from typing import Any, Dict, List, Optional
import logging
import os
import json
import sys
import shutil
from datetime import datetime, timedelta

# Fail loudly on unsupported Python. The mcp/FastMCP dependency requires 3.11+.
# On Python 3.10 or older the server dies during import, and when it is launched
# by a GUI client (e.g. Claude Desktop) that failure is silent — no tools appear
# and no log file is written, which is impossible to diagnose. See issue #19.
if sys.version_info < (3, 11):
    sys.exit(
        "mcp-gsc requires Python 3.11 or newer, but you are running "
        f"{sys.version_info.major}.{sys.version_info.minor}. "
        "Install Python 3.11+ (https://www.python.org/downloads/) and recreate "
        "your virtual environment, or use the uvx install method which manages "
        "the Python version for you."
    )

from platformdirs import user_config_dir

import google.auth
from google.auth.transport.requests import Request
from google.oauth2 import service_account
from google.oauth2.credentials import Credentials
from google_auth_oauthlib.flow import InstalledAppFlow
from googleapiclient.discovery import build
from googleapiclient.errors import HttpError

# Suppress the noisy file_cache warning from google-api-python-client.
# Some MCP hosts (e.g. GitHub Copilot CLI) treat any stderr output as a
# fatal error, so this prevents false crashes.
logging.getLogger("googleapiclient.discovery_cache").setLevel(logging.ERROR)

# MCP
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("gsc-server")

def _expand_path(path: Optional[str]) -> Optional[str]:
    """Expand ``~`` and environment variables in a path, returning None for empty input.

    This lets users write ``GSC_CREDENTIALS_PATH=~/creds.json`` or
    ``GSC_CREDENTIALS_PATH=$HOME/creds.json`` without hitting silent lookup failures.
    """
    if not path:
        return None
    return os.path.expandvars(os.path.expanduser(path))


# Path to your service account JSON or user credentials JSON
# First check if GSC_CREDENTIALS_PATH environment variable is set
# Then try looking in the script directory and current working directory as fallbacks
GSC_CREDENTIALS_PATH = _expand_path(os.environ.get("GSC_CREDENTIALS_PATH"))
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
POSSIBLE_CREDENTIAL_PATHS = [
    GSC_CREDENTIALS_PATH,  # First try the environment variable if set
    os.path.join(SCRIPT_DIR, "service_account_credentials.json"),
    os.path.join(os.getcwd(), "service_account_credentials.json"),
    # Add any other potential paths here
]

# OAuth client secrets file path
OAUTH_CLIENT_SECRETS_FILE = _expand_path(os.environ.get("GSC_OAUTH_CLIENT_SECRETS_FILE"))
# Track whether the user explicitly set the env var (vs. using the SCRIPT_DIR fallback).
# Needed so get_gsc_service() can fail-fast when the explicit path is wrong, while
# leaving the fallback behavior unchanged for clone-install users.
GSC_OAUTH_CLIENT_SECRETS_FILE_EXPLICIT = OAUTH_CLIENT_SECRETS_FILE is not None
if not OAUTH_CLIENT_SECRETS_FILE:
    OAUTH_CLIENT_SECRETS_FILE = os.path.join(SCRIPT_DIR, "client_secrets.json")

# Token file path for storing OAuth tokens.
# Stored in the user config directory so it survives uvx updates (which replace SCRIPT_DIR).
# Override with GSC_CONFIG_DIR env var for Docker/power users.
_CONFIG_DIR = os.environ.get("GSC_CONFIG_DIR") or user_config_dir("mcp-gsc")
os.makedirs(_CONFIG_DIR, exist_ok=True)
TOKEN_FILE = os.path.join(_CONFIG_DIR, "token.json")

# Silently migrate token from old location (SCRIPT_DIR) on first run after upgrade.
# Existing users never need to re-authenticate.
_OLD_TOKEN = os.path.join(SCRIPT_DIR, "token.json")
if os.path.exists(_OLD_TOKEN) and not os.path.exists(TOKEN_FILE):
    shutil.move(_OLD_TOKEN, TOKEN_FILE)

# Environment variable to skip OAuth authentication
SKIP_OAUTH = os.environ.get("GSC_SKIP_OAUTH", "").lower() in ("true", "1", "yes")

# Safety flag for destructive operations (add_site, delete_site, delete_sitemap).
# Default is false — set GSC_ALLOW_DESTRUCTIVE=true to enable these tools.
ALLOW_DESTRUCTIVE = os.environ.get("GSC_ALLOW_DESTRUCTIVE", "false").lower() in ("true", "1", "yes")

# Data state for search analytics queries.
# "all"   → includes fresh/unconfirmed data, matches the GSC dashboard (default)
# "final" → only confirmed data, which lags 2-3 days behind the dashboard
_raw_data_state = os.environ.get("GSC_DATA_STATE", "all").lower().strip()
if _raw_data_state not in ("all", "final"):
    raise ValueError(
        f"Invalid GSC_DATA_STATE value '{_raw_data_state}'. "
        "Accepted values are 'all' (default, matches GSC dashboard) or 'final' (2-3 day lag)."
    )
DATA_STATE = _raw_data_state

SCOPES = ["https://www.googleapis.com/auth/webmasters"]

def get_gsc_service():
    """
    Returns an authorized Search Console service object.
    First tries OAuth authentication, then falls back to service account.
    """
    # Fail-fast if credential env vars are set but point to files that don't exist.
    # Without this, a typo'd or uvx-incompatible path would silently fall through to
    # SCRIPT_DIR/cwd fallbacks (which don't work under uvx) and emit a misleading
    # "file not found" error with no mention of the env var the user actually set.
    # We rely on the module-level captured values (GSC_CREDENTIALS_PATH,
    # GSC_OAUTH_CLIENT_SECRETS_FILE_EXPLICIT) rather than re-reading os.environ at
    # call time, because the env var may have been cleared by the caller after module
    # import (e.g. by test frameworks or MCP hosts that scope env vars narrowly).
    if GSC_CREDENTIALS_PATH and not os.path.exists(GSC_CREDENTIALS_PATH):
        raise FileNotFoundError(
            f"GSC_CREDENTIALS_PATH is set to {GSC_CREDENTIALS_PATH!r} but the file "
            f"does not exist. "
            f"If running via uvx, this MUST be an absolute path to your service account "
            f"credentials JSON file — placing the file in your project folder is not "
            f"sufficient because uvx runs the code from an internal cache directory."
        )
    if GSC_OAUTH_CLIENT_SECRETS_FILE_EXPLICIT and not os.path.exists(OAUTH_CLIENT_SECRETS_FILE):
        raise FileNotFoundError(
            f"GSC_OAUTH_CLIENT_SECRETS_FILE is set to {OAUTH_CLIENT_SECRETS_FILE!r} "
            f"but the file does not exist. "
            f"If running via uvx, this MUST be an absolute path to your OAuth "
            f"client_secrets.json file."
        )

    # Try OAuth authentication first if not skipped
    if not SKIP_OAUTH:
        try:
            return get_gsc_service_oauth()
        except Exception as e:
            # If OAuth fails, try service account
            logging.warning("OAuth authentication failed: %s", e)
            pass
    
    # Try service account authentication
    for cred_path in POSSIBLE_CREDENTIAL_PATHS:
        if cred_path and os.path.exists(cred_path):
            try:
                creds = service_account.Credentials.from_service_account_file(
                    cred_path, scopes=SCOPES
                )
                return build("searchconsole", "v1", credentials=creds, cache_discovery=False)
            except Exception as e:
                continue  # Try the next path if this one fails
    
    # If we get here, none of the authentication methods worked.
    # Note: uvx users can't place files "in the script directory" because uvx runs
    # the code from ~/.cache/uv/archive-v0/<hash>/lib/python*/site-packages/ — an
    # internal cache they cannot reach. They must use env vars with absolute paths.
    raise FileNotFoundError(
        f"Authentication failed. Please either:\n"
        f"1. Set up OAuth by setting GSC_OAUTH_CLIENT_SECRETS_FILE to an absolute path, "
        f"or (for clone installs) placing a client_secrets.json file in the script "
        f"directory, then call the 'reauthenticate' tool to open a browser login window "
        f"and complete authentication, or\n"
        f"2. Set GSC_CREDENTIALS_PATH to an absolute path, or (for clone installs) "
        f"place a service account credentials file in one of these locations: "
        f"{', '.join([p for p in POSSIBLE_CREDENTIAL_PATHS[1:] if p])}\n"
        f"\n"
        f"If you installed via uvx, the 'script directory' is an internal uv cache "
        f"that you cannot access — you MUST use the environment variables with "
        f"absolute paths."
    )

def get_gsc_service_oauth():
    """
    Returns an authorized Search Console service object using OAuth.
    """
    creds = None
    
    # Check if token file exists
    if os.path.exists(TOKEN_FILE):
        try:
            creds = Credentials.from_authorized_user_file(TOKEN_FILE, SCOPES)
        except Exception as e:
            # If token file is corrupted, delete it
            if os.path.exists(TOKEN_FILE):
                os.remove(TOKEN_FILE)
            creds = None
    
    # If credentials don't exist or are invalid, get new ones
    if not creds or not creds.valid:
        if creds and creds.expired and creds.refresh_token:
            try:
                creds.refresh(Request())
                # Save the refreshed credentials
                with open(TOKEN_FILE, 'w') as token:
                    token.write(creds.to_json())
            except Exception as e:
                # If refresh fails, delete the bad token and trigger new OAuth flow
                if os.path.exists(TOKEN_FILE):
                    os.remove(TOKEN_FILE)
                # Fall through to the OAuth flow below
                creds = None
        
        # Start new OAuth flow if we don't have valid credentials
        if not creds or not creds.valid:
            # Check if client secrets file exists
            if not os.path.exists(OAUTH_CLIENT_SECRETS_FILE):
                raise FileNotFoundError(
                    f"OAuth client secrets file not found. Please place a client_secrets.json file in the script directory "
                    f"or set the GSC_OAUTH_CLIENT_SECRETS_FILE environment variable."
                )

            # Start OAuth flow — opens a browser window on macOS even from MCP subprocess.
            flow = InstalledAppFlow.from_client_secrets_file(OAUTH_CLIENT_SECRETS_FILE, SCOPES)
            creds = flow.run_local_server(port=0)
            
            # Save the credentials for future use
            with open(TOKEN_FILE, 'w') as token:
                token.write(creds.to_json())
    
    # Build and return the service
    return build("searchconsole", "v1", credentials=creds, cache_discovery=False)


def _site_not_found_error(site_url: str) -> str:
    """Return a helpful message when a GSC property returns 404."""
    lines = [f"Property '{site_url}' not found (404). Possible causes:\n"]
    lines.append(
        "1. The site_url doesn't exactly match what is in GSC. "
        "Run list_properties to get the exact string to use."
    )
    if site_url.startswith("sc-domain:"):
        lines.append(
            "2. Domain properties require the service account to be explicitly added "
            "under GSC Settings > Users and permissions for that specific domain property. "
            "OAuth users must also have verified access to it."
        )
    else:
        lines.append(
            "2. If your property is a domain property (covers all subdomains), "
            "the correct format is 'sc-domain:example.com', not a full URL."
        )
    lines.append(
        "3. The authenticated account may not have access to this property."
    )
    return "\n".join(lines)


@mcp.tool()
async def get_capabilities() -> str:
    """
    Get a full list of all available tools, current auth status, and how to get started.
    ALWAYS call this first when asked what tools are available, what this server can do,
    or how to get started. Returns all tool names grouped by category in a single call —
    faster than searching individually. Also shows if authentication is needed.
    """
    # Check auth status
    try:
        get_gsc_service()
        auth_status = "✅ Authenticated — ready to use all tools."
    except Exception as e:
        auth_status = f"❌ Not authenticated — call the 'reauthenticate' tool first to open a browser login window.\nDetails: {e}"

    return f"""Google Search Console MCP Server

AUTH STATUS:
{auth_status}

GETTING STARTED:
1. If not authenticated, call the 'reauthenticate' tool to complete OAuth login.
2. Call 'list_properties' to see all your GSC sites and get the exact site_url for other tools.
3. Use any tool below with the site_url from step 2.

AVAILABLE TOOLS:

Authentication:
  - reauthenticate: Open browser OAuth login window. Call this if you see auth errors.

Properties:
  - list_properties: List all GSC sites/properties you have access to (start here)
  - get_site_details: Get verification and ownership details for a site

Analytics & Reporting:
  - get_search_analytics: Top queries and pages with clicks, impressions, CTR, position
  - get_performance_overview: Summary of site performance for a time period
  - compare_search_periods: Compare performance between two time periods
  - get_search_by_page_query: Search terms driving traffic to a specific page
  - get_advanced_search_analytics: Advanced filtering by country, device, query, page

URL Inspection & Indexing:
  - inspect_url_enhanced: Detailed crawl/index status for a specific URL
  - batch_url_inspection: Inspect up to 10 URLs at once
  - check_indexing_issues: Check multiple URLs for indexing problems

Sitemaps:
  - get_sitemaps: List all sitemaps for a site
  - list_sitemaps_enhanced: Detailed sitemap info including errors and warnings
  - manage_sitemaps: Submit or delete sitemaps (requires GSC_ALLOW_DESTRUCTIVE=true for delete)

Destructive (disabled by default, set GSC_ALLOW_DESTRUCTIVE=true to enable):
  - add_site: Add a new property to GSC
  - delete_site: Remove a property from GSC
"""


@mcp.tool()
async def list_properties() -> str:
    """
    List all Google Search Console (GSC) properties and sites the user has access to.
    Use this to see all verified sites, domain properties, and URL-prefix properties
    in the connected Google Search Console account. Always call this first to get the
    exact site_url needed for other tools.
    """
    try:
        service = get_gsc_service()
        site_list = service.sites().list().execute()

        # site_list is typically something like:
        # {
        #   "siteEntry": [
        #       {"siteUrl": "...", "permissionLevel": "..."},
        #       ...
        #   ]
        # }
        sites = site_list.get("siteEntry", [])

        if not sites:
            return "No Search Console properties found."

        return json.dumps({
            "count": len(sites),
            "properties": [
                {
                    "site_url": site.get("siteUrl", "Unknown"),
                    "permission_level": site.get("permissionLevel", "Unknown"),
                }
                for site in sites
            ],
        })
    except FileNotFoundError as e:
        return f"Error: {str(e)}"
    except Exception as e:
        return f"Error retrieving properties: {str(e)}"

@mcp.tool()
async def add_site(site_url: str) -> str:
    """
    Add a site to your Search Console properties.
    
    Args:
        site_url: The URL of the site to add (must be exact match e.g. https://example.com, or https://www.example.com, or https://subdomain.example.com/path/, for domain properties use format: sc-domain:example.com)
    """
    if not ALLOW_DESTRUCTIVE:
        return (
            "Safety: add_site is a destructive operation that modifies your GSC account. "
            "Set GSC_ALLOW_DESTRUCTIVE=true in your environment to enable add/delete tools."
        )
    try:
        service = get_gsc_service()
        
        # Add the site
        response = service.sites().add(siteUrl=site_url).execute()
        
        # Format the response
        result_lines = [f"Site {site_url} has been added to Search Console."]
        
        # Add permission level if available
        if "permissionLevel" in response:
            result_lines.append(f"Permission level: {response['permissionLevel']}")
        
        return "\n".join(result_lines)
    except HttpError as e:
        error_content = json.loads(e.content.decode('utf-8'))
        error_details = error_content.get('error', {})
        error_code = e.resp.status
        error_message = error_details.get('message', str(e))
        error_reason = error_details.get('errors', [{}])[0].get('reason', '')
        
        if error_code == 409:
            return f"Site {site_url} is already added to Search Console."
        elif error_code == 403:
            if error_reason == 'forbidden':
                return f"Error: You don't have permission to add this site. Please verify ownership first."
            elif error_reason == 'quotaExceeded':
                return f"Error: API quota exceeded. Please try again later."
            else:
                return f"Error: Permission denied. {error_message}"
        elif error_code == 400:
            if error_reason == 'invalidParameter':
                return f"Error: Invalid site URL format. Please check the URL format and try again."
            else:
                return f"Error: Bad request. {error_message}"
        elif error_code == 401:
            return f"Error: Unauthorized. Please check your credentials."
        elif error_code == 429:
            return f"Error: Too many requests. Please try again later."
        elif error_code == 500:
            return f"Error: Internal server error from Google Search Console API. Please try again later."
        elif error_code == 503:
            return f"Error: Service unavailable. Google Search Console API is currently down. Please try again later."
        else:
            return f"Error adding site (HTTP {error_code}): {error_message}"
    except Exception as e:
        return f"Error adding site: {str(e)}"

@mcp.tool()
async def delete_site(site_url: str) -> str:
    """
    Remove a site from your Search Console properties.
    
    Args:
        site_url: The URL of the site to remove (must be exact match e.g. https://example.com, or https://www.example.com, or https://subdomain.example.com/path/, for domain properties use format: sc-domain:example.com)
    """
    if not ALLOW_DESTRUCTIVE:
        return (
            "Safety: delete_site permanently removes a property from your GSC account. "
            "Set GSC_ALLOW_DESTRUCTIVE=true in your environment to enable add/delete tools."
        )
    try:
        service = get_gsc_service()
        
        # Delete the site
        service.sites().delete(siteUrl=site_url).execute()
        
        return f"Site {site_url} has been removed from Search Console."
    except HttpError as e:
        error_content = json.loads(e.content.decode('utf-8'))
        error_details = error_content.get('error', {})
        error_code = e.resp.status
        error_message = error_details.get('message', str(e))
        error_reason = error_details.get('errors', [{}])[0].get('reason', '')
        
        if error_code == 404:
            return f"Site {site_url} was not found in Search Console."
        elif error_code == 403:
            if error_reason == 'forbidden':
                return f"Error: You don't have permission to remove this site."
            elif error_reason == 'quotaExceeded':
                return f"Error: API quota exceeded. Please try again later."
            else:
                return f"Error: Permission denied. {error_message}"
        elif error_code == 400:
            if error_reason == 'invalidParameter':
                return f"Error: Invalid site URL format. Please check the URL format and try again."
            else:
                return f"Error: Bad request. {error_message}"
        elif error_code == 401:
            return f"Error: Unauthorized. Please check your credentials."
        elif error_code == 429:
            return f"Error: Too many requests. Please try again later."
        elif error_code == 500:
            return f"Error: Internal server error from Google Search Console API. Please try again later."
        elif error_code == 503:
            return f"Error: Service unavailable. Google Search Console API is currently down. Please try again later."
        else:
            return f"Error removing site (HTTP {error_code}): {error_message}"
    except Exception as e:
        return f"Error removing site: {str(e)}"

@mcp.tool()
async def get_search_analytics(site_url: str, days: int = 28, dimensions: str = "query", row_limit: int = 20) -> str:
    """
    Get search analytics data for a specific property.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        days: Number of days to look back (default: 28)
        dimensions: Dimensions to group by (default: query). Options: query, page, device, country, date
                   You can provide multiple dimensions separated by comma (e.g., "query,page")
        row_limit: Number of rows to return (default: 20, max: 500). Use 5-20 for quick overviews,
                   50-200 for deeper analysis, up to 500 for comprehensive reports. For bulk exports
                   beyond 500 rows, use get_advanced_search_analytics which supports pagination.
    """
    try:
        service = get_gsc_service()
        
        # Calculate date range
        end_date = datetime.now().date()
        start_date = end_date - timedelta(days=days)
        
        # Parse dimensions
        dimension_list = [d.strip() for d in dimensions.split(",")]
        
        # Build request
        request = {
            "startDate": start_date.strftime("%Y-%m-%d"),
            "endDate": end_date.strftime("%Y-%m-%d"),
            "dimensions": dimension_list,
            "rowLimit": min(max(1, row_limit), 500),
            "dataState": DATA_STATE
        }
        
        # Execute request
        response = service.searchanalytics().query(siteUrl=site_url, body=request).execute()
        
        if not response.get("rows"):
            return f"No search analytics data found for {site_url} in the last {days} days."

        rows = []
        for row in response.get("rows", []):
            entry = {}
            for i, dim in enumerate(dimension_list):
                entry[dim] = row.get("keys", [])[i] if i < len(row.get("keys", [])) else None
            entry["clicks"] = row.get("clicks", 0)
            entry["impressions"] = row.get("impressions", 0)
            entry["ctr"] = round(row.get("ctr", 0), 4)
            entry["position"] = round(row.get("position", 0), 1)
            rows.append(entry)

        return json.dumps({
            "site_url": site_url,
            "date_range": {
                "start": start_date.strftime("%Y-%m-%d"),
                "end": end_date.strftime("%Y-%m-%d"),
                "days": days,
            },
            "dimensions": dimension_list,
            "row_count": len(rows),
            "rows": rows,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error retrieving search analytics: {str(e)}"

@mcp.tool()
async def get_site_details(site_url: str) -> str:
    """
    Get detailed information about a specific Search Console property.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
    """
    try:
        service = get_gsc_service()
        
        # Get site details
        site_info = service.sites().get(siteUrl=site_url).execute()
        
        result = {
            "site_url": site_url,
            "permission_level": site_info.get("permissionLevel", "Unknown"),
        }

        if "siteVerificationInfo" in site_info:
            verify_info = site_info["siteVerificationInfo"]
            result["verification"] = {
                "state": verify_info.get("verificationState", "Unknown"),
                "verified_user": verify_info.get("verifiedUser"),
                "method": verify_info.get("verificationMethod"),
            }

        if "ownershipInfo" in site_info:
            owner_info = site_info["ownershipInfo"]
            result["ownership"] = {
                "owner": owner_info.get("owner", "Unknown"),
                "verification_method": owner_info.get("verificationMethod"),
            }

        return json.dumps(result)
    except Exception as e:
        return f"Error retrieving site details: {str(e)}"

@mcp.tool()
async def get_sitemaps(site_url: str) -> str:
    """
    List all sitemaps for a specific Search Console property.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
    """
    try:
        service = get_gsc_service()
        
        # Get sitemaps list
        sitemaps = service.sitemaps().list(siteUrl=site_url).execute()
        
        if not sitemaps.get("sitemap"):
            return f"No sitemaps found for {site_url}."

        sitemap_list = []
        for sitemap in sitemaps.get("sitemap", []):
            last_downloaded = sitemap.get("lastDownloaded")
            if last_downloaded:
                try:
                    dt = datetime.fromisoformat(last_downloaded.replace("Z", "+00:00"))
                    last_downloaded = dt.strftime("%Y-%m-%d %H:%M")
                except Exception:
                    pass

            errors = int(sitemap.get("errors", 0))
            warnings = int(sitemap.get("warnings", 0))

            status = "Valid"
            if errors > 0:
                status = "Has errors"
            elif warnings > 0:
                status = "Has warnings"

            indexed_urls = None
            if "contents" in sitemap:
                for content in sitemap["contents"]:
                    if content.get("type") == "web":
                        indexed_urls = content.get("submitted")
                        break

            sitemap_list.append({
                "path": sitemap.get("path", "Unknown"),
                "last_downloaded": last_downloaded,
                "status": status,
                "indexed_urls": indexed_urls,
                "errors": errors,
                "warnings": warnings,
            })

        return json.dumps({
            "site_url": site_url,
            "count": len(sitemap_list),
            "sitemaps": sitemap_list,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error retrieving sitemaps: {str(e)}"

@mcp.tool()
async def inspect_url_enhanced(site_url: str, page_url: str) -> str:
    """
    Enhanced URL inspection to check indexing status and rich results in Google.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        page_url: The specific URL to inspect
    """
    try:
        service = get_gsc_service()
        
        # Build request
        request = {
            "inspectionUrl": page_url,
            "siteUrl": site_url
        }
        
        # Execute request
        response = service.urlInspection().index().inspect(body=request).execute()
        
        if not response or "inspectionResult" not in response:
            return f"No inspection data found for {page_url}."

        inspection = response["inspectionResult"]
        index_status = inspection.get("indexStatusResult", {})

        last_crawled = None
        if "lastCrawlTime" in index_status:
            try:
                crawl_time = datetime.fromisoformat(index_status["lastCrawlTime"].replace("Z", "+00:00"))
                last_crawled = crawl_time.strftime("%Y-%m-%d %H:%M")
            except Exception:
                last_crawled = index_status["lastCrawlTime"]

        rich_results = None
        if "richResultsResult" in inspection:
            rich = inspection["richResultsResult"]
            rich_results = {
                "verdict": rich.get("verdict", "UNKNOWN"),
                "detected_types": [
                    item.get("richResultType", "Unknown")
                    for item in rich.get("detectedItems", [])
                ],
                "issues": [
                    {"severity": issue.get("severity"), "message": issue.get("message")}
                    for issue in rich.get("richResultsIssues", [])
                ],
            }

        return json.dumps({
            "page_url": page_url,
            "site_url": site_url,
            "inspection_result_link": inspection.get("inspectionResultLink"),
            "verdict": index_status.get("verdict", "UNKNOWN"),
            "coverage_state": index_status.get("coverageState"),
            "last_crawled": last_crawled,
            "page_fetch_state": index_status.get("pageFetchState"),
            "robots_txt_state": index_status.get("robotsTxtState"),
            "indexing_state": index_status.get("indexingState"),
            "google_canonical": index_status.get("googleCanonical"),
            "user_canonical": index_status.get("userCanonical"),
            "crawled_as": index_status.get("crawledAs"),
            "referring_urls": index_status.get("referringUrls", [])[:5],
            "rich_results": rich_results,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error inspecting URL: {str(e)}"

@mcp.tool()
async def batch_url_inspection(site_url: str, urls: str) -> str:
    """
    Inspect multiple URLs in batch (within API limits).
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        urls: List of URLs to inspect, one per line
    """
    try:
        service = get_gsc_service()
        
        # Parse URLs
        url_list = [url.strip() for url in urls.split('\n') if url.strip()]
        
        if not url_list:
            return "No URLs provided for inspection."
        
        if len(url_list) > 10:
            return f"Too many URLs provided ({len(url_list)}). Please limit to 10 URLs per batch to avoid API quota issues."
        
        # Process each URL
        results = []
        
        for page_url in url_list:
            # Build request
            request = {
                "inspectionUrl": page_url,
                "siteUrl": site_url
            }
            
            try:
                # Execute request with a small delay to avoid rate limits
                response = service.urlInspection().index().inspect(body=request).execute()
                
                if not response or "inspectionResult" not in response:
                    results.append(f"{page_url}: No inspection data found")
                    continue
                
                inspection = response["inspectionResult"]
                index_status = inspection.get("indexStatusResult", {})
                
                # Get key information
                verdict = index_status.get("verdict", "UNKNOWN")
                coverage = index_status.get("coverageState", "Unknown")
                last_crawl = "Never"
                
                if "lastCrawlTime" in index_status:
                    try:
                        crawl_time = datetime.fromisoformat(index_status["lastCrawlTime"].replace('Z', '+00:00'))
                        last_crawl = crawl_time.strftime('%Y-%m-%d')
                    except:
                        last_crawl = index_status["lastCrawlTime"]
                
                # Check for rich results
                rich_results = "None"
                if "richResultsResult" in inspection:
                    rich = inspection["richResultsResult"]
                    if rich.get("verdict") == "PASS" and "detectedItems" in rich and rich["detectedItems"]:
                        rich_types = [item.get("richResultType", "Unknown") for item in rich["detectedItems"]]
                        rich_results = ", ".join(rich_types)
                
                results.append({
                    "url": page_url,
                    "verdict": verdict,
                    "coverage_state": coverage,
                    "last_crawled": last_crawl,
                    "rich_results": rich_results,
                })

            except Exception as e:
                results.append({"url": page_url, "error": str(e)})

        return json.dumps({
            "site_url": site_url,
            "count": len(results),
            "results": results,
        })

    except Exception as e:
        return f"Error performing batch inspection: {str(e)}"

@mcp.tool()
async def check_indexing_issues(site_url: str, urls: str) -> str:
    """
    Check for specific indexing issues across multiple URLs.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        urls: List of URLs to check, one per line
    """
    try:
        service = get_gsc_service()
        
        # Parse URLs
        url_list = [url.strip() for url in urls.split('\n') if url.strip()]
        
        if not url_list:
            return "No URLs provided for inspection."
        
        if len(url_list) > 10:
            return f"Too many URLs provided ({len(url_list)}). Please limit to 10 URLs per batch to avoid API quota issues."
        
        # Track issues by category
        issues_summary = {
            "not_indexed": [],
            "canonical_issues": [],
            "robots_blocked": [],
            "fetch_issues": [],
            "indexed": []
        }
        
        # Process each URL
        for page_url in url_list:
            # Build request
            request = {
                "inspectionUrl": page_url,
                "siteUrl": site_url
            }
            
            try:
                # Execute request
                response = service.urlInspection().index().inspect(body=request).execute()
                
                if not response or "inspectionResult" not in response:
                    issues_summary["not_indexed"].append(f"{page_url} - No inspection data found")
                    continue
                
                inspection = response["inspectionResult"]
                index_status = inspection.get("indexStatusResult", {})
                
                # Check indexing status
                verdict = index_status.get("verdict", "UNKNOWN")
                coverage = index_status.get("coverageState", "Unknown")
                
                if verdict != "PASS" or "not indexed" in coverage.lower() or "excluded" in coverage.lower():
                    issues_summary["not_indexed"].append(f"{page_url} - {coverage}")
                else:
                    issues_summary["indexed"].append(page_url)
                
                # Check canonical issues
                google_canonical = index_status.get("googleCanonical", "")
                user_canonical = index_status.get("userCanonical", "")
                
                if google_canonical and user_canonical and google_canonical != user_canonical:
                    issues_summary["canonical_issues"].append(
                        f"{page_url} - Google chose: {google_canonical} instead of user-declared: {user_canonical}"
                    )
                
                # Check robots.txt status
                robots_state = index_status.get("robotsTxtState", "")
                if robots_state == "BLOCKED":
                    issues_summary["robots_blocked"].append(page_url)
                
                # Check fetch issues
                fetch_state = index_status.get("pageFetchState", "")
                if fetch_state != "SUCCESSFUL":
                    issues_summary["fetch_issues"].append(f"{page_url} - {fetch_state}")
            
            except Exception as e:
                issues_summary["not_indexed"].append(f"{page_url} - Error: {str(e)}")
        
        return json.dumps({
            "site_url": site_url,
            "summary": {
                "total_checked": len(url_list),
                "indexed": len(issues_summary["indexed"]),
                "not_indexed": len(issues_summary["not_indexed"]),
                "canonical_issues": len(issues_summary["canonical_issues"]),
                "robots_blocked": len(issues_summary["robots_blocked"]),
                "fetch_issues": len(issues_summary["fetch_issues"]),
            },
            "issues": {
                "not_indexed": issues_summary["not_indexed"],
                "canonical_issues": issues_summary["canonical_issues"],
                "robots_blocked": issues_summary["robots_blocked"],
                "fetch_issues": issues_summary["fetch_issues"],
            },
            "indexed_urls": issues_summary["indexed"],
        })

    except Exception as e:
        return f"Error checking indexing issues: {str(e)}"

@mcp.tool()
async def get_performance_overview(site_url: str, days: int = 28) -> str:
    """
    Get a performance overview for a specific property.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        days: Number of days to look back (default: 28)
    """
    try:
        service = get_gsc_service()
        
        # Calculate date range
        end_date = datetime.now().date()
        start_date = end_date - timedelta(days=days)
        
        # Get total metrics
        total_request = {
            "startDate": start_date.strftime("%Y-%m-%d"),
            "endDate": end_date.strftime("%Y-%m-%d"),
            "dimensions": [],  # No dimensions for totals
            "rowLimit": 1,
            "dataState": DATA_STATE
        }
        
        total_response = service.searchanalytics().query(siteUrl=site_url, body=total_request).execute()
        
        # Get by date for trend
        date_request = {
            "startDate": start_date.strftime("%Y-%m-%d"),
            "endDate": end_date.strftime("%Y-%m-%d"),
            "dimensions": ["date"],
            "rowLimit": days,
            "dataState": DATA_STATE
        }
        
        date_response = service.searchanalytics().query(siteUrl=site_url, body=date_request).execute()
        
        if not total_response.get("rows"):
            return f"No performance data available for {site_url} in the last {days} days."

        totals_row = total_response["rows"][0]
        totals = {
            "clicks": totals_row.get("clicks", 0),
            "impressions": totals_row.get("impressions", 0),
            "ctr": round(totals_row.get("ctr", 0), 4),
            "position": round(totals_row.get("position", 0), 1),
        }

        daily_trend = []
        if date_response.get("rows"):
            sorted_rows = sorted(date_response["rows"], key=lambda x: x["keys"][0])
            for row in sorted_rows:
                daily_trend.append({
                    "date": row["keys"][0],
                    "clicks": row.get("clicks", 0),
                    "impressions": row.get("impressions", 0),
                    "ctr": round(row.get("ctr", 0), 4),
                    "position": round(row.get("position", 0), 1),
                })

        return json.dumps({
            "site_url": site_url,
            "date_range": {
                "start": start_date.strftime("%Y-%m-%d"),
                "end": end_date.strftime("%Y-%m-%d"),
                "days": days,
            },
            "totals": totals,
            "daily_trend": daily_trend,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error retrieving performance overview: {str(e)}"

@mcp.tool()
async def get_advanced_search_analytics(
    site_url: str, 
    start_date: str = None, 
    end_date: str = None, 
    dimensions: str = "query", 
    search_type: str = "WEB",
    row_limit: int = 1000,
    start_row: int = 0,
    sort_by: str = "clicks",
    sort_direction: str = "descending",
    filter_dimension: str = None,
    filter_operator: str = "contains", 
    filter_expression: str = None,
    filters: str = None,
    data_state: str = None
) -> str:
    """
    Get advanced search analytics data with sorting, filtering, and pagination.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        start_date: Start date in YYYY-MM-DD format (defaults to 28 days ago)
        end_date: End date in YYYY-MM-DD format (defaults to today)
        dimensions: Dimensions to group by, comma-separated (e.g., "query,page,device")
        search_type: Type of search results (WEB, IMAGE, VIDEO, NEWS, DISCOVER)
        row_limit: Maximum number of rows to return (max 25000)
        start_row: Starting row for pagination
        sort_by: Metric to sort by (clicks, impressions, ctr, position)
        sort_direction: Sort direction (ascending or descending)
        filter_dimension: Single filter dimension (query, page, country, device). Use 'filters' instead for multiple filters.
        filter_operator: Single filter operator (contains, equals, notContains, notEquals)
        filter_expression: Single filter expression value
        filters: JSON array of filter objects for AND logic across multiple dimensions. Overrides
                 filter_dimension/filter_operator/filter_expression when provided. Each object must
                 have 'dimension', 'operator', and 'expression' keys. Valid dimensions: query, page,
                 country, device. Valid operators: contains, equals, notContains, notEquals.
                 Example: [{"dimension":"country","operator":"equals","expression":"usa"},
                           {"dimension":"device","operator":"equals","expression":"MOBILE"}]
        data_state: Data freshness — "all" (default, matches GSC dashboard) or "final" (confirmed data only, 2-3 day lag)
    """
    try:
        service = get_gsc_service()
        
        # Calculate date range if not provided
        if not end_date:
            end_date = datetime.now().date().strftime("%Y-%m-%d")
        if not start_date:
            start_date = (datetime.now().date() - timedelta(days=28)).strftime("%Y-%m-%d")
        
        # Resolve and validate data_state (per-call override or fall back to global setting)
        resolved_data_state = (data_state or DATA_STATE).lower().strip()
        if resolved_data_state not in ("all", "final"):
            return (
                f"Invalid data_state value '{data_state}'. "
                "Accepted values are 'all' (matches GSC dashboard) or 'final' (2-3 day lag)."
            )
        
        # Parse dimensions
        dimension_list = [d.strip() for d in dimensions.split(",")]
        
        # Build request
        request = {
            "startDate": start_date,
            "endDate": end_date,
            "dimensions": dimension_list,
            "rowLimit": min(row_limit, 25000),  # Cap at API maximum
            "startRow": start_row,
            "searchType": search_type.upper(),
            "dataState": resolved_data_state
        }
        
        # Add sorting
        if sort_by:
            metric_map = {
                "clicks": "CLICK_COUNT",
                "impressions": "IMPRESSION_COUNT",
                "ctr": "CTR",
                "position": "POSITION"
            }
            
            if sort_by in metric_map:
                request["orderBy"] = [{
                    "metric": metric_map[sort_by],
                    "direction": sort_direction.lower()
                }]
        
        # Build filter groups — multi-filter JSON takes priority over single-filter params
        active_filters = []
        if filters:
            try:
                filter_list = json.loads(filters)
            except json.JSONDecodeError:
                return "Invalid filters JSON. Please provide a valid JSON array of filter objects."
            if not isinstance(filter_list, list) or len(filter_list) == 0:
                return "Invalid filters value. Expected a non-empty JSON array of filter objects."
            for f in filter_list:
                if not all(k in f for k in ("dimension", "operator", "expression")):
                    return (
                        "Each filter object must have 'dimension', 'operator', and 'expression' keys. "
                        f"Invalid filter: {f}"
                    )
            request["dimensionFilterGroups"] = [{"filters": filter_list}]
            active_filters = filter_list
        elif filter_dimension and filter_expression:
            single_filter = {
                "dimension": filter_dimension,
                "operator": filter_operator,
                "expression": filter_expression
            }
            request["dimensionFilterGroups"] = [{"filters": [single_filter]}]
            active_filters = [single_filter]
        
        # Execute request
        response = service.searchanalytics().query(siteUrl=site_url, body=request).execute()
        
        if not response.get("rows"):
            no_data_msg = (
                f"No search analytics data found for {site_url} with the specified parameters.\n\n"
                f"Parameters used:\n"
                f"- Date range: {start_date} to {end_date}\n"
                f"- Dimensions: {dimensions}\n"
                f"- Search type: {search_type}\n"
            )
            if active_filters:
                no_data_msg += "- Filters:\n"
                for f in active_filters:
                    no_data_msg += f"    {f['dimension']} {f['operator']} '{f['expression']}'\n"
            else:
                no_data_msg += "- No filter applied\n"
            return no_data_msg
        
        rows = []
        for row in response.get("rows", []):
            entry = {}
            for i, dim in enumerate(dimension_list):
                entry[dim] = row.get("keys", [])[i] if i < len(row.get("keys", [])) else None
            entry["clicks"] = row.get("clicks", 0)
            entry["impressions"] = row.get("impressions", 0)
            entry["ctr"] = round(row.get("ctr", 0), 4)
            entry["position"] = round(row.get("position", 0), 1)
            rows.append(entry)

        has_more = len(response.get("rows", [])) == row_limit
        return json.dumps({
            "site_url": site_url,
            "date_range": {"start": start_date, "end": end_date},
            "search_type": search_type,
            "dimensions": dimension_list,
            "filters_applied": active_filters,
            "pagination": {
                "start_row": start_row,
                "row_count": len(rows),
                "has_more": has_more,
                "next_start_row": start_row + row_limit if has_more else None,
            },
            "rows": rows,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error retrieving advanced search analytics: {str(e)}"

@mcp.tool()
async def compare_search_periods(
    site_url: str,
    period1_start: str,
    period1_end: str,
    period2_start: str,
    period2_end: str,
    dimensions: str = "query",
    limit: int = 10
) -> str:
    """
    Compare search analytics data between two time periods.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        period1_start: Start date for period 1 (YYYY-MM-DD)
        period1_end: End date for period 1 (YYYY-MM-DD)
        period2_start: Start date for period 2 (YYYY-MM-DD)
        period2_end: End date for period 2 (YYYY-MM-DD)
        dimensions: Dimensions to group by (default: query)
        limit: Number of top results to compare (default: 10)
    """
    try:
        service = get_gsc_service()
        
        # Parse dimensions
        dimension_list = [d.strip() for d in dimensions.split(",")]
        
        # Build requests for both periods
        period1_request = {
            "startDate": period1_start,
            "endDate": period1_end,
            "dimensions": dimension_list,
            "rowLimit": 1000,  # Get more to ensure we can match items between periods
            "dataState": DATA_STATE
        }
        
        period2_request = {
            "startDate": period2_start,
            "endDate": period2_end,
            "dimensions": dimension_list,
            "rowLimit": 1000,
            "dataState": DATA_STATE
        }
        
        # Execute requests
        period1_response = service.searchanalytics().query(siteUrl=site_url, body=period1_request).execute()
        period2_response = service.searchanalytics().query(siteUrl=site_url, body=period2_request).execute()
        
        period1_rows = period1_response.get("rows", [])
        period2_rows = period2_response.get("rows", [])
        
        if not period1_rows and not period2_rows:
            return f"No data found for either period for {site_url}."
        
        # Create dictionaries for easy lookup
        period1_data = {tuple(row.get("keys", [])): row for row in period1_rows}
        period2_data = {tuple(row.get("keys", [])): row for row in period2_rows}
        
        # Find common keys and calculate differences
        all_keys = set(period1_data.keys()) | set(period2_data.keys())
        comparison_data = []
        
        for key in all_keys:
            p1_row = period1_data.get(key, {"clicks": 0, "impressions": 0, "ctr": 0, "position": 0})
            p2_row = period2_data.get(key, {"clicks": 0, "impressions": 0, "ctr": 0, "position": 0})
            
            # Calculate differences
            click_diff = p2_row.get("clicks", 0) - p1_row.get("clicks", 0)
            click_pct = (click_diff / p1_row.get("clicks", 1)) * 100 if p1_row.get("clicks", 0) > 0 else float('inf')
            
            imp_diff = p2_row.get("impressions", 0) - p1_row.get("impressions", 0)
            imp_pct = (imp_diff / p1_row.get("impressions", 1)) * 100 if p1_row.get("impressions", 0) > 0 else float('inf')
            
            ctr_diff = p2_row.get("ctr", 0) - p1_row.get("ctr", 0)
            pos_diff = p1_row.get("position", 0) - p2_row.get("position", 0)  # Note: lower position is better
            
            comparison_data.append({
                "key": key,
                "p1_clicks": p1_row.get("clicks", 0),
                "p2_clicks": p2_row.get("clicks", 0),
                "click_diff": click_diff,
                "click_pct": click_pct,
                "p1_impressions": p1_row.get("impressions", 0),
                "p2_impressions": p2_row.get("impressions", 0),
                "imp_diff": imp_diff,
                "imp_pct": imp_pct,
                "p1_ctr": p1_row.get("ctr", 0),
                "p2_ctr": p2_row.get("ctr", 0),
                "ctr_diff": ctr_diff,
                "p1_position": p1_row.get("position", 0),
                "p2_position": p2_row.get("position", 0),
                "pos_diff": pos_diff
            })
        
        # Sort by absolute click difference
        comparison_data.sort(key=lambda x: abs(x["click_diff"]), reverse=True)

        serialisable = []
        for item in comparison_data[:limit]:
            click_pct = item["click_pct"] if item["click_pct"] != float("inf") else None
            imp_pct = item["imp_pct"] if item["imp_pct"] != float("inf") else None
            serialisable.append({
                "key": list(item["key"]),
                "p1_clicks": item["p1_clicks"],
                "p2_clicks": item["p2_clicks"],
                "click_diff": item["click_diff"],
                "click_pct": round(click_pct, 1) if click_pct is not None else None,
                "p1_impressions": item["p1_impressions"],
                "p2_impressions": item["p2_impressions"],
                "imp_diff": item["imp_diff"],
                "imp_pct": round(imp_pct, 1) if imp_pct is not None else None,
                "p1_ctr": round(item["p1_ctr"], 4),
                "p2_ctr": round(item["p2_ctr"], 4),
                "ctr_diff": round(item["ctr_diff"], 4),
                "p1_position": round(item["p1_position"], 1),
                "p2_position": round(item["p2_position"], 1),
                "position_diff": round(item["pos_diff"], 1),
            })

        return json.dumps({
            "site_url": site_url,
            "period1": {"start": period1_start, "end": period1_end},
            "period2": {"start": period2_start, "end": period2_end},
            "dimensions": dimension_list,
            "total_items": len(comparison_data),
            "showing": len(serialisable),
            "comparison": serialisable,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error comparing search periods: {str(e)}"

@mcp.tool()
async def get_search_by_page_query(
    site_url: str,
    page_url: str,
    days: int = 28,
    row_limit: int = 20
) -> str:
    """
    Get search analytics data for a specific page, broken down by query.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        page_url: The specific page URL to analyze
        days: Number of days to look back (default: 28)
        row_limit: Number of rows to return (default: 20, max: 500). Use 5-20 for quick overviews,
                   50-200 for deeper analysis, up to 500 for comprehensive reports. For bulk exports
                   beyond 500 rows, use get_advanced_search_analytics which supports pagination.
    """
    try:
        service = get_gsc_service()
        
        # Calculate date range
        end_date = datetime.now().date()
        start_date = end_date - timedelta(days=days)
        
        # Build request with page filter
        request = {
            "startDate": start_date.strftime("%Y-%m-%d"),
            "endDate": end_date.strftime("%Y-%m-%d"),
            "dimensions": ["query"],
            "dimensionFilterGroups": [{
                "filters": [{
                    "dimension": "page",
                    "operator": "equals",
                    "expression": page_url
                }]
            }],
            "rowLimit": min(max(1, row_limit), 500),
            "orderBy": [{"metric": "CLICK_COUNT", "direction": "descending"}],
            "dataState": DATA_STATE
        }
        
        # Execute request
        response = service.searchanalytics().query(siteUrl=site_url, body=request).execute()
        
        if not response.get("rows"):
            return f"No search data found for page {page_url} in the last {days} days."
        
        rows = []
        for row in response.get("rows", []):
            rows.append({
                "query": row.get("keys", ["Unknown"])[0],
                "clicks": row.get("clicks", 0),
                "impressions": row.get("impressions", 0),
                "ctr": round(row.get("ctr", 0), 4),
                "position": round(row.get("position", 0), 1),
            })

        total_clicks = sum(r["clicks"] for r in rows)
        total_impressions = sum(r["impressions"] for r in rows)

        return json.dumps({
            "site_url": site_url,
            "page_url": page_url,
            "date_range": {
                "start": start_date.strftime("%Y-%m-%d"),
                "end": end_date.strftime("%Y-%m-%d"),
                "days": days,
            },
            "totals": {
                "clicks": total_clicks,
                "impressions": total_impressions,
                "avg_ctr": round(total_clicks / total_impressions, 4) if total_impressions > 0 else 0,
            },
            "row_count": len(rows),
            "rows": rows,
        })
    except Exception as e:
        return f"Error retrieving page query data: {str(e)}"

@mcp.tool()
async def list_sitemaps_enhanced(site_url: str, sitemap_index: str = None) -> str:
    """
    List all sitemaps for a specific Search Console property with detailed information.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        sitemap_index: Optional sitemap index URL to list child sitemaps
    """
    try:
        service = get_gsc_service()
        
        # Get sitemaps list
        if sitemap_index:
            sitemaps = service.sitemaps().list(siteUrl=site_url, sitemapIndex=sitemap_index).execute()
            source = f"child sitemaps from index: {sitemap_index}"
        else:
            sitemaps = service.sitemaps().list(siteUrl=site_url).execute()
            source = "all submitted sitemaps"
        
        if not sitemaps.get("sitemap"):
            return f"No sitemaps found for {site_url}" + (f" in index {sitemap_index}" if sitemap_index else ".")

        def _fmt_date(raw):
            if not raw:
                return None
            try:
                return datetime.fromisoformat(raw.replace("Z", "+00:00")).strftime("%Y-%m-%d %H:%M")
            except Exception:
                return raw

        sitemap_list = []
        for sitemap in sitemaps.get("sitemap", []):
            errors = int(sitemap.get("errors", 0))
            warnings = int(sitemap.get("warnings", 0))
            url_count = None
            if "contents" in sitemap:
                for content in sitemap["contents"]:
                    if content.get("type") == "web":
                        url_count = content.get("submitted")
                        break
            sitemap_list.append({
                "path": sitemap.get("path", "Unknown"),
                "last_submitted": _fmt_date(sitemap.get("lastSubmitted")),
                "last_downloaded": _fmt_date(sitemap.get("lastDownloaded")),
                "type": "Index" if sitemap.get("isSitemapsIndex", False) else "Sitemap",
                "is_pending": sitemap.get("isPending", False),
                "url_count": url_count,
                "errors": errors,
                "warnings": warnings,
            })

        pending_count = sum(1 for s in sitemap_list if s["is_pending"])

        return json.dumps({
            "site_url": site_url,
            "sitemap_index": sitemap_index,
            "count": len(sitemap_list),
            "pending_count": pending_count,
            "sitemaps": sitemap_list,
        })
    except Exception as e:
        if "404" in str(e):
            return _site_not_found_error(site_url)
        return f"Error retrieving sitemaps: {str(e)}"

@mcp.tool()
async def get_sitemap_details(site_url: str, sitemap_url: str) -> str:
    """
    Get detailed information about a specific sitemap.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        sitemap_url: The full URL of the sitemap to inspect
    """
    try:
        service = get_gsc_service()
        
        # Get sitemap details
        details = service.sitemaps().get(siteUrl=site_url, feedpath=sitemap_url).execute()
        
        if not details:
            return f"No details found for sitemap {sitemap_url}."

        def _fmt_date(raw):
            if not raw:
                return None
            try:
                return datetime.fromisoformat(raw.replace("Z", "+00:00")).strftime("%Y-%m-%d %H:%M")
            except Exception:
                return raw

        is_index = details.get("isSitemapsIndex", False)
        content_breakdown = [
            {
                "type": content.get("type", "unknown").upper(),
                "submitted": content.get("submitted", 0),
                "indexed": content.get("indexed"),
            }
            for content in details.get("contents", [])
        ]

        return json.dumps({
            "sitemap_url": sitemap_url,
            "site_url": site_url,
            "type": "Index" if is_index else "Sitemap",
            "status": "pending" if details.get("isPending", False) else "processed",
            "last_submitted": _fmt_date(details.get("lastSubmitted")),
            "last_downloaded": _fmt_date(details.get("lastDownloaded")),
            "errors": int(details.get("errors", 0)),
            "warnings": int(details.get("warnings", 0)),
            "content_breakdown": content_breakdown,
            "is_index": is_index,
        })
    except Exception as e:
        return f"Error retrieving sitemap details: {str(e)}"

@mcp.tool()
async def submit_sitemap(site_url: str, sitemap_url: str) -> str:
    """
    Submit a new sitemap or resubmit an existing one to Google.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        sitemap_url: The full URL of the sitemap to submit
    """
    try:
        service = get_gsc_service()
        
        # Submit the sitemap
        service.sitemaps().submit(siteUrl=site_url, feedpath=sitemap_url).execute()
        
        # Verify submission by getting details
        try:
            details = service.sitemaps().get(siteUrl=site_url, feedpath=sitemap_url).execute()
            
            # Format response
            result_lines = [f"Successfully submitted sitemap: {sitemap_url}"]
            
            # Add submission time if available
            if "lastSubmitted" in details:
                try:
                    dt = datetime.fromisoformat(details["lastSubmitted"].replace('Z', '+00:00'))
                    result_lines.append(f"Submission time: {dt.strftime('%Y-%m-%d %H:%M')}")
                except:
                    result_lines.append(f"Submission time: {details['lastSubmitted']}")
            
            # Add processing status
            is_pending = details.get("isPending", True)
            result_lines.append(f"Status: {'Pending processing' if is_pending else 'Processing started'}")
            
            # Add note about processing time
            result_lines.append("\nNote: Google may take some time to process the sitemap. Check back later for full details.")
            
            return "\n".join(result_lines)
        except:
            # If we can't get details, just return basic success message
            return f"Successfully submitted sitemap: {sitemap_url}\n\nGoogle will queue it for processing."
    
    except Exception as e:
        return f"Error submitting sitemap: {str(e)}"

@mcp.tool()
async def delete_sitemap(site_url: str, sitemap_url: str) -> str:
    """
    Delete (unsubmit) a sitemap from Google Search Console.
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        sitemap_url: The full URL of the sitemap to delete
    """
    if not ALLOW_DESTRUCTIVE:
        return (
            "Safety: delete_sitemap permanently removes a sitemap from GSC. "
            "Set GSC_ALLOW_DESTRUCTIVE=true in your environment to enable add/delete tools."
        )
    try:
        service = get_gsc_service()
        
        # First check if the sitemap exists
        try:
            service.sitemaps().get(siteUrl=site_url, feedpath=sitemap_url).execute()
        except Exception as e:
            if "404" in str(e):
                return f"Sitemap not found: {sitemap_url}. It may have already been deleted or was never submitted."
            else:
                raise e
        
        # Delete the sitemap
        service.sitemaps().delete(siteUrl=site_url, feedpath=sitemap_url).execute()
        
        return f"Successfully deleted sitemap: {sitemap_url}\n\nNote: This only removes the sitemap from Search Console. Any URLs already indexed will remain in Google's index."
    
    except Exception as e:
        return f"Error deleting sitemap: {str(e)}"

@mcp.tool()
async def manage_sitemaps(site_url: str, action: str, sitemap_url: str = None, sitemap_index: str = None) -> str:
    """
    All-in-one tool to manage sitemaps (list, get details, submit, delete).
    
    Args:
        site_url: Exact GSC property URL from list_properties (e.g. "https://example.com/" or
                  "sc-domain:example.com"). Domain properties cover all subdomains — use the
                  domain property as site_url and filter by page to analyze a specific subdomain.
        action: The action to perform (list, details, submit, delete)
        sitemap_url: The full URL of the sitemap (required for details, submit, delete)
        sitemap_index: Optional sitemap index URL for listing child sitemaps (only used with 'list' action)
    """
    try:
        # Validate inputs
        action = action.lower().strip()
        valid_actions = ["list", "details", "submit", "delete"]
        
        if action not in valid_actions:
            return f"Invalid action: {action}. Please use one of: {', '.join(valid_actions)}"
        
        if action in ["details", "submit", "delete"] and not sitemap_url:
            return f"The {action} action requires a sitemap_url parameter."
        
        # Perform the requested action
        if action == "list":
            return await list_sitemaps_enhanced(site_url, sitemap_index)
        elif action == "details":
            return await get_sitemap_details(site_url, sitemap_url)
        elif action == "submit":
            return await submit_sitemap(site_url, sitemap_url)
        elif action == "delete":
            return await delete_sitemap(site_url, sitemap_url)
    
    except Exception as e:
        return f"Error managing sitemaps: {str(e)}"

@mcp.tool()
async def get_creator_info() -> str:
    """
    Provides information about Amin Foroutan, the creator of the MCP-GSC tool.
    """
    creator_info = """
# About the Creator: Amin Foroutan

Amin Foroutan is an SEO consultant with over a decade of experience, specializing in technical SEO, Python-driven tools, and data analysis for SEO performance.

## Connect with Amin:

- **LinkedIn**: [Amin Foroutan](https://www.linkedin.com/in/ma-foroutan/)
- **Personal Website**: [aminforoutan.com](https://aminforoutan.com/)
- **YouTube**: [Amin Forout](https://www.youtube.com/channel/UCW7tPXg-rWdH4YzLrcAdBIw)
- **X (Twitter)**: [@aminfseo](https://x.com/aminfseo)

## Notable Projects:

Amin has created several popular SEO tools including:
- Advanced GSC Visualizer (6.4K+ users)
- SEO Render Insight Tool (3.5K+ users)
- Google AI Overview Impact Analysis (1.2K+ users)
- Google AI Overview Citation Analysis (900+ users)
- SEMRush Enhancer (570+ users)
- SEO Page Inspector (115+ users)

## Expertise:

Amin combines technical SEO knowledge with programming skills to create innovative solutions for SEO challenges.
"""
    return creator_info

@mcp.tool()
async def reauthenticate() -> str:
    """
    Perform a logout and new login sequence.
    Deletes the current OAuth token file and triggers the browser authentication flow.
    Useful when you need to switch to a different Google account.
    """
    try:
        # Delete existing token to force re-authentication
        if os.path.exists(TOKEN_FILE):
            os.remove(TOKEN_FILE)
            token_deleted = True
        else:
            token_deleted = False

        # Check if OAuth client secrets file exists
        if not os.path.exists(OAUTH_CLIENT_SECRETS_FILE):
            return (
                "Error: OAuth client secrets file not found. "
                "Cannot start new authentication flow. "
                "Please ensure client_secrets.json is present or set the "
                "GSC_OAUTH_CLIENT_SECRETS_FILE environment variable."
            )

        # Trigger new OAuth flow — opens a browser window on the local machine.
        # run_local_server() works on macOS even from an MCP subprocess because
        # macOS can open browsers via webbrowser.open() regardless of TTY state.
        flow = InstalledAppFlow.from_client_secrets_file(OAUTH_CLIENT_SECRETS_FILE, SCOPES)
        creds = flow.run_local_server(port=0)

        # Save the new credentials for future use
        with open(TOKEN_FILE, "w") as token:
            token.write(creds.to_json())

        msg = "Successfully authenticated with a new Google account."
        if token_deleted:
            msg = "Previous session deleted. " + msg
        return msg

    except Exception as e:
        return f"Error during reauthentication: {str(e)}"


def main():
    """Entry point for the MCP server. Supports stdio (default) and SSE transports."""
    transport = os.environ.get("MCP_TRANSPORT", "stdio").lower()
    host = os.environ.get("MCP_HOST", "127.0.0.1")
    try:
        port = int(os.environ.get("MCP_PORT", "3001"))
    except ValueError:
        raise ValueError("MCP_PORT must be an integer")

    if transport == "stdio":
        mcp.run(transport="stdio")
    elif transport in {"sse", "http"}:
        # mcp SDK >= 1.27 removed the host/port kwargs from run() (they must be
        # set on mcp.settings instead) and enabled DNS-rebinding protection with
        # a Host allowlist of localhost only — which 421s the remote/Docker
        # binding the operator explicitly opted into via MCP_HOST. Disable that
        # check for the documented remote path. See issues #30 and #33.
        mcp.settings.host = host
        mcp.settings.port = port
        try:
            mcp.settings.transport_security.enable_dns_rebinding_protection = False
        except Exception:
            pass
        mcp.run(transport="sse")
    else:
        raise ValueError(
            f"Unknown MCP_TRANSPORT '{transport}'. "
            "Use 'stdio' (default) or 'sse'."
        )


if __name__ == "__main__":
    main()