)
                cutoff2 = datetime.utcnow().replace(tzinfo=timezone.utc) - timedelta(
                    seconds=600
                )
                sql_trail = """
                    SELECT timestamp, character_name, ew, ns, z
                    FROM telemetry_events
                    WHERE timestamp >= :cutoff
                    ORDER BY character_name, timestamp
                """
                rows2 = await conn.fetch_all(sql_trail, {"cutoff": cutoff2})
                _cached_trails["trails"] = [
                    {
                        "timestamp": r["timestamp"],
                        "character_name": r["character_name"],
                        "ew": r["ew"],
                        "ns": r["ns"],
                        "z": r["z"],
                    }
                    for r in rows2
                ]

            # Reset failure counter on success
            consecutive_failures = 0
            logger.debug(
                f"Cache refreshed: {len(_cached_live['players'])} players, {len(_cached_trails['trails'])} trail points"
            )

        except Exception as e:
            consecutive_failures += 1
            logger.error(
                f"Cache refresh failed ({consecutive_failures}/{max_consecutive_failures}): {e}",
                exc_info=True,
            )

            # If too many consecutive failures, wait longer and try to reconnect
            if consecutive_failures >= max_consecutive_failures:
                logger.warning(
                    f"Too many consecutive cache refresh failures. Attempting database reconnection..."
                )
                try:
                    await database.disconnect()
                    await asyncio.sleep(2)
                    await database.connect()
                    logger.info("Database reconnected successfully")
                    consecutive_failures = 0
                except Exception as reconnect_error:
                    logger.error(f"Database reconnection failed: {reconnect_error}")
                    await asyncio.sleep(10)  # Wait longer before retrying
                    continue

        await asyncio.sleep(5)


async def _refresh_total_rares_cache() -> None:
    """Background task: refresh total rares cache every 5 minutes."""
    consecutive_failures = 0
    max_consecutive_failures = 3

    while True:
        try:
            async with database.connection() as conn:
                # Get all-time total rares (sum of all characters) - gracefully handle missing table
                try:
                    all_time_query = (
                        "SELECT COALESCE(SUM(total_rares), 0) as total FROM rare_stats"
                    )
                    all_time_result = await conn.fetch_one(all_time_query)
                    all_time_total = all_time_result["total"] if all_time_result else 0
                except Exception as e:
                    logger.debug(f"rare_stats table not available: {e}")
                    all_time_total = 0

                # Get today's rares from rare_events table - gracefully handle missing table
                try:
                    today_query = """
                        SELECT COUNT(*) as today_count 
                        FROM rare_events 
                        WHERE timestamp >= CURRENT_DATE
                    """
                    today_result = await conn.fetch_one(today_query)
                    today_total = today_result["today_count"] if today_result else 0
                except Exception as e:
                    logger.debug(f"rare_events table not available or empty: {e}")
                    today_total = 0

                # Get total kills from char_stats table (all-time, all characters)
                try:
                    kills_query = (
                        "SELECT COALESCE(SUM(total_kills), 0) as total FROM char_stats"
                    )
                    kills_result = await conn.fetch_one(kills_query)
                    total_kills = kills_result["total"] if kills_result else 0
                except Exception as e:
                    logger.debug(f"char_stats table not available: {e}")
                    total_kills = 0

                # Update caches
                _cached_total_rares["all_time"] = all_time_total
                _cached_total_rares["today"] = today_total
                _cached_total_rares["last_updated"] = datetime.now(timezone.utc)

                _cached_total_kills["total"] = total_kills
                _cached_total_kills["last_updated"] = datetime.now(timezone.utc)

                consecutive_failures = 0
                logger.debug(
                    f"Stats cache updated: Rares all-time: {all_time_total}, today: {today_total}, Kills: {total_kills}"
                )

        except Exception as e:
            consecutive_failures += 1
            logger.error(
                f"Total rares cache refresh failed ({consecutive_failures}/{max_consecutive_failures}): {e}",
                exc_info=True,
            )

            if consecutive_failures >= max_consecutive_failures:
                logger.warning(
                    "Too many consecutive total rares cache failures, waiting longer..."
                )
                await asyncio.sleep(60)  # Wait longer on repeated failures
                continue

        # Sleep for 5 minutes (300 seconds)
        await asyncio.sleep(300)


# ------------------------------------------------------------------
app = FastAPI()
# In-memory store mapping character_name to the most recent telemetry snapshot
live_snapshots: Dict[str, dict] = {}
live_vitals: Dict[str, dict] = {}
live_character_stats: Dict[str, dict] = {}
live_equipment_cantrip_states: Dict[str, dict] = {}

# Shared secret used to authenticate plugin WebSocket connections (override for production)
SHARED_SECRET = "your_shared_secret"
# LOG_FILE      = "telemetry_log.jsonl"
# ------------------------------------------------------------------
ACTIVE_WINDOW = timedelta(
    seconds=30
)  # Time window defining “online” players (last 30 seconds)

"""
Data models for plugin events:
  - TelemetrySnapshot: periodic telemetry data from a player client
  - SpawnEvent: information about a mob spawn event
  - RareEvent: details of a rare mob event
"""


class TelemetrySnapshot(BaseModel):
    character_name: str
    char_tag: Optional[str] = None
    session_id: str
    timestamp: datetime

    ew: float  # +E / –W
    ns: float  # +N / –S
    z: float

    kills: int
    kills_per_hour: Optional[float] = None
    onlinetime: Optional[str] = None
    deaths: int
    total_deaths: Optional[int] = None
    # Removed from telemetry payload; always enforced to 0 and tracked via rare events
    rares_found: Optional[int] = 0
    prismatic_taper_count: int
    vt_state: str
    # Optional telemetry metrics
    mem_mb: Optional[float] = None
    cpu_pct: Optional[float] = None
    mem_handles: Optional[int] = None
    latency_ms: Optional[float] = None


class SpawnEvent(BaseModel):
    """
    Model for a spawn event emitted by plugin clients when a mob appears.
    Records character context, mob type, timestamp, and spawn location.
    """

    character_name: str
    mob: str
    timestamp: datetime
    ew: float
    ns: float
    z: float = 0.0


class RareEvent(BaseModel):
    """
    Model for a rare mob event when a player encounters or discovers a rare entity.
    Includes character, event name, timestamp, and location coordinates.
    """

    character_name: str
    name: str
    timestamp: datetime
    ew: float
    ns: float
    z: float = 0.0


class FullInventoryMessage(BaseModel):
    """
    Model for the full_inventory WebSocket message type.
    Contains complete character inventory snapshot with raw item data.
    """

    character_name: str
    timestamp: datetime
    item_count: int
    items: List[Dict[str, Any]]


class VitalsMessage(BaseModel):
    """
    Model for the vitals WebSocket message type.
    Contains character health, stamina, mana, and vitae information.
    """

    character_name: str
    timestamp: datetime
    health_current: int
    health_max: int
    health_percentage: float
    stamina_current: int
    stamina_max: int
    stamina_percentage: float
    mana_current: int
    mana_max: int
    mana_percentage: float
    vitae: int


class CharacterStatsMessage(BaseModel):
    """
    Model for the character_stats WebSocket message type.
    Contains character attributes, skills, allegiance, and progression data.
    Sent by plugin on login and every 10 minutes.
    """

    character_name: str
    timestamp: datetime
    level: Optional[int] = None
    total_xp: Optional[int] = None
    unassigned_xp: Optional[int] = None
    luminance_earned: Optional[int] = None
    luminance_total: Optional[int] = None
    deaths: Optional[int] = None
    race: Optional[str] = None
    gender: Optional[str] = None
    birth: Optional[str] = None
    current_title: Optional[int] = None
    skill_credits: Optional[int] = None
    burden: Optional[int] = None
    burden_units: Optional[int] = None
    encumbrance_capacity: Optional[int] = None
    attributes: Optional[dict] = None
    vitals: Optional[dict] = None
    skills: Optional[dict] = None
    allegiance: Optional[dict] = None
    active_item_enchantments: Optional[list] = None
    properties: Optional[dict] = (
        None  # Dict[int, int] — DWORD properties (augs, ratings, etc.)
    )
    titles: Optional[list] = None  # List[str] — character title names


@app.on_event("startup")
async def on_startup():
    """Event handler triggered when application starts up.

    Attempts to connect to the database with retry logic to accommodate
    potential startup delays (e.g., waiting for Postgres to be ready).
    """
    max_attempts = 5
    for attempt in range(1, max_attempts + 1):
        try:
            await database.connect()
            await init_db_async()
            logger.info(f"Database connected successfully on attempt {attempt}")
            # Log connection pool configuration
            try:
                logger.info(f"Database connection established with pool configuration")
            except Exception as pool_error:
                logger.debug(f"Could not access pool details: {pool_error}")
            break
        except Exception as e:
            logger.warning(
                f"Database connection failed (attempt {attempt}/{max_attempts}): {e}"
            )
            if attempt < max_attempts:
                await asyncio.sleep(5)
    else:
        raise RuntimeError(
            f"Could not connect to database after {max_attempts} attempts"
        )
    # Start background cache refresh (live & trails)
    global _cache_task, _rares_cache_task, _server_health_task, _cleanup_task
    _cache_task = asyncio.create_task(_refresh_cache_loop())
    _rares_cache_task = asyncio.create_task(_refresh_total_rares_cache())
    _server_health_task = asyncio.create_task(monitor_server_health())
    _cleanup_task = asyncio.create_task(cleanup_connections_loop())
    logger.info(
        "Background cache refresh, server monitoring, and connection cleanup tasks started"
    )


@app.on_event("shutdown")
async def on_shutdown():
    """Event handler triggered when application is shutting down.

    Ensures the database connection is closed cleanly.
    """
    # Stop cache refresh tasks
    global _cache_task, _rares_cache_task, _server_health_task, _cleanup_task
    if _cache_task:
        logger.info("Stopping background cache refresh task")
        _cache_task.cancel()
        try:
            await _cache_task
        except asyncio.CancelledError:
            pass

    if _rares_cache_task:
        logger.info("Stopping total rares cache refresh task")
        _rares_cache_task.cancel()
        try:
            await _rares_cache_task
        except asyncio.CancelledError:
            pass

    if _server_health_task:
        logger.info("Stopping server health monitoring task")
        _server_health_task.cancel()
        try:
            await _server_health_task
        except asyncio.CancelledError:
            pass

    if _cleanup_task:
        logger.info("Stopping WebSocket connection cleanup task")
        _cleanup_task.cancel()
        try:
            await _cleanup_task
        except asyncio.CancelledError:
            pass
    # Cancel any in-flight broadcast tasks
    if _broadcast_tasks:
        logger.info(f"Cancelling {len(_broadcast_tasks)} in-flight broadcast tasks")
        for task in _broadcast_tasks:
            task.cancel()
        await asyncio.gather(*_broadcast_tasks, return_exceptions=True)
        _broadcast_tasks.clear()

    logger.info("Disconnecting from database")
    await database.disconnect()


# ------------------------  GET  -----------------------------------
@app.get("/debug")
def debug():
    return {"status": "OK"}


@app.get("/debug/player-flapping")
async def get_player_flapping_debug():
    """Return player tracking data for debugging flapping issues."""
    try:
        # Analyze flapping patterns
        flapping_analysis = _analyze_flapping_patterns()

        # Analyze telemetry timing
        timing_analysis = _analyze_telemetry_timing()

        # Get recent events (last 50)
        recent_events = (
            _player_events[-50:] if len(_player_events) > 50 else _player_events
        )

        # Convert timestamps to ISO format for JSON serialization
        formatted_events = []
        for event in recent_events:
            formatted_event = event.copy()
            formatted_event["timestamp"] = event["timestamp"].isoformat()
            formatted_events.append(formatted_event)

        # Format history
        formatted_history = []
        for entry in _player_history:
            formatted_entry = {
                "timestamp": entry["timestamp"].isoformat(),
                "player_count": entry["player_count"],
                "player_names": entry["player_names"],
            }
            formatted_history.append(formatted_entry)

        # Format timing data for JSON serialization
        formatted_timing = {}
        for player_name, timing_data in timing_analysis["all_players"].items():
            formatted_timing[player_name] = timing_data.copy()
            # Round last_message_age for readability
            formatted_timing[player_name]["last_message_age"] = round(
                timing_data["last_message_age"], 1
            )

        return {
            "current_players": len(_cached_live.get("players", [])),
            "history": formatted_history,
            "recent_events": formatted_events,
            "flapping_analysis": flapping_analysis,
            "timing_analysis": {
                "all_players": formatted_timing,
                "problem_players": timing_analysis["problem_players"],
                "summary": timing_analysis["summary"],
            },
            "tracking_stats": {
                "history_entries": len(_player_history),
                "total_events": len(_player_events),
                "tracked_players": len(_player_telemetry_times),
                "max_history_size": _max_history_size,
                "max_events_size": _max_events_size,
            },
        }
    except Exception as e:
        logger.error(f"Failed to get player flapping debug data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/debug/websocket-health")
async def get_websocket_health():
    """Return simple WebSocket connection counts."""
    try:
        return {
            "plugin_connections": _plugin_connections,
            "browser_connections": _browser_connections,
            "total_connections": _plugin_connections + _browser_connections,
        }
    except Exception as e:
        logger.error(f"Failed to get WebSocket health data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/debug/database-performance")
async def get_database_performance():
    """Return simple database query performance statistics."""
    try:
        avg_query_time = (
            (_total_query_time / _total_queries) if _total_queries > 0 else 0.0
        )
        return {
            "total_queries": _total_queries,
            "total_query_time": round(_total_query_time, 3),
            "average_query_time": round(avg_query_time, 3),
        }
    except Exception as e:
        logger.error(f"Failed to get database performance data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/debug/recent-activity")
async def get_recent_activity():
    """Return recent telemetry activity feed."""
    try:
        return {
            "recent_messages": _recent_telemetry_messages.copy(),
            "total_messages": len(_recent_telemetry_messages),
            "max_messages": _max_recent_messages,
        }
    except Exception as e:
        logger.error(f"Failed to get recent activity data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/server-health")
async def get_server_health():
    """Return current server health status."""
    try:
        # Get latest status from database if cache is stale
        if not _server_status_cache.get("last_check") or (
            datetime.now(timezone.utc)
            - datetime.fromisoformat(
                _server_status_cache["last_check"].replace("Z", "+00:00")
            )
            > timedelta(minutes=2)
        ):
            row = await database.fetch_one(
                "SELECT * FROM server_status WHERE server_name = :name",
                {"name": "Coldeve"},
            )

            if row:
                _server_status_cache.update(
                    {
                        "status": row["current_status"],
                        "latency_ms": row["last_latency_ms"],
                        "player_count": row["last_player_count"],
                        "last_check": row["last_check"].isoformat()
                        if row["last_check"]
                        else None,
                        "uptime_seconds": row["total_uptime_seconds"],
                        "last_restart": row["last_restart"].isoformat()
                        if row["last_restart"]
                        else None,
                    }
                )

        # Format uptime
        uptime_seconds = _server_status_cache.get("uptime_seconds", 0)
        days = uptime_seconds // 86400
        hours = (uptime_seconds % 86400) // 3600
        minutes = (uptime_seconds % 3600) // 60

        uptime_str = (
            f"{days}d {hours}h {minutes}m" if days > 0 else f"{hours}h {minutes}m"
        )

        return {
            "server_name": "Coldeve",
            "status": _server_status_cache.get("status", "unknown"),
            "latency_ms": _server_status_cache.get("latency_ms"),
            "player_count": _server_status_cache.get("player_count"),
            "uptime": uptime_str,
            "uptime_seconds": uptime_seconds,
            "last_restart": _server_status_cache.get("last_restart"),
            "last_check": _server_status_cache.get("last_check"),
        }

    except Exception as e:
        logger.error(f"Failed to get server health data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/quest-status")
async def get_quest_status():
    """Return current cached quest status for all players."""
    try:
        # Return the quest cache with structured data
        return {
            "quest_data": _quest_status_cache,
            "tracked_quests": [
                "Stipend Collection Timer",
                "Blank Augmentation Gem Pickup Timer",
                "Insatiable Eater Jaw",
            ],
            "player_count": len(_quest_status_cache),
        }
    except Exception as e:
        logger.error(f"Failed to get quest status data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/portals")
async def get_portals():
    """Return all active portals (less than 1 hour old)."""
    try:
        # No need for cutoff check - cleanup job handles expiration
        query = """
            SELECT portal_name, ns, ew, z, discovered_at, discovered_by
            FROM portals
            ORDER BY discovered_at DESC
        """

        rows = await database.fetch_all(query)

        portals = []
        for row in rows:
            portal = {
                "portal_name": row["portal_name"],
                "coordinates": {"ns": row["ns"], "ew": row["ew"], "z": row["z"]},
                "discovered_at": row["discovered_at"].isoformat(),
                "discovered_by": row["discovered_by"],
            }
            portals.append(portal)

        return {"portals": portals, "portal_count": len(portals)}

    except Exception as e:
        logger.error(f"Failed to get portals data: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/live", response_model=dict)
@app.get("/live/", response_model=dict)
async def get_live_players():
    """Return cached live telemetry per character."""
    try:
        return JSONResponse(content=jsonable_encoder(_cached_live))
    except Exception as e:
        logger.error(f"Failed to get live players: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


# --- GET Trails ---------------------------------
@app.get("/trails")
@app.get("/trails/")
async def get_trails(
    seconds: int = Query(600, ge=0, description="Lookback window in seconds"),
):
    """Return cached trails (updated every 5 seconds)."""
    try:
        return JSONResponse(content=jsonable_encoder(_cached_trails))
    except Exception as e:
        logger.error(f"Failed to get trails: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/total-rares")
@app.get("/total-rares/")
async def get_total_rares():
    """Return cached total rares statistics (updated every 5 minutes)."""
    try:
        return JSONResponse(content=jsonable_encoder(_cached_total_rares))
    except Exception as e:
        logger.error(f"Failed to get total rares: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/total-kills")
@app.get("/total-kills/")
async def get_total_kills():
    """Return cached total kills statistics (updated every 5 minutes)."""
    try:
        return JSONResponse(content=jsonable_encoder(_cached_total_kills))
    except Exception as e:
        logger.error(f"Failed to get total kills: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


# --- GET Spawn Heat Map Endpoint ---------------------------------
@app.get("/spawns/heatmap")
async def get_spawn_heatmap_data(
    hours: int = Query(
        24, ge=1, le=168, description="Lookback window in hours (1-168)"
    ),
    limit: int = Query(
        10000, ge=100, le=50000, description="Maximum number of spawn points to return"
    ),
):
    """
    Aggregate spawn locations for heat-map visualization.

    Returns spawn event coordinates grouped by location with intensity counts
    for the specified time window.

    Response format:
    {
        "spawn_points": [{"ew": float, "ns": float, "intensity": int}, ...],
        "total_points": int,
        "timestamp": "UTC-ISO"
    }
    """
    try:
        cutoff = datetime.now(timezone.utc) - timedelta(hours=hours)

        # Aggregate spawn events by coordinates within time window
        query = """
            SELECT ew, ns, COUNT(*) AS spawn_count
            FROM spawn_events
            WHERE timestamp >= :cutoff
            GROUP BY ew, ns
            ORDER BY spawn_count DESC
            LIMIT :limit
        """

        rows = await database.fetch_all(query, {"cutoff": cutoff, "limit": limit})

        spawn_points = [
            {
                "ew": float(row["ew"]),
                "ns": float(row["ns"]),
                "intensity": int(row["spawn_count"]),
            }
            for row in rows
        ]

        result = {
            "spawn_points": spawn_points,
            "total_points": len(spawn_points),
            "timestamp": datetime.now(timezone.utc).isoformat(),
            "hours_window": hours,
        }

        logger.debug(
            f"Heat map data: {len(spawn_points)} unique spawn locations from last {hours} hours"
        )
        return JSONResponse(content=jsonable_encoder(result))

    except Exception as e:
        logger.error(f"Heat map query failed: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Spawn heat map query failed")


# --- GET Inventory Endpoints ---------------------------------
@app.get("/inventory/{character_name}")
async def get_character_inventory(character_name: str):
    """Get the complete inventory for a specific character - inventory service only."""
    try:
        async with httpx.AsyncClient(timeout=10.0) as client:
            response = await client.get(
                f"{INVENTORY_SERVICE_URL}/inventory/{character_name}"
            )

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            elif response.status_code == 404:
                raise HTTPException(
                    status_code=404,
                    detail=f"No inventory found for character '{character_name}'",
                )
            else:
                logger.error(
                    f"Inventory service returned {response.status_code} for {character_name}"
                )
                raise HTTPException(status_code=502, detail="Inventory service error")

    except httpx.RequestError as e:
        logger.error(f"Could not reach inventory service: {e}")
        raise HTTPException(status_code=503, detail="Inventory service unavailable")
    except HTTPException:
        raise
    except Exception as e:
        logger.error(
            f"Failed to get inventory for {character_name}: {e}", exc_info=True
        )
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/inventory/{character_name}/search")
async def search_character_inventory(
    character_name: str,
    name: str = Query(None, description="Search by item name (partial match)"),
    object_class: int = Query(None, description="Filter by ObjectClass"),
    min_value: int = Query(None, description="Minimum item value"),
    max_value: int = Query(None, description="Maximum item value"),
    min_burden: int = Query(None, description="Minimum burden"),
    max_burden: int = Query(None, description="Maximum burden"),
):
    """Search and filter inventory items for a character with various criteria."""
    try:
        conditions = ["character_name = :character_name"]
        params = {"character_name": character_name}

        if name:
            conditions.append("name ILIKE :name")
            params["name"] = f"%{name}%"

        if object_class is not None:
            conditions.append("object_class = :object_class")
            params["object_class"] = object_class

        if min_value is not None:
            conditions.append("value >= :min_value")
            params["min_value"] = min_value

        if max_value is not None:
            conditions.append("value <= :max_value")
            params["max_value"] = max_value

        if min_burden is not None:
            conditions.append("burden >= :min_burden")
            params["min_burden"] = min_burden

        if max_burden is not None:
            conditions.append("burden <= :max_burden")
            params["max_burden"] = max_burden

        query = f"""
            SELECT name, icon, object_class, value, burden, has_id_data, item_data, timestamp
            FROM character_inventories 
            WHERE {" AND ".join(conditions)}
            ORDER BY value DESC, name
        """

        rows = await database.fetch_all(query, params)

        items = []
        for row in rows:
            item = dict(row)
            items.append(item)

        return JSONResponse(
            content=jsonable_encoder(
                {
                    "character_name": character_name,
                    "item_count": len(items),
                    "search_criteria": {
                        "name": name,
                        "object_class": object_class,
                        "min_value": min_value,
                        "max_value": max_value,
                        "min_burden": min_burden,
                        "max_burden": max_burden,
                    },
                    "items": items,
                }
            )
        )
    except Exception as e:
        logger.error(
            f"Failed to search inventory for {character_name}: {e}", exc_info=True
        )
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/inventories")
async def list_characters_with_inventories():
    """List all characters that have stored inventories with item counts."""
    try:
        query = """
            SELECT character_name, COUNT(*) as item_count, MAX(timestamp) as last_updated
            FROM character_inventories 
            GROUP BY character_name
            ORDER BY last_updated DESC
        """
        rows = await database.fetch_all(query)

        characters = []
        for row in rows:
            characters.append(
                {
                    "character_name": row["character_name"],
                    "item_count": row["item_count"],
                    "last_updated": row["last_updated"],
                }
            )

        return JSONResponse(
            content=jsonable_encoder(
                {"characters": characters, "total_characters": len(characters)}
            )
        )
    except Exception as e:
        logger.error(f"Failed to list inventory characters: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


# --- Inventory Service Character List Proxy ---------------------
@app.get("/inventory-characters")
async def get_inventory_characters():
    """Get character list from inventory service - proxy to avoid routing conflicts."""
    try:
        async with httpx.AsyncClient(timeout=10.0) as client:
            response = await client.get(f"{INVENTORY_SERVICE_URL}/characters/list")

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            else:
                logger.error(
                    f"Inventory service returned {response.status_code}: {response.text}"
                )
                raise HTTPException(
                    status_code=response.status_code,
                    detail="Failed to get characters from inventory service",
                )

    except Exception as e:
        logger.error(
            f"Failed to proxy inventory characters request: {e}", exc_info=True
        )
        raise HTTPException(
            status_code=500, detail="Failed to get inventory characters"
        )


# --- Inventory Search Service Proxy Endpoints -------------------
@app.get("/search/items")
async def search_items_proxy(
    text: str = Query(
        None, description="Search item names, descriptions, or properties"
    ),
    character: str = Query(None, description="Limit search to specific character"),
    include_all_characters: bool = Query(
        False, description="Search across all characters"
    ),
    equipment_status: str = Query(None, description="equipped, unequipped, or all"),
    equipment_slot: int = Query(None, description="Equipment slot mask"),
    # Item category filtering
    armor_only: bool = Query(False, description="Show only armor items"),
    jewelry_only: bool = Query(False, description="Show only jewelry items"),
    weapon_only: bool = Query(False, description="Show only weapon items"),
    # Spell filtering
    has_spell: str = Query(None, description="Must have this specific spell (by name)"),
    spell_contains: str = Query(None, description="Spell name contains this text"),
    legendary_cantrips: str = Query(
        None, description="Comma-separated list of legendary cantrip names"
    ),
    # Combat properties
    min_damage: int = Query(None, description="Minimum damage"),
    max_damage: int = Query(None, description="Maximum damage"),
    min_armor: int = Query(None, description="Minimum armor level"),
    max_armor: int = Query(None, description="Maximum armor level"),
    min_attack_bonus: float = Query(None, description="Minimum attack bonus"),
    min_crit_damage_rating: int = Query(
        None, description="Minimum critical damage rating"
    ),
    min_damage_rating: int = Query(None, description="Minimum damage rating"),
    min_heal_boost_rating: int = Query(None, description="Minimum heal boost rating"),
    max_level: int = Query(None, description="Maximum wield level requirement"),
    min_level: int = Query(None, description="Minimum wield level requirement"),
    material: str = Query(None, description="Material type (partial match)"),
    min_workmanship: float = Query(None, description="Minimum workmanship"),
    has_imbue: bool = Query(None, description="Has imbue effects"),
    item_set: str = Query(None, description="Item set name (partial match)"),
    min_tinks: int = Query(None, description="Minimum tinker count"),
    bonded: bool = Query(None, description="Bonded status"),
    attuned: bool = Query(None, description="Attuned status"),
    unique: bool = Query(None, description="Unique item status"),
    is_rare: bool = Query(None, description="Rare item status"),
    min_condition: int = Query(None, description="Minimum condition percentage"),
    min_value: int = Query(None, description="Minimum item value"),
    max_value: int = Query(None, description="Maximum item value"),
    max_burden: int = Query(None, description="Maximum burden"),
    sort_by: str = Query(
        "name", description="Sort field: name, value, damage, armor, workmanship"
    ),
    sort_dir: str = Query("asc", description="Sort direction: asc or desc"),
    page: int = Query(1, ge=1, description="Page number"),
    limit: int = Query(50, ge=1, le=200, description="Items per page"),
):
    """Proxy to inventory service comprehensive item search."""
    try:
        # Build query parameters
        params = {}
        if text:
            params["text"] = text
        if character:
            params["character"] = character
        if include_all_characters:
            params["include_all_characters"] = include_all_characters
        if equipment_status:
            params["equipment_status"] = equipment_status
        if equipment_slot is not None:
            params["equipment_slot"] = equipment_slot
        # Category filtering
        if armor_only:
            params["armor_only"] = armor_only
        if jewelry_only:
            params["jewelry_only"] = jewelry_only
        if weapon_only:
            params["weapon_only"] = weapon_only
        # Spell filtering
        if has_spell:
            params["has_spell"] = has_spell
        if spell_contains:
            params["spell_contains"] = spell_contains
        if legendary_cantrips:
            params["legendary_cantrips"] = legendary_cantrips
        # Combat properties
        if min_damage is not None:
            params["min_damage"] = min_damage
        if max_damage is not None:
            params["max_damage"] = max_damage
        if min_armor is not None:
            params["min_armor"] = min_armor
        if max_armor is not None:
            params["max_armor"] = max_armor
        if min_attack_bonus is not None:
            params["min_attack_bonus"] = min_attack_bonus
        if min_crit_damage_rating is not None:
            params["min_crit_damage_rating"] = min_crit_damage_rating
        if min_damage_rating is not None:
            params["min_damage_rating"] = min_damage_rating
        if min_heal_boost_rating is not None:
            params["min_heal_boost_rating"] = min_heal_boost_rating
        if max_level is not None:
            params["max_level"] = max_level
        if min_level is not None:
            params["min_level"] = min_level
        if material:
            params["material"] = material
        if min_workmanship is not None:
            params["min_workmanship"] = min_workmanship
        if has_imbue is not None:
            params["has_imbue"] = has_imbue
        if item_set:
            params["item_set"] = item_set
        if min_tinks is not None:
            params["min_tinks"] = min_tinks
        if bonded is not None:
            params["bonded"] = bonded
        if attuned is not None:
            params["attuned"] = attuned
        if unique is not None:
            params["unique"] = unique
        if is_rare is not None:
            params["is_rare"] = is_rare
        if min_condition is not None:
            params["min_condition"] = min_condition
        if min_value is not None:
            params["min_value"] = min_value
        if max_value is not None:
            params["max_value"] = max_value
        if max_burden is not None:
            params["max_burden"] = max_burden
        params["sort_by"] = sort_by
        params["sort_dir"] = sort_dir
        params["page"] = page
        params["limit"] = limit

        async with httpx.AsyncClient(timeout=30.0) as client:
            response = await client.get(
                f"{INVENTORY_SERVICE_URL}/search/items", params=params
            )

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            else:
                logger.error(
                    f"Inventory search service returned {response.status_code}"
                )
                raise HTTPException(
                    status_code=response.status_code,
                    detail="Inventory search service error",
                )

    except httpx.RequestError as e:
        logger.error(f"Could not reach inventory service: {e}")
        raise HTTPException(status_code=503, detail="Inventory service unavailable")
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Failed to search items: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/search/equipped/{character_name}")
async def search_equipped_items_proxy(
    character_name: str,
    slot: int = Query(None, description="Specific equipment slot mask"),
):
    """Proxy to inventory service equipped items search."""
    try:
        params = {}
        if slot is not None:
            params["slot"] = slot

        async with httpx.AsyncClient(timeout=10.0) as client:
            response = await client.get(
                f"{INVENTORY_SERVICE_URL}/search/equipped/{character_name}",
                params=params,
            )

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            elif response.status_code == 404:
                raise HTTPException(
                    status_code=404,
                    detail=f"No equipped items found for character '{character_name}'",
                )
            else:
                logger.error(
                    f"Inventory service returned {response.status_code} for equipped items search"
                )
                raise HTTPException(
                    status_code=response.status_code, detail="Inventory service error"
                )

    except httpx.RequestError as e:
        logger.error(f"Could not reach inventory service: {e}")
        raise HTTPException(status_code=503, detail="Inventory service unavailable")
    except HTTPException:
        raise
    except Exception as e:
        logger.error(
            f"Failed to search equipped items for {character_name}: {e}", exc_info=True
        )
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/search/upgrades/{character_name}/{slot}")
async def find_equipment_upgrades_proxy(
    character_name: str,
    slot: int,
    upgrade_type: str = Query(
        "damage", description="What to optimize for: damage, armor, workmanship, value"
    ),
):
    """Proxy to inventory service equipment upgrades search."""
    try:
        params = {"upgrade_type": upgrade_type}

        async with httpx.AsyncClient(timeout=10.0) as client:
            response = await client.get(
                f"{INVENTORY_SERVICE_URL}/search/upgrades/{character_name}/{slot}",
                params=params,
            )

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            elif response.status_code == 404:
                raise HTTPException(
                    status_code=404,
                    detail=f"No upgrade options found for character '{character_name}' slot {slot}",
                )
            else:
                logger.error(
                    f"Inventory service returned {response.status_code} for upgrades search"
                )
                raise HTTPException(
                    status_code=response.status_code, detail="Inventory service error"
                )

    except httpx.RequestError as e:
        logger.error(f"Could not reach inventory service: {e}")
        raise HTTPException(status_code=503, detail="Inventory service unavailable")
    except HTTPException:
        raise
    except Exception as e:
        logger.error(
            f"Failed to find equipment upgrades for {character_name} slot {slot}: {e}",
            exc_info=True,
        )
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/sets/list")
async def list_equipment_sets_proxy():
    """Proxy to inventory service equipment sets list."""
    try:
        async with httpx.AsyncClient(timeout=10.0) as client:
            response = await client.get(f"{INVENTORY_SERVICE_URL}/sets/list")

            if response.status_code == 200:
                return JSONResponse(content=response.json())
            else:
                logger.error(
                    f"Inventory service returned {response.status_code} for sets list"
                )
                raise HTTPException(
                    status_code=response.status_code, detail="Inventory service error"
                )

    except httpx.RequestError as e:
        logger.error(f"Could not reach inventory service: {e}")
        raise HTTPException(status_code=503, detail="Inventory service unavailable")
    except HTTPException:
        raise
    except Exception as e:
        logger.error(f"Failed to list equipment sets: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


# --------------------  WebSocket endpoints -----------------------
## WebSocket connection tracking
# Set of browser WebSocket clients subscribed to live updates
browser_conns: set[WebSocket] = set()
# Mapping of plugin clients by character_name to their WebSocket for command forwarding
plugin_conns: Dict[str, WebSocket] = {}


async def _send_to_browser(ws: WebSocket, data: dict) -> WebSocket | None:
    """Send data to a single browser client. Returns the ws if it failed, None if ok."""
    try:
        await asyncio.wait_for(ws.send_json(data), timeout=1.0)
    except (WebSocketDisconnect, RuntimeError, ConnectionAbortedError) as e:
        logger.debug(f"Detected disconnected browser client: {e}")
        return ws
    except asyncio.TimeoutError:
        logger.warning(
            "Timed out broadcasting to browser client; removing stale connection"
        )
        return ws
    except Exception as e:
        logger.warning(f"Unexpected error broadcasting to browser client: {e}")
        return ws
    return None


async def _do_broadcast(data: dict):
    """Send data to all browser clients concurrently. Runs as a background task."""
    clients = list(browser_conns)
    if not clients:
        return
    results = await asyncio.gather(*(_send_to_browser(ws, data) for ws in clients))
    for ws in results:
        if ws is not None:
            browser_conns.discard(ws)


async def _broadcast_to_browser_clients(snapshot: dict):
    """Broadcast a telemetry or chat message to all connected browser clients.

    Fires off a background task so the plugin receive loop is never blocked
    by slow browser connections.
    """
    data = jsonable_encoder(snapshot)
    task = asyncio.create_task(_do_broadcast(data))
    _broadcast_tasks.add(task)
    task.add_done_callback(_broadcast_tasks.discard)


async def _forward_to_inventory_service(inventory_msg: FullInventoryMessage):
    """Forward inventory data to the inventory microservice for processing."""
    try:
        # Prepare data for inventory service
        inventory_data = {
            "character_name": inventory_msg.character_name,
            "timestamp": inventory_msg.timestamp.isoformat(),
            "items": inventory_msg.items,
        }

        async with httpx.AsyncClient(timeout=30.0) as client:
            response = await client.post(
                f"{INVENTORY_SERVICE_URL}/process-inventory", json=inventory_data
            )

            if response.status_code == 200:
                result = response.json()
                logger.info(
                    f"Inventory service processed {result['processed']} items for {inventory_msg.character_name}"
                )
            else:
                logger.error(
                    f"Inventory service error {response.status_code}: {response.text}"
                )

    except Exception as e:
        logger.error(f"Failed to forward inventory to service: {e}")
        # Don't raise - this shouldn't block the main storage


async def _store_inventory(inventory_msg: FullInventoryMessage):
    """Forward inventory data to inventory microservice for processing and storage."""
    try:
        # Forward to inventory microservice for enhanced processing and storage
        await _forward_to_inventory_service(inventory_msg)

        # Optional: Create JSON file for debugging (can be removed in production)
        inventory_dir = Path("./inventory")
        inventory_dir.mkdir(exist_ok=True)

        file_path = inventory_dir / f"{inventory_msg.character_name}_inventory.json"
        inventory_data = {
            "character_name": inventory_msg.character_name,
            "timestamp": inventory_msg.timestamp.isoformat(),
            "item_count": inventory_msg.item_count,
            "items": inventory_msg.items,
        }

        with open(file_path, "w") as f:
            json.dump(inventory_data, f, indent=2)

    except Exception as e:
        logger.error(
            f"Failed to forward inventory for {inventory_msg.character_name}: {e}",
            exc_info=True,
        )
        raise


@app.websocket("/ws/position")
async def ws_receive_snapshots(
    websocket: WebSocket,
    secret: str | None = Query(None),
    x_plugin_secret: str | None = Header(None),
):
    """WebSocket endpoint for plugin clients to send telemetry and events.

    Validates a shared secret for authentication, then listens for messages of
    various types (register, spawn, telemetry, rare, chat) and handles each:
      - register: record plugin WebSocket for command forwarding
      - spawn: persist spawn event
      - telemetry: store snapshot, update stats, broadcast to browsers
      - rare: update total and session rare counts, persist event
      - chat: broadcast chat messages to browsers
    """
    global _plugin_connections

    # Authenticate plugin connection using shared secret
    key = secret or x_plugin_secret
    if key != SHARED_SECRET:
        # Reject without completing the WebSocket handshake
        logger.warning(
            f"Plugin WebSocket authentication failed from {websocket.client}"
        )
        await websocket.close(code=1008)
        return
    # Accept the WebSocket connection
    await websocket.accept()
    logger.info(f"🔌 PLUGIN_CONNECTED: {websocket.client}")

    # Track plugin connection
    _plugin_connections += 1

    try:
        while True:
            # Read next text frame
            try:
                raw = await websocket.receive_text()
                # Debug: log all incoming plugin WebSocket messages
                logger.debug(f"Plugin WebSocket RX from {websocket.client}: {raw}")
            except WebSocketDisconnect:
                logger.info(f"🔌 PLUGIN_DISCONNECTED: {websocket.client}")
                break
            # Parse JSON payload
            try:
                data = json.loads(raw)
            except json.JSONDecodeError as e:
                logger.warning(f"Invalid JSON from plugin {websocket.client}: {e}")
                continue
            msg_type = data.get("type")
            # --- Registration: associate character_name with this plugin socket ---
            if msg_type == "register":
                name = data.get("character_name") or data.get("player_name")
                if isinstance(name, str):
                    plugin_conns[name] = websocket
                    live_equipment_cantrip_states.pop(name, None)
                    logger.info(f"📋 PLUGIN_REGISTERED: {name} from {websocket.client}")
                continue
            # --- Spawn event: persist to spawn_events table ---
            if msg_type == "spawn":
                payload = data.copy()
                payload.pop("type", None)
                try:
                    spawn = SpawnEvent.parse_obj(payload)
                    await database.execute(spawn_events.insert().values(**spawn.dict()))
                    logger.debug(
                        f"Recorded spawn event: {spawn.mob} by {spawn.character_name}"
                    )
                except Exception as e:
                    logger.error(f"Failed to process spawn event: {e}")
                    continue
                continue
            # --- Telemetry message: persist snapshot and update kill stats ---
            if msg_type == "telemetry":
                # Parse telemetry snapshot and update in-memory state
                payload = data.copy()
                payload.pop("type", None)
                character_name = payload.get("character_name", "unknown")

                # Track message receipt and start timing
                telemetry_start_time = time.time()
                logger.info(
                    f"📨 TELEMETRY_RECEIVED: {character_name} from {websocket.client}"
                )

                try:
                    snap = TelemetrySnapshot.parse_obj(payload)
                    live_snapshots[snap.character_name] = snap.dict()
                    # Prepare data and compute kill delta
                    db_data = snap.dict()
                    db_data["rares_found"] = 0
                    key = (snap.session_id, snap.character_name)

                    # Get last recorded kill count for this session
                    if key in ws_receive_snapshots._last_kills:
                        last = ws_receive_snapshots._last_kills[key]
                    else:
                        # Cache miss - check database for last kill count for this session
                        row = await database.fetch_one(
                            "SELECT kills FROM telemetry_events WHERE character_name = :char AND session_id = :session ORDER BY timestamp DESC LIMIT 1",
                            {"char": snap.character_name, "session": snap.session_id},
                        )
                        last = row["kills"] if row else 0
                        logger.debug(
                            f"Cache miss for {snap.character_name} session {snap.session_id[:8]}: loaded last_kills={last} from database"
                        )

                    delta = snap.kills - last
                    # Persist snapshot and any kill delta in a single transaction
                    db_start_time = time.time()

                    # Log connection pool status before database operation
                    try:
                        pool_status = (
                            f"pool_size:{database._pool._queue.qsize()}"
                            if hasattr(database, "_pool")
                            and hasattr(database._pool, "_queue")
                            else "pool_status:unknown"
                        )
                    except:
                        pool_status = "pool_status:error"

                    logger.info(
                        f"💾 TELEMETRY_DB_WRITE_ATTEMPT: {snap.character_name} session:{snap.session_id[:8]} kills:{snap.kills} delta:{delta} {pool_status}"
                    )

                    try:
                        async with database.transaction():
                            await database.execute(
                                telemetry_events.insert().values(**db_data)
                            )
                            if delta > 0:
                                stmt = (
                                    pg_insert(char_stats)
                                    .values(
                                        character_name=snap.character_name,
                                        total_kills=delta,
                                    )
                                    .on_conflict_do_update(
                                        index_elements=["character_name"],
                                        set_={
                                            "total_kills": char_stats.c.total_kills
                                            + delta
                                        },
                                    )
                                )
                                await database.execute(stmt)
                                logger.debug(
                                    f"Updated kills for {snap.character_name}: +{delta} (total from {last} to {snap.kills})"
                                )

                        # Success: log timing and update cache
                        db_duration = (time.time() - db_start_time) * 1000
                        ws_receive_snapshots._last_kills[key] = snap.kills

                        # Track database performance (Phase 2)
                        global _total_queries, _total_query_time
                        _total_queries += 1
                        _total_query_time += (
                            db_duration / 1000.0
                        )  # Convert ms to seconds

                        # Track recent activity (Phase 3)
                        global _recent_telemetry_messages, _max_recent_messages
                        activity_entry = {
                            "timestamp": datetime.now(timezone.utc).isoformat(),
                            "character_name": snap.character_name,
                            "kills": snap.kills,
                            "kill_delta": delta,
                            "query_time": round(db_duration, 1),
                        }
                        _recent_telemetry_messages.append(activity_entry)
                        if len(_recent_telemetry_messages) > _max_recent_messages:
                            _recent_telemetry_messages.pop(0)

                        # Log final pool status after successful operation
                        try:
                            final_pool_status = (
                                f"pool_size:{database._pool._queue.qsize()}"
                                if hasattr(database, "_pool")
                                and hasattr(database._pool, "_queue")
                                else "pool_status:unknown"
                            )
                        except:
                            final_pool_status = "pool_status:error"

                        logger.info(
                            f"✅ TELEMETRY_DB_WRITE_SUCCESS: {snap.character_name} took {db_duration:.1f}ms {final_pool_status}"
                        )

                    except Exception as db_error:
                        db_duration = (time.time() - db_start_time) * 1000

                        # Log pool status during failure
                        try:
                            error_pool_status = (
                                f"pool_size:{database._pool._queue.qsize()}"
                                if hasattr(database, "_pool")
                                and hasattr(database._pool, "_queue")
                                else "pool_status:unknown"
                            )
                        except:
                            error_pool_status = "pool_status:error"

                        logger.error(
                            f"❌ TELEMETRY_DB_WRITE_FAILED: {snap.character_name} session:{snap.session_id[:8]} took {db_duration:.1f}ms {error_pool_status} error:{db_error}",
                            exc_info=True,
                        )
                        continue
                    # Broadcast updated snapshot to all browser clients
                    await _broadcast_to_browser_clients(snap.dict())

                    # Log successful processing completion with timing
                    total_duration = (time.time() - telemetry_start_time) * 1000
                    logger.info(
                        f"⏱️ TELEMETRY_PROCESSING_COMPLETE: {snap.character_name} took {total_duration:.1f}ms total"
                    )

                except Exception as e:
                    total_duration = (time.time() - telemetry_start_time) * 1000
                    logger.error(
                        f"❌ TELEMETRY_PROCESSING_FAILED: {character_name} took {total_duration:.1f}ms error:{e}",
                        exc_info=True,
                    )
                continue
            # --- Rare event: update total and session counters and persist ---
            if msg_type == "rare":
                name = data.get("character_name")
                if isinstance(name, str) and name.strip():
                    try:
                        # Total rare count per character
                        stmt_tot = (
                            pg_insert(rare_stats)
                            .values(character_name=name, total_rares=1)
                            .on_conflict_do_update(
                                index_elements=["character_name"],
                                set_={"total_rares": rare_stats.c.total_rares + 1},
                            )
                        )
                        await database.execute(stmt_tot)
                        # Session-specific rare count (use live cache or fallback to latest telemetry)
                        session_id = live_snapshots.get(name, {}).get("session_id")
                        if not session_id:
                            row = await database.fetch_one(
                                "SELECT session_id FROM telemetry_events"
                                " WHERE character_name = :name"
                                " ORDER BY timestamp DESC LIMIT 1",
                                {"name": name},
                            )
                            if row:
                                session_id = row["session_id"]
                        if session_id:
                            stmt_sess = (
                                pg_insert(rare_stats_sessions)
                                .values(
                                    character_name=name,
                                    session_id=session_id,
                                    session_rares=1,
                                )
                                .on_conflict_do_update(
                                    index_elements=["character_name", "session_id"],
                                    set_={
                                        "session_rares": rare_stats_sessions.c.session_rares
                                        + 1
                                    },
                                )
                            )
                            await database.execute(stmt_sess)
                        # Persist individual rare event for future analysis
                        payload = data.copy()
                        payload.pop("type", None)
                        try:
                            rare_ev = RareEvent.parse_obj(payload)
                            await database.execute(
                                rare_events.insert().values(**rare_ev.dict())
                            )
                            logger.info(
                                f"Recorded rare event: {rare_ev.name} by {name}"
                            )
                            # Broadcast rare event to browser clients for epic notifications
                            await _broadcast_to_browser_clients(data)
                        except Exception as e:
                            logger.error(f"Failed to persist rare event: {e}")
                    except Exception as e:
                        logger.error(
                            f"Failed to process rare event for {name}: {e}",
                            exc_info=True,
                        )
                continue
            # --- Chat message: forward chat payload to browser clients ---
            if msg_type == "chat":
                await _broadcast_to_browser_clients(data)
                logger.debug(
                    f"Broadcasted chat message from {data.get('character_name', 'unknown')}"
                )
                continue
            # --- Full inventory message: store complete inventory snapshot ---
            if msg_type == "full_inventory":
                payload = data.copy()
                payload.pop("type", None)
                try:
                    inventory_msg = FullInventoryMessage.parse_obj(payload)
                    await _store_inventory(inventory_msg)
                    logger.info(
                        f"Stored inventory for {inventory_msg.character_name}: {inventory_msg.item_count} items"
                    )
                except Exception as e:
                    logger.error(
                        f"Failed to process inventory for {data.get('character_name', 'unknown')}: {e}",
                        exc_info=True,
                    )
                continue
            # --- Inventory delta: single item add/remove/update ---
            if msg_type == "inventory_delta":
                try:
                    action = data.get("action")
                    char_name = data.get("character_name", "unknown")

                    if action == "remove":
                        item_id = data.get("item_id")
                        if item_id is not None:
                            async with httpx.AsyncClient(timeout=10.0) as client:
                                resp = await client.delete(
                                    f"{INVENTORY_SERVICE_URL}/inventory/{char_name}/item/{item_id}"
                                )
                                if resp.status_code >= 400:
                                    logger.warning(
                                        f"Inventory service returned {resp.status_code} for delta remove item_id={item_id}"
                                    )
                    elif action in ("add", "update"):
                        item = data.get("item")
                        if item:
                            async with httpx.AsyncClient(timeout=10.0) as client:
                                resp = await client.post(
                                    f"{INVENTORY_SERVICE_URL}/inventory/{char_name}/item",
                                    json=item,
                                )
                                if resp.status_code < 400:
                                    # Use enriched item from inventory-service response for broadcast
                                    resp_json = resp.json()
                                    enriched_item = resp_json.get("item")
                                    if enriched_item:
                                        data = {
                                            "type": "inventory_delta",
                                            "action": action,
                                            "character_name": char_name,
                                            "item": enriched_item,
                                        }
                                else:
                                    logger.warning(
                                        f"Inventory service returned {resp.status_code} for delta {action}"
                                    )

                    # Broadcast delta to all browser clients
                    await _broadcast_to_browser_clients(data)
                    logger.debug(f"Inventory delta ({action}) for {char_name}")
                except Exception as e:
                    logger.error(
                        f"Failed to process inventory delta: {e}", exc_info=True
                    )
                continue
            # --- Vitals message: store character health/stamina/mana and broadcast ---
            if msg_type == "vitals":
                payload = data.copy()
                payload.pop("type", None)
                try:
                    vitals_msg = VitalsMessage.parse_obj(payload)
                    live_vitals[vitals_msg.character_name] = vitals_msg.dict()
                    await _broadcast_to_browser_clients(data)
                    logger.debug(
                        f"Updated vitals for {vitals_msg.character_name}: {vitals_msg.health_percentage}% HP, {vitals_msg.stamina_percentage}% Stam, {vitals_msg.mana_percentage}% Mana"
                    )
                except Exception as e:
                    logger.error(
                        f"Failed to process vitals for {data.get('character_name', 'unknown')}: {e}",
                        exc_info=True,
                    )
                continue
            # --- Character stats message: store character attributes/skills/progression and broadcast ---
            if msg_type == "character_stats":
                payload = data.copy()
                payload.pop("type", None)
                try:
                    stats_msg = CharacterStatsMessage.parse_obj(payload)
                    stats_dict = stats_msg.dict()

                    # Cache in memory
                    live_character_stats[stats_msg.character_name] = stats_dict

                    # Build stats_data JSONB (everything except extracted columns)
                    stats_data = {}
                    for key in (
                        "attributes",
                        "vitals",
                        "skills",
                        "allegiance",
                        "active_item_enchantments",
                        "race",
                        "gender",
                        "birth",
                        "current_title",
                        "skill_credits",
                        "burden",
                        "burden_units",
                        "encumbrance_capacity",
                        "properties",
                        "titles",
                    ):
                        if stats_dict.get(key) is not None:
                            stats_data[key] = stats_dict[key]

                    # Upsert to database
                    await database.execute(
                        """
                        INSERT INTO character_stats
                            (character_name, timestamp, level, total_xp, unassigned_xp,
                             luminance_earned, luminance_total, deaths, stats_data)
                        VALUES
                            (:character_name, :timestamp, :level, :total_xp, :unassigned_xp,
                             :luminance_earned, :luminance_total, :deaths, :stats_data)
                        ON CONFLICT (character_name) DO UPDATE SET
                            timestamp = EXCLUDED.timestamp,
                            level = EXCLUDED.level,
                            total_xp = EXCLUDED.total_xp,
                            unassigned_xp = EXCLUDED.unassigned_xp,
                            luminance_earned = EXCLUDED.luminance_earned,
                            luminance_total = EXCLUDED.luminance_total,
                            deaths = EXCLUDED.deaths,
                            stats_data = EXCLUDED.stats_data
                        """,
                        {
                            "character_name": stats_msg.character_name,
                            "timestamp": stats_msg.timestamp,
                            "level": stats_msg.level,
                            "total_xp": stats_msg.total_xp,
                            "unassigned_xp": stats_msg.unassigned_xp,
                            "luminance_earned": stats_msg.luminance_earned,
                            "luminance_total": stats_msg.luminance_total,
                            "deaths": stats_msg.deaths,
                            "stats_data": json.dumps(stats_data),
                        },
                    )

                    # Broadcast to browser clients
                    await _broadcast_to_browser_clients(data)
                    logger.info(
                        f"Updated character stats for {stats_msg.character_name}: Level {stats_msg.level}"
                    )
                except Exception as e:
                    logger.error(
                        f"Failed to process character_stats for {data.get('character_name', 'unknown')}: {e}",
                        exc_info=True,
                    )
                continue
            # --- Equipment cantrip state: live-only overlay for mana panel ---
            if msg_type == "equipment_cantrip_state":
                try:
                    character_name = data.get("character_name")
                    if character_name:
                        live_equipment_cantrip_states[character_name] = data
                    await _broadcast_to_browser_clients(data)
                    logger.debug(
                        f"Updated equipment cantrip state for {character_name}"
                    )
                except Exception as e:
                    logger.error(
                        f"Failed to process equipment_cantrip_state for {data.get('character_name', 'unknown')}: {e}",
                        exc_info=True,
                    )
                continue
            # --- Quest message: update cache and broadcast (no database storage) ---
            if msg_type == "quest":
                character_name = data.get("character_name")
                quest_name = data.get("quest_name")
                countdown = data.get("countdown")

                if character_name and quest_name and countdown is not None:
                    # Only track specific quest types
                    allowed_quests = {
                        "Stipend Collection Timer",
                        "Blank Augmentation Gem Pickup Timer",
                        "Insatiable Eater Jaw",
                    }

                    if quest_name in allowed_quests:
                        # Update quest cache
                        if character_name not in _quest_status_cache:
                            _quest_status_cache[character_name] = {}
                        _quest_status_cache[character_name][quest_name] = countdown

                        # Broadcast to browser clients for real-time updates
                        await _broadcast_to_browser_clients(data)
                        logger.debug(
                            f"Updated quest status for {character_name}: {quest_name} = {countdown}"
                        )
                    else:
                        logger.debug(f"Ignoring non-tracked quest: {quest_name}")
                else:
                    logger.warning(
                        f"Invalid quest message format from {websocket.client}: missing required fields"
                    )
                continue
            # --- Portal message: store in database and broadcast ---
            if msg_type == "portal":
                character_name = data.get("character_name")
                portal_name = data.get("portal_name")
                ns = data.get("ns")
                ew = data.get("ew")
                z = data.get("z")
                timestamp_str = data.get("timestamp")

                if all([character_name, portal_name, ns, ew, z, timestamp_str]):
                    try:
                        # Parse timestamp
                        timestamp = datetime.fromisoformat(
                            timestamp_str.replace("Z", "+00:00")
                        )

                        # Convert coordinates to floats for database storage
                        ns = float(ns)
                        ew = float(ew)
                        z = float(z)

                        # Round coordinates for display (0.1 tolerance to match DB constraint)
                        ns_rounded = round(ns, 1)
                        ew_rounded = round(ew, 1)

                        # Use PostgreSQL UPSERT to handle race conditions atomically
                        # This completely eliminates duplicate key errors
                        result = await database.fetch_one(
                            """
                            INSERT INTO portals (portal_name, ns, ew, z, discovered_at, discovered_by)
                            VALUES (:portal_name, :ns, :ew, :z, :timestamp, :character_name)
                            ON CONFLICT (ROUND(ns::numeric, 1), ROUND(ew::numeric, 1))
                            DO UPDATE SET
                                discovered_at = EXCLUDED.discovered_at,
                                discovered_by = EXCLUDED.discovered_by,
                                portal_name = EXCLUDED.portal_name
                            RETURNING (xmax = 0) AS was_inserted
                            """,
                            {
                                "portal_name": portal_name,
                                "ns": ns,
                                "ew": ew,
                                "z": z,
                                "timestamp": timestamp,
                                "character_name": character_name,
                            },
                        )

                        # Log whether this was a new discovery or an update
                        # xmax = 0 means it was an INSERT (new portal)
                        # xmax != 0 means it was an UPDATE (existing portal)
                        if result and result["was_inserted"]:
                            logger.info(
                                f"New portal discovered: {portal_name} at {ns_rounded}, {ew_rounded} by {character_name}"
                            )
                        else:
                            logger.debug(
                                f"Portal timestamp updated: {portal_name} at {ns_rounded}, {ew_rounded} by {character_name}"
                            )

                        # Broadcast to browser clients for map updates
                        await _broadcast_to_browser_clients(data)

                    except Exception as e:
                        logger.error(
                            f"Failed to process portal discovery for {character_name}: {e}",
                            exc_info=True,
                        )
                else:
                    logger.warning(
                        f"Invalid portal message format from {websocket.client}: missing required fields"
                    )
                continue
            # Unknown message types are ignored
            if msg_type:
                logger.warning(
                    f"Unknown message type '{msg_type}' from {websocket.client}"
                )
    finally:
        # Track plugin disconnection
        _plugin_connections = max(0, _plugin_connections - 1)
        disconnected_names = [
            name for name, ws in plugin_conns.items() if ws is websocket
        ]
        for name in disconnected_names:
            plugin_conns.pop(name, None)
            live_equipment_cantrip_states.pop(name, None)

        # Clean up any plugin registrations for this socket
        to_remove = [n for n, ws in plugin_conns.items() if ws is websocket]
        for n in to_remove:
            # Use pop() instead of del to avoid KeyError if already removed
            plugin_conns.pop(n, None)

        # Also clean up any entries in the kill tracking cache for this session
        # Remove entries that might be associated with disconnected clients
        stale_keys = []
        for (session_id, char_name), _ in ws_receive_snapshots._last_kills.items():
            if char_name in to_remove:
                stale_keys.append((session_id, char_name))
        for key in stale_keys:
            ws_receive_snapshots._last_kills.pop(key, None)

        if to_remove:
            logger.info(
                f"Cleaned up plugin connections for characters: {to_remove} from {websocket.client}"
            )
            if stale_keys:
                logger.debug(
                    f"Cleaned up {len(stale_keys)} kill tracking cache entries"
                )
        else:
            logger.debug(f"No plugin registrations to clean up for {websocket.client}")


# In-memory cache of last seen kill counts per (session_id, character_name)
# Used to compute deltas for updating persistent kill statistics efficiently
ws_receive_snapshots._last_kills = {}


async def cleanup_stale_connections():
    """Periodic cleanup of stale WebSocket connections.

    This function can be called periodically to clean up connections
    that may have become stale but weren't properly cleaned up.
    """
    # Clean up plugin connections that no longer have valid WebSockets
    stale_plugins = []
    for char_name, ws in list(plugin_conns.items()):
        try:
            # Test if the WebSocket is still alive by checking its state
            if ws.client_state.name != "CONNECTED":
                stale_plugins.append(char_name)
        except Exception:
            # If we can't check the state, consider it stale
            stale_plugins.append(char_name)

    for char_name in stale_plugins:
        plugin_conns.pop(char_name, None)
        logger.info(f"Cleaned up stale plugin connection: {char_name}")

    # Clean up browser connections
    stale_browsers = []
    for ws in list(browser_conns):
        try:
            if ws.client_state.name != "CONNECTED":
                stale_browsers.append(ws)
        except Exception:
            stale_browsers.append(ws)

    for ws in stale_browsers:
        browser_conns.discard(ws)

    if stale_browsers:
        logger.info(f"Cleaned up {len(stale_browsers)} stale browser connections")

    logger.debug(
        f"Connection health check: {len(plugin_conns)} plugins, {len(browser_conns)} browsers"
    )


@app.websocket("/ws/live")
async def ws_live_updates(websocket: WebSocket):
    """WebSocket endpoint for browser clients to receive live updates and send commands.

    Manages a set of connected browser clients; listens for incoming command messages
    and forwards them to the appropriate plugin client WebSocket.
    """
    global _browser_connections
    # Add new browser client to the set
    await websocket.accept()
    browser_conns.add(websocket)
    logger.info(f"Browser WebSocket connected: {websocket.client}")

    # Track browser connection
    _browser_connections += 1

    try:
        while True:
            # Receive command messages from browser
            try:
                data = await websocket.receive_json()
                # Debug: log all incoming browser WebSocket messages
                logger.debug(f"Browser WebSocket RX from {websocket.client}: {data}")
            except WebSocketDisconnect:
                logger.info(f"Browser WebSocket disconnected: {websocket.client}")
                break
            # Determine command envelope format (new or legacy)
            if "player_name" in data and "command" in data:
                # New format: { player_name, command }
                target_name = data["player_name"]
                payload = data
            elif (
                data.get("type") == "command"
                and "character_name" in data
                and "text" in data
            ):
                # Legacy format: { type: 'command', character_name, text }
                target_name = data.get("character_name")
                payload = {"player_name": target_name, "command": data.get("text")}
            else:
                # Not a recognized command envelope
                continue
            # Forward command envelope to the appropriate plugin WebSocket
            target_ws = plugin_conns.get(target_name)
            if target_ws:
                try:
                    await target_ws.send_json(payload)
                    logger.debug(
                        f"Forwarded command to plugin for {target_name}: {payload}"
                    )
                except (WebSocketDisconnect, RuntimeError, ConnectionAbortedError) as e:
                    logger.warning(f"Failed to forward command to {target_name}: {e}")
                    # Remove stale connection
                    plugin_conns.pop(target_name, None)
                except Exception as e:
                    logger.error(
                        f"Unexpected error forwarding command to {target_name}: {e}"
                    )
                    # Remove potentially corrupted connection
                    plugin_conns.pop(target_name, None)
            else:
                logger.warning(
                    f"No plugin connection found for target character: {target_name}"
                )
    except WebSocketDisconnect:
        pass
    finally:
        # Track browser disconnection
        _browser_connections = max(0, _browser_connections - 1)

        browser_conns.discard(websocket)
        logger.debug(
            f"Removed browser WebSocket from connection pool: {websocket.client}"
        )


## --------------------  static frontend  ---------------------------
## (static mount moved to end of file, below API routes)

# list routes for convenience
logger.info("🔍 Registered HTTP API routes:")
for route in app.routes:
    if isinstance(route, APIRoute):
        # Log the path and allowed methods for each API route
        logger.info(f"{route.path} -> {route.methods}")


# Add stats endpoint for per-character metrics
@app.get("/stats/{character_name}")
async def get_stats(character_name: str):
    """
    HTTP GET endpoint to retrieve per-character metrics:
      - latest_snapshot: most recent telemetry entry for the character
      - total_kills: accumulated kills from char_stats
      - total_rares: accumulated rares from rare_stats
    Returns 404 if character has no recorded telemetry.
    """
    try:
        # Single optimized query with LEFT JOINs to get all data in one round trip
        sql = """
            WITH latest AS (
                SELECT * FROM telemetry_events 
                WHERE character_name = :cn 
                ORDER BY timestamp DESC LIMIT 1
            )
            SELECT 
                l.*,
                COALESCE(cs.total_kills, 0) as total_kills,
                COALESCE(rs.total_rares, 0) as total_rares
            FROM latest l
            LEFT JOIN char_stats cs ON l.character_name = cs.character_name
            LEFT JOIN rare_stats rs ON l.character_name = rs.character_name
        """
        row = await database.fetch_one(sql, {"cn": character_name})
        if not row:
            logger.warning(f"No telemetry data found for character: {character_name}")
            raise HTTPException(status_code=404, detail="Character not found")

        # Extract latest snapshot data (exclude the added total_kills/total_rares)
        snap_dict = {
            k: v
            for k, v in dict(row).items()
            if k not in ("total_kills", "total_rares")
        }

        result = {
            "character_name": character_name,
            "latest_snapshot": snap_dict,
            "total_kills": row["total_kills"],
            "total_rares": row["total_rares"],
        }
        logger.debug(
            f"Retrieved stats for character: {character_name} (optimized query)"
        )
        return JSONResponse(content=jsonable_encoder(result))
    except HTTPException:
        raise
    except Exception as e:
        logger.error(
            f"Failed to get stats for character {character_name}: {e}", exc_info=True
        )
        raise HTTPException(status_code=500, detail="Internal server error")


# --- Character Stats API -------------------------------------------


@app.post("/character-stats/test")
async def test_character_stats_default():
    """Inject mock character_stats data for frontend development."""
    return await test_character_stats("TestCharacter")


@app.post("/character-stats/test/{name}")
async def test_character_stats(name: str):
    """Inject mock character_stats data for a specific character name.
    Processes through the same pipeline as real plugin data."""
    mock_data = {
        "type": "character_stats",
        "timestamp": datetime.utcnow().isoformat() + "Z",
        "character_name": name,
        "level": 275,
        "race": "Aluvian",
        "gender": "Male",
        "birth": "2018-03-15 14:22:33",
        "total_xp": 191226310247,
        "unassigned_xp": 4500000,
        "skill_credits": 2,
        "luminance_earned": 500000,
        "luminance_total": 1500000,
        "deaths": 3175,
        "current_title": 42,
        "attributes": {
            "strength": {"base": 290, "creation": 100},
            "endurance": {"base": 200, "creation": 100},
            "coordination": {"base": 240, "creation": 100},
            "quickness": {"base": 220, "creation": 10},
            "focus": {"base": 250, "creation": 100},
            "self": {"base": 200, "creation": 100},
        },
        "vitals": {
            "health": {"base": 341},
            "stamina": {"base": 400},
            "mana": {"base": 300},
        },
        "skills": {
            "war_magic": {"base": 533, "training": "Specialized"},
            "life_magic": {"base": 440, "training": "Specialized"},
            "creature_enchantment": {"base": 430, "training": "Specialized"},
            "item_enchantment": {"base": 420, "training": "Specialized"},
            "void_magic": {"base": 510, "training": "Specialized"},
            "melee_defense": {"base": 488, "training": "Specialized"},
            "missile_defense": {"base": 470, "training": "Specialized"},
            "magic_defense": {"base": 460, "training": "Specialized"},
            "two_handed_combat": {"base": 420, "training": "Specialized"},
            "heavy_weapons": {"base": 410, "training": "Specialized"},
            "finesse_weapons": {"base": 400, "training": "Trained"},
            "light_weapons": {"base": 390, "training": "Trained"},
            "missile_weapons": {"base": 380, "training": "Trained"},
            "shield": {"base": 350, "training": "Trained"},
            "dual_wield": {"base": 340, "training": "Trained"},
            "arcane_lore": {"base": 330, "training": "Trained"},
            "mana_conversion": {"base": 320, "training": "Trained"},
            "healing": {"base": 300, "training": "Trained"},
            "lockpick": {"base": 280, "training": "Trained"},
            "assess_creature": {"base": 10, "training": "Untrained"},
            "assess_person": {"base": 10, "training": "Untrained"},
            "deception": {"base": 10, "training": "Untrained"},
            "leadership": {"base": 10, "training": "Untrained"},
            "loyalty": {"base": 10, "training": "Untrained"},
            "jump": {"base": 10, "training": "Untrained"},
            "run": {"base": 10, "training": "Untrained"},
            "salvaging": {"base": 10, "training": "Untrained"},
            "cooking": {"base": 10, "training": "Untrained"},
            "fletching": {"base": 10, "training": "Untrained"},
            "alchemy": {"base": 10, "training": "Untrained"},
            "sneak_attack": {"base": 10, "training": "Untrained"},
            "dirty_fighting": {"base": 10, "training": "Untrained"},
            "recklessness": {"base": 10, "training": "Untrained"},
            "summoning": {"base": 10, "training": "Untrained"},
        },
        "allegiance": {
            "name": "Knights of Dereth",
            "monarch": {"name": "HighKing", "race": 1, "rank": 0, "gender": 0},
            "patron": {"name": "SirLancelot", "race": 1, "rank": 5, "gender": 0},
            "rank": 8,
            "followers": 12,
        },
    }

    # Process through the same pipeline as real data
    payload = mock_data.copy()
    payload.pop("type", None)
    try:
        stats_msg = CharacterStatsMessage.parse_obj(payload)
        stats_dict = stats_msg.dict()
        live_character_stats[stats_msg.character_name] = stats_dict

        stats_data = {}
        for key in (
            "attributes",
            "vitals",
            "skills",
            "allegiance",
            "active_item_enchantments",
            "race",
            "gender",
            "birth",
            "current_title",
            "skill_credits",
            "burden",
            "burden_units",
            "encumbrance_capacity",
        ):
            if stats_dict.get(key) is not None:
                stats_data[key] = stats_dict[key]

        await database.execute(
            """
            INSERT INTO character_stats
                (character_name, timestamp, level, total_xp, unassigned_xp,
                 luminance_earned, luminance_total, deaths, stats_data)
            VALUES
                (:character_name, :timestamp, :level, :total_xp, :unassigned_xp,
                 :luminance_earned, :luminance_total, :deaths, :stats_data)
            ON CONFLICT (character_name) DO UPDATE SET
                timestamp = EXCLUDED.timestamp,
                level = EXCLUDED.level,
                total_xp = EXCLUDED.total_xp,
                unassigned_xp = EXCLUDED.unassigned_xp,
                luminance_earned = EXCLUDED.luminance_earned,
                luminance_total = EXCLUDED.luminance_total,
                deaths = EXCLUDED.deaths,
                stats_data = EXCLUDED.stats_data
            """,
            {
                "character_name": stats_msg.character_name,
                "timestamp": stats_msg.timestamp,
                "level": stats_msg.level,
                "total_xp": stats_msg.total_xp,
                "unassigned_xp": stats_msg.unassigned_xp,
                "luminance_earned": stats_msg.luminance_earned,
                "luminance_total": stats_msg.luminance_total,
                "deaths": stats_msg.deaths,
                "stats_data": json.dumps(stats_data),
            },
        )

        await _broadcast_to_browser_clients(mock_data)
        return {"status": "ok", "character_name": stats_msg.character_name}
    except Exception as e:
        logger.error(f"Test endpoint failed: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail=str(e))


@app.get("/character-stats/{name}")
async def get_character_stats(name: str):
    """Return latest character stats. Checks in-memory cache first, falls back to DB."""
    try:
        # Try in-memory cache first
        if name in live_character_stats:
            return JSONResponse(content=jsonable_encoder(live_character_stats[name]))

        # Fall back to database
        row = await database.fetch_one(
            "SELECT * FROM character_stats WHERE character_name = :name", {"name": name}
        )
        if row:
            result = dict(row._mapping)
            # Parse stats_data back from JSONB
            if isinstance(result.get("stats_data"), str):
                result["stats_data"] = json.loads(result["stats_data"])
            # Merge stats_data fields into top level for frontend compatibility
            stats_data = result.pop("stats_data", {})
            result.update(stats_data)
            return JSONResponse(content=jsonable_encoder(result))

        return JSONResponse(
            content={"error": "No stats available for this character"}, status_code=404
        )
    except Exception as e:
        logger.error(f"Failed to get character stats for {name}: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail="Internal server error")


@app.get("/equipment-cantrip-state/{name}")
async def get_equipment_cantrip_state(name: str):
    """Return latest live equipment cantrip state overlay for a character."""
    try:
        data = live_equipment_cantrip_states.get(name)
        if data:
            return JSONResponse(content=jsonable_encoder(data))

        return JSONResponse(
            content={
                "type": "equipment_cantrip_state",
                "character_name": name,
                "items": [],
            }
        )
    except Exception as e:
        logger.error(
            f"Failed to get equipment cantrip state for {name}: {e}", exc_info=True
        )
        raise HTTPException(status_code=500, detail="Internal server error")


# --------------------  static frontend  ---------------------------
# Custom icon handler that prioritizes clean icons over originals
from fastapi.responses import FileResponse


@app.get("/icons/{icon_filename}")
async def serve_icon(icon_filename: str):
    """Serve icons from static/icons directory"""

    # Serve from static/icons directory
    icon_path = Path("static/icons") / icon_filename
    if icon_path.exists():
        return FileResponse(icon_path, media_type="image/png")

    # Icon not found
    raise HTTPException(status_code=404, detail="Icon not found")


# -------------------- Inventory Service Proxy ---------------------------


@app.get("/inv/test")
async def test_inventory_route():
    """Test route to verify inventory proxy is working"""
    return {"message": "Inventory proxy route is working"}


@app.post("/inv/suitbuilder/search")
async def proxy_suitbuilder_search(request: Request):
    """Stream suitbuilder search results - SSE requires streaming proxy."""
    inventory_service_url = os.getenv(
        "INVENTORY_SERVICE_URL", "http://inventory-service:8000"
    )
    logger.info(f"Streaming proxy to suitbuilder search")

    # Read body BEFORE creating generator (request context needed)
    body = await request.body()

    async def stream_response():
        try:
            # Use streaming request with long timeout for searches
            async with httpx.AsyncClient(
                timeout=httpx.Timeout(300.0, connect=10.0)
            ) as client:
                async with client.stream(
                    method="POST",
                    url=f"{inventory_service_url}/suitbuilder/search",
                    content=body,
                    headers={"Content-Type": "application/json"},
                ) as response:
                    async for chunk in response.aiter_bytes():
                        yield chunk
        except httpx.ReadTimeout:
            yield b'event: error\ndata: {"message": "Search timeout"}\n\n'
        except Exception as e:
            logger.error(f"Streaming proxy error: {e}")
            yield f'event: error\ndata: {{"message": "Proxy error: {str(e)}"}}\n\n'.encode()

    return StreamingResponse(
        stream_response(),
        media_type="text/event-stream",
        headers={
            "Cache-Control": "no-cache",
            "Connection": "keep-alive",
            "X-Accel-Buffering": "no",  # Disable nginx buffering
        },
    )


@app.api_route("/inv/{path:path}", methods=["GET", "POST"])
async def proxy_inventory_service(path: str, request: Request):
    """Proxy all inventory service requests"""
    try:
        inventory_service_url = os.getenv(
            "INVENTORY_SERVICE_URL", "http://inventory-service:8000"
        )
        logger.info(f"Proxying to inventory service: {inventory_service_url}/{path}")

        # Forward the request to inventory service (60s timeout for large queries)
        async with httpx.AsyncClient(timeout=60.0) as client:
            response = await client.request(
                method=request.method,
                url=f"{inventory_service_url}/{path}",
                params=request.query_params,
                headers=dict(request.headers),
                content=await request.body(),
            )
            return Response(
                content=response.content,
                status_code=response.status_code,
                headers=dict(response.headers),
            )
    except Exception as e:
        logger.error(f"Failed to proxy inventory request: {e}")
        raise HTTPException(status_code=500, detail="Inventory service unavailable")


# Icons are now served from static/icons directory
# Serve SPA files (catch-all for frontend routes)
# Mount the single-page application frontend (static assets) at root path
app.mount("/", StaticFiles(directory="static", html=True), name="static")