# main.py (Modified to add crawl4ai and adjust fetching logic) import os import re import logging import asyncio import json import html import contextlib import traceback from typing import Optional, Dict, Any, Tuple # --- Frameworks --- from starlette.applications import Starlette from starlette.routing import Route from starlette.responses import PlainTextResponse, JSONResponse, Response from starlette.requests import Request # --- Telegram Bot --- from telegram import Update, InlineKeyboardButton, InlineKeyboardMarkup, Bot from telegram.ext import ( Application, CommandHandler, MessageHandler, filters, ContextTypes, CallbackQueryHandler, ) from telegram.constants import ParseMode from telegram.error import NetworkError, RetryAfter, TimedOut, BadRequest, TelegramError from telegram.request import HTTPXRequest, BaseRequest # --- Other Libraries --- import httpx from youtube_transcript_api import YouTubeTranscriptApi, TranscriptsDisabled, NoTranscriptFound from bs4 import BeautifulSoup from tenacity import retry, stop_after_attempt, wait_exponential, retry_if_exception_type, before_sleep_log try: import lxml DEFAULT_PARSER = 'lxml' except ImportError: DEFAULT_PARSER = 'html.parser' # --- Google Gemini --- try: import google.generativeai as genai from google.generativeai.types import HarmCategory, HarmBlockThreshold _gemini_available = True except ImportError: genai = None HarmCategory = None HarmBlockThreshold = None _gemini_available = False # logger will be defined later, log warning after logger setup # --- Crawl4AI (New Primary Web Scraper) --- try: from crawl4ai import AsyncWebCrawler _crawl4ai_available = True except ImportError: AsyncWebCrawler = None _crawl4ai_available = False # logger will be defined later # --- Logging Setup --- logging.basicConfig( format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', level=logging.INFO ) logging.getLogger("httpx").setLevel(logging.WARNING) logging.getLogger("telegram.ext").setLevel(logging.INFO) logging.getLogger('telegram.bot').setLevel(logging.INFO) logging.getLogger("urllib3").setLevel(logging.INFO) logging.getLogger('gunicorn.error').setLevel(logging.INFO) logging.getLogger('uvicorn').setLevel(logging.INFO) logging.getLogger('starlette').setLevel(logging.INFO) if _gemini_available: logging.getLogger("google.ai.generativelanguage").setLevel(logging.WARNING) # Reduce crawl4ai logging noise if needed if _crawl4ai_available: logging.getLogger("crawl4ai").setLevel(logging.WARNING) # Or INFO for more detail logging.getLogger("playwright").setLevel(logging.WARNING) logger = logging.getLogger(__name__) logger.info(f"Logging configured. Using BS4 parser: {DEFAULT_PARSER}") if not _gemini_available: logger.warning("google-generativeai library not found. Gemini functionality disabled.") if not _crawl4ai_available: logger.warning("crawl4ai library not found. Primary website scraping will be disabled.") # --- Global variable for PTB app --- ptb_app: Optional[Application] = None # --- Environment Variable Loading & Configuration --- logger.info("Attempting to load secrets and configuration...") def get_secret(secret_name): value = os.environ.get(secret_name) if value: status = "Found"; log_length = min(len(value), 8); value_start = value[:log_length]; logger.info(f"Secret '{secret_name}': {status} (Value starts with: {value_start}...)") else: status = "Not Found"; logger.warning(f"Secret '{secret_name}': {status}") return value TELEGRAM_TOKEN = get_secret('TELEGRAM_TOKEN') OPENROUTER_API_KEY = get_secret('OPENROUTER_API_KEY') # Now Fallback URLTOTEXT_API_KEY = get_secret('URLTOTEXT_API_KEY') # Fallback Web 2 SUPADATA_API_KEY = get_secret('SUPADATA_API_KEY') # Fallback YT 1 APIFY_API_TOKEN = get_secret('APIFY_API_TOKEN') # Fallback YT 2 WEBHOOK_SECRET = get_secret('WEBHOOK_SECRET') GEMINI_API_KEY = get_secret('GEMINI_API_KEY') # Primary Summarizer # Models (User can still configure via env vars) OPENROUTER_MODEL = os.environ.get("OPENROUTER_MODEL", "deepseek/deepseek-chat-v3-0324:free") # Fallback Summarizer Model APIFY_ACTOR_ID = os.environ.get("APIFY_ACTOR_ID", "karamelo~youtube-transcripts") GEMINI_MODEL = os.environ.get("GEMINI_MODEL", "gemini-2.0-flash-001") # Primary Summarizer Model if not TELEGRAM_TOKEN: logger.critical("❌ FATAL: TELEGRAM_TOKEN not found."); raise RuntimeError("Exiting: Telegram token missing.") if not GEMINI_API_KEY: logger.error("❌ ERROR: GEMINI_API_KEY not found. Primary summarization (Gemini) will fail.") if not OPENROUTER_API_KEY: logger.warning("⚠️ WARNING: OPENROUTER_API_KEY not found. Fallback summarization will fail.") _gemini_primary_enabled = _gemini_available and bool(GEMINI_API_KEY) if not _gemini_available: logger.warning("⚠️ WARNING: google-generativeai library missing. Gemini disabled.") elif not GEMINI_API_KEY: logger.warning("⚠️ WARNING: GEMINI_API_KEY not found or empty. Gemini disabled.") _openrouter_fallback_enabled = bool(OPENROUTER_API_KEY) if not _openrouter_fallback_enabled: logger.warning("⚠️ WARNING: OPENROUTER_API_KEY not found. Fallback summarization disabled.") _crawl4ai_primary_web_enabled = _crawl4ai_available if not _crawl4ai_primary_web_enabled: logger.warning("⚠️ WARNING: crawl4ai library not found. Primary web scraping disabled.") _bs4_fallback_web_enabled = True # Assumes bs4 is always available _urltotext_fallback_web_enabled = bool(URLTOTEXT_API_KEY) if not _urltotext_fallback_web_enabled: logger.info("ℹ️ INFO: URLTOTEXT_API_KEY not found. Secondary web fallback disabled.") # Fallback YT checks if not SUPADATA_API_KEY: logger.info("ℹ️ INFO: SUPADATA_API_KEY not found. First YT fallback disabled.") _apify_token_exists = bool(APIFY_API_TOKEN) if not _apify_token_exists: logger.info("ℹ️ INFO: APIFY_API_TOKEN not found. Second YT fallback disabled.") if not WEBHOOK_SECRET: logger.info("Optional secret 'WEBHOOK_SECRET' not found. Webhook security disabled.") logger.info("Secret loading and configuration check finished.") logger.info(f"Primary Web Scraper: {'Crawl4AI' if _crawl4ai_primary_web_enabled else 'DISABLED'}") logger.info(f"Fallback Web Scraper 1: {'BeautifulSoup' if _bs4_fallback_web_enabled else 'DISABLED'}") logger.info(f"Fallback Web Scraper 2: {'urltotext.com API' if _urltotext_fallback_web_enabled else 'DISABLED'}") logger.info(f"Using Gemini Model (Primary Summarizer): {GEMINI_MODEL if _gemini_primary_enabled else 'DISABLED'}") logger.info(f"Using OpenRouter Model (Fallback Summarizer): {OPENROUTER_MODEL if _openrouter_fallback_enabled else 'DISABLED'}") logger.info(f"Using Apify Actor (via REST): {APIFY_ACTOR_ID}") if _gemini_primary_enabled: try: genai.configure(api_key=GEMINI_API_KEY); logger.info("Google GenAI client configured successfully.") except Exception as e: logger.error(f"Failed to configure Google GenAI client: {e}"); _gemini_primary_enabled = False # --- Retry Decorator --- @retry( stop=stop_after_attempt(4), wait=wait_exponential(multiplier=1, min=2, max=15), retry=retry_if_exception_type((NetworkError, RetryAfter, TimedOut, BadRequest)), before_sleep=before_sleep_log(logger, logging.WARNING), reraise=True ) async def retry_bot_operation(func, *args, **kwargs): try: return await func(*args, **kwargs) except BadRequest as e: ignore_errors = [ "message is not modified", "query is too old", "message to edit not found", "chat not found", "bot was blocked by the user", ] if any(err in str(e).lower() for err in ignore_errors): logger.warning(f"Ignoring non-critical BadRequest: {e}"); return None logger.error(f"Potentially critical BadRequest: {e}"); raise except TelegramError as e: logger.warning(f"TelegramError (will retry if applicable): {e}"); raise except Exception as e: logger.error(f"Unexpected error during bot operation: {e}", exc_info=True); raise # --- Helper Functions --- def is_youtube_url(url): youtube_regex = re.compile( r'(?:https?://)?(?:www\.)?(?:m\.)?(?:youtube(?:-nocookie)?\.com|youtu\.be)/' r'(?:watch\?v=|embed/|v/|shorts/|live/|attribution_link\?a=.*&u=/watch\?v=)?' r'([\w-]{11})' r'(?:\S+)?', re.IGNORECASE) match = youtube_regex.search(url); logger.debug(f"is_youtube_url '{url}': {bool(match)}"); return bool(match) def extract_youtube_id(url): youtube_regex = re.compile( r'(?:https?://)?(?:www\.)?(?:m\.)?(?:youtube(?:-nocookie)?\.com|youtu\.be)/' r'(?:watch\?v=|embed/|v/|shorts/|live/|attribution_link\?a=.*&u=/watch\?v=)?' r'([\w-]{11})' r'(?:\S+)?', re.IGNORECASE) match = youtube_regex.search(url) if match: video_id = match.group(1); logger.debug(f"Extracted YT ID '{video_id}' from {url}"); return video_id else: logger.warning(f"Could not extract YT ID from {url}"); return None # --- Content Fetching Functions --- # --- YouTube Fetching (Unchanged) --- async def get_transcript_via_supadata(video_id: str, api_key: str) -> Optional[str]: if not video_id: logger.error("[Supadata] No video_id provided"); return None if not api_key: logger.error("[Supadata] API key missing."); return None logger.info(f"[Supadata] Attempting fetch for video ID: {video_id}") api_endpoint = "https://api.supadata.ai/v1/youtube/transcript" # Corrected URL params = {"videoId": video_id, "format": "text"}; headers = {"X-API-Key": api_key} try: async with httpx.AsyncClient(timeout=30.0) as client: response = await client.get(api_endpoint, headers=headers, params=params) logger.debug(f"[Supadata] Status code {response.status_code} for {video_id}") if response.status_code == 200: try: try: data = response.json() except json.JSONDecodeError: data = None content = None if data: content = data if isinstance(data, str) else data.get("transcript") or data.get("text") or data.get("data") if not content and response.text: content = response.text if content and isinstance(content, str): logger.info(f"[Supadata] Success for {video_id}. Length: {len(content)}"); return content.strip() else: logger.warning(f"[Supadata] Success but content empty/invalid for {video_id}. Response: {response.text[:200]}"); return None except Exception as e: logger.error(f"[Supadata] Error processing success response for {video_id}: {e}", exc_info=True); return None elif response.status_code in [401, 403]: logger.error(f"[Supadata] Auth error ({response.status_code}). Check API key."); return None elif response.status_code == 404: logger.warning(f"[Supadata] Not found (404) for {video_id}."); return None else: logger.error(f"[Supadata] Unexpected status {response.status_code} for {video_id}. Resp: {response.text[:200]}"); return None except httpx.TimeoutException: logger.error(f"[Supadata] Timeout connecting for {video_id}"); return None except httpx.RequestError as e: if "CERTIFICATE_VERIFY_FAILED" in str(e): logger.error(f"[Supadata] SSL Cert Verify Failed for {video_id}: {e}") else: logger.error(f"[Supadata] Request error for {video_id}: {e}") return None except Exception as e: logger.error(f"[Supadata] Unexpected error for {video_id}: {e}", exc_info=True); return None async def get_transcript_via_apify(video_url: str, api_token: str) -> Optional[str]: """Fetches YouTube transcript using Apify REST API (run-sync-get-dataset-items endpoint).""" global APIFY_ACTOR_ID if not video_url: logger.error("[Apify SyncItems] No video_url provided"); return None if not api_token: logger.error("[Apify SyncItems] API token missing."); return None logger.info(f"[Apify SyncItems] Attempting fetch for URL: {video_url} (Actor: {APIFY_ACTOR_ID})") sync_items_endpoint = f"https://api.apify.com/v2/acts/{APIFY_ACTOR_ID}/run-sync-get-dataset-items" params = {"token": api_token} payload = { "urls": [video_url], "outputFormat": "singleStringText", "maxRetries": 5, "channelHandleBoolean": False, "channelNameBoolean": False, "datePublishedBoolean": False, "relativeDateTextBoolean": False, } headers = {"Content-Type": "application/json"} try: async with httpx.AsyncClient(timeout=120.0) as client: log_headers = {k: v for k, v in headers.items()} logger.debug(f"[Apify SyncItems] POST Request Details:\nURL: {sync_items_endpoint}\nParams: {params}\nHeaders: {log_headers}\nPayload: {json.dumps(payload)}") response = await client.post(sync_items_endpoint, headers=headers, params=params, json=payload) logger.debug(f"[Apify SyncItems] Received status code {response.status_code} for {video_url}") if response.status_code == 200: try: results = response.json() if isinstance(results, list) and len(results) > 0: item = results[0] content = None if "captions" in item and isinstance(item["captions"], str): logger.info("[Apify SyncItems] Found 'captions' key with string content."); content = item["captions"] elif "text" in item and isinstance(item["text"], str): logger.info("[Apify SyncItems] Found 'text' key with string content."); content = item["text"] elif "transcript" in item and isinstance(item["transcript"], str): logger.info("[Apify SyncItems] Found 'transcript' key with string content."); content = item["transcript"] elif "captions" in item and isinstance(item["captions"], list): logger.warning("[Apify SyncItems] Received list format for 'captions' unexpectedly. Processing...") if len(item["captions"]) > 0 and isinstance(item["captions"][0], dict) and 'text' in item["captions"][0]: content = " ".join(line.get("text", "") for line in item["captions"] if line.get("text")) elif len(item["captions"]) > 0 and isinstance(item["captions"][0], str): content = " ".join(item["captions"]) if content and isinstance(content, str): logger.info(f"[Apify SyncItems] Success via REST for {video_url}. Length: {len(content)}"); return content.strip() else: logger.warning(f"[Apify SyncItems] Dataset item parsed but transcript content empty/invalid format for {video_url}. Item keys: {list(item.keys())}"); return None else: logger.warning(f"[Apify SyncItems] Actor success but dataset was empty for {video_url}. Response: {results}"); return None except json.JSONDecodeError: logger.error(f"[Apify SyncItems] Failed JSON decode. Status:{response.status_code}. Resp:{response.text[:200]}"); return None except Exception as e: logger.error(f"[Apify SyncItems] Error processing success response for {video_url}: {e}", exc_info=True); return None elif response.status_code == 400: logger.error(f"[Apify SyncItems] Bad Request (400) for {video_url}. Check payload. Resp:{response.text[:200]}"); return None elif response.status_code == 401: logger.error("[Apify SyncItems] Auth error (401). Check token."); return None elif response.status_code == 404: error_info = ""; try: error_info = response.json().get("error", {}).get("message", "") except Exception: pass logger.error(f"[Apify SyncItems] Endpoint/Actor Not Found (404). Error: '{error_info}' Resp:{response.text[:200]}"); return None else: logger.error(f"[Apify SyncItems] Unexpected status {response.status_code} for {video_url}. Resp:{response.text[:200]}"); return None except httpx.TimeoutException as e: logger.error(f"[Apify SyncItems] Timeout during API interaction for {video_url}: {e}"); return None except httpx.HTTPStatusError as e: logger.error(f"[Apify SyncItems] HTTP Status Error during API interaction for {video_url}: {e}"); return None except httpx.RequestError as e: logger.error(f"[Apify SyncItems] Request error during API interaction for {video_url}: {e}"); return None except Exception as e: logger.error(f"[Apify SyncItems] Unexpected error during Apify SyncItems REST call for {video_url}: {e}", exc_info=True); return None async def get_youtube_transcript(video_id: str, video_url: str) -> Optional[str]: global SUPADATA_API_KEY, APIFY_API_TOKEN if not video_id: logger.error("get_youtube_transcript: No video_id"); return None logger.info(f"Fetching transcript for video ID: {video_id} (URL: {video_url})") transcript_text = None logger.info("[Primary YT] Attempting youtube-transcript-api...") try: transcript_list = await asyncio.to_thread( YouTubeTranscriptApi.get_transcript, video_id, languages=['en', 'en-GB', 'en-US'] ) if transcript_list: transcript_text = " ".join([item['text'] for item in transcript_list if 'text' in item]) if transcript_text: logger.info(f"[Primary YT] Success via lib for {video_id} (len: {len(transcript_text)})"); return transcript_text else: logger.warning(f"[Primary YT] Transcript list/text empty for {video_id}"); transcript_text = None except Exception as e: logger.warning(f"[Primary YT] Error via lib for {video_id}: {e}") if "No transcript found" in str(e): logger.warning(f"[Primary YT] No transcript found.") elif "disabled" in str(e): logger.warning(f"[Primary YT] Transcripts disabled.") transcript_text = None if transcript_text is None: logger.info("[Fallback YT 1] Trying Supadata API...") if SUPADATA_API_KEY: transcript_text = await get_transcript_via_supadata(video_id, SUPADATA_API_KEY) if transcript_text: logger.info(f"[Fallback YT 1] Success via Supadata for {video_id}"); return transcript_text else: logger.warning(f"[Fallback YT 1] Supadata failed or no content for {video_id}.") else: logger.warning("[Fallback YT 1] Supadata API key unavailable. Skipping.") if transcript_text is None: logger.info("[Fallback YT 2] Trying Apify REST API (SyncItems)...") if APIFY_API_TOKEN: transcript_text = await get_transcript_via_apify(video_url, APIFY_API_TOKEN) if transcript_text: logger.info(f"[Fallback YT 2] Success via Apify SyncItems REST for {video_url}"); return transcript_text else: logger.warning(f"[Fallback YT 2] Apify SyncItems REST failed or no content for {video_url}.") else: logger.warning("[Fallback YT 2] Apify API token unavailable. Skipping.") if transcript_text is None: logger.error(f"All methods failed for YT transcript: {video_id}"); return None return transcript_text # --- Website Content Fetching --- # NEW: Primary Method using Crawl4AI async def get_website_content_via_crawl4ai(url: str) -> Optional[str]: """Fetches website content using Crawl4AI. Returns Markdown.""" global _crawl4ai_primary_web_enabled if not _crawl4ai_primary_web_enabled: logger.warning("[Crawl4AI Primary] Called but disabled/unavailable.") return None if not url: logger.error("[Crawl4AI Primary] No URL provided"); return None logger.info(f"[Crawl4AI Primary] Attempting fetch for: {url}") try: # Using async with for proper resource cleanup async with AsyncWebCrawler(headless=True) as crawler: # Headless is generally preferred for server environments # Timeout can be added here if needed: crawler_params={"timeout": 60000} # milliseconds result = await crawler.arun(url=url) if result and result.markdown: logger.info(f"[Crawl4AI Primary] Success via Crawl4AI for {url}. Markdown length: {len(result.markdown)}") return result.markdown.strip() elif result and not result.markdown: logger.warning(f"[Crawl4AI Primary] Crawl4AI ran but returned empty markdown for {url}.") return None else: logger.warning(f"[Crawl4AI Primary] Crawl4AI returned no result object for {url}.") return None except ImportError: logger.error("[Crawl4AI Primary] Import Error - library might be missing.") _crawl4ai_primary_web_enabled = False # Disable if import fails at runtime return None except Exception as e: # Catch potential Playwright errors, timeouts, or other issues logger.error(f"[Crawl4AI Primary] Error during Crawl4AI execution for {url}: {e}", exc_info=True) return None # HELPER: Used by Fallback 1 (BS4) - no changes needed here async def fetch_url_content_for_scrape(url: str, timeout: int = 25) -> Optional[str]: headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36', 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8', 'Accept-Language': 'en-US,en;q=0.5', 'Connection': 'keep-alive', 'DNT': '1', 'Upgrade-Insecure-Requests': '1' } try: async with httpx.AsyncClient(follow_redirects=True, timeout=timeout, headers=headers) as client: logger.debug(f"[Web Scrape Helper] Sending request to {url}") response = await client.get(url) logger.debug(f"[Web Scrape Helper] Received response {response.status_code} from {url}") response.raise_for_status() content_type = response.headers.get('content-type', '').lower() if 'html' not in content_type: logger.warning(f"[Web Scrape Helper] Non-HTML content type from {url}: {content_type}"); return None try: return response.text except Exception as e: logger.error(f"[Web Scrape Helper] Error decoding response for {url}: {e}"); return None except httpx.HTTPStatusError as e: logger.error(f"[Web Scrape Helper] HTTP error {e.response.status_code} fetching {url}: {e}") except httpx.TimeoutException: logger.error(f"[Web Scrape Helper] Timeout error fetching {url}") except httpx.TooManyRedirects: logger.error(f"[Web Scrape Helper] Too many redirects fetching {url}") except httpx.RequestError as e: logger.error(f"[Web Scrape Helper] Request error fetching {url}: {e}") except Exception as e: logger.error(f"[Web Scrape Helper] Unexpected error fetching {url}: {e}", exc_info=True) return None # Fallback 1: Direct BS4 Scraping (Renamed original function) async def get_website_content_via_bs4(url: str) -> Optional[str]: """Fetches and extracts text content using BeautifulSoup (Fallback 1).""" global _bs4_fallback_web_enabled if not _bs4_fallback_web_enabled: logger.warning("[BS4 Fallback] Called but disabled.") # Should not happen unless manually disabled return None if not url: logger.error("[BS4 Fallback] No URL provided"); return None logger.info(f"[BS4 Fallback] Fetching website content for: {url}") html_content = await fetch_url_content_for_scrape(url) # Use the helper if not html_content: logger.warning(f"[BS4 Fallback] fetch_url_content_for_scrape failed for {url}") return None try: def parse_html(content): soup = BeautifulSoup(content, DEFAULT_PARSER) # Keep the existing cleaning logic for element in soup(["script", "style", "header", "footer", "nav", "aside", "form", "button", "input", "iframe", "img", "svg", "link", "meta", "noscript", "figure"]): element.extract() main_content = soup.find('main') or soup.find('article') or soup.find(id='content') or soup.find(class_='content') or soup.find(id='main-content') or soup.find(class_='main-content') or soup.find(role='main') target_element = main_content if main_content else soup.body if not target_element: logger.warning(f"[BS4 Fallback] Could not find body/main for parsing {url}"); return None lines = [line.strip() for line in target_element.get_text(separator='\n', strip=True).splitlines() if line.strip()] text = " ".join(lines) if not text: logger.warning(f"[BS4 Fallback] Extracted text empty after clean for {url}"); return None return text text_content = await asyncio.to_thread(parse_html, html_content) if text_content: logger.info(f"[BS4 Fallback] Success scrape for {url} (final len: {len(text_content)})"); return text_content else: logger.warning(f"[BS4 Fallback] parse_html returned None for {url}") return None except Exception as e: logger.error(f"[BS4 Fallback] Error scraping/parsing {url}: {e}", exc_info=True); return None # Fallback 2: urltotext.com API (Unchanged function) async def get_website_content_via_api(url: str, api_key: str) -> Optional[str]: """Fetches website content using urltotext.com API (Fallback 2).""" global _urltotext_fallback_web_enabled if not _urltotext_fallback_web_enabled: logger.warning("[urltotext API Fallback] Called but disabled (no API key).") return None if not url: logger.error("[urltotext API Fallback] No URL"); return None if not api_key: logger.error("[urltotext API Fallback] urltotext.com API key missing."); return None # Redundant check but safe logger.info(f"[urltotext API Fallback] Attempting fetch for: {url} using urltotext.com API") api_endpoint = "https://urltotext.com/api/v1/urltotext/" payload = { "url": url, "output_format": "text", "extract_main_content": True, "render_javascript": True, "residential_proxy": False } headers = { "Authorization": f"Token {api_key}", "Content-Type": "application/json" } try: async with httpx.AsyncClient(timeout=45.0) as client: logger.debug(f"[urltotext API Fallback] Sending request to urltotext.com API for {url}") response = await client.post(api_endpoint, headers=headers, json=payload) logger.debug(f"[urltotext API Fallback] Received status {response.status_code} from urltotext.com API for {url}") if response.status_code == 200: try: data = response.json() content = data.get("data", {}).get("content"); credits = data.get("credits_used", "N/A"); warning = data.get("data", {}).get("warning") if warning: logger.warning(f"[urltotext API Fallback] urltotext.com API Warning for {url}: {warning}") if content: logger.info(f"[urltotext API Fallback] Success via urltotext.com API for {url}. Len: {len(content)}. Credits: {credits}"); return content.strip() else: logger.warning(f"[urltotext API Fallback] urltotext.com API success but content empty for {url}. Resp: {data}"); return None except json.JSONDecodeError: logger.error(f"[urltotext API Fallback] Failed JSON decode urltotext.com for {url}. Resp:{response.text[:500]}"); return None except Exception as e: logger.error(f"[urltotext API Fallback] Error processing urltotext.com success response for {url}: {e}", exc_info=True); return None elif response.status_code in [400, 401, 402, 403, 422, 500]: logger.error(f"[urltotext API Fallback] Error {response.status_code} from urltotext.com API for {url}. Resp:{response.text[:200]}"); return None else: logger.error(f"[urltotext API Fallback] Unexpected status {response.status_code} from urltotext.com API for {url}. Resp:{response.text[:200]}"); return None except httpx.TimeoutException: logger.error(f"[urltotext API Fallback] Timeout connecting to urltotext.com API for {url}"); return None except httpx.RequestError as e: logger.error(f"[urltotext API Fallback] Request error connecting to urltotext.com API for {url}: {e}"); return None except Exception as e: logger.error(f"[urltotext API Fallback] Unexpected error during urltotext.com API call for {url}: {e}", exc_info=True); return None # --- Summarization Functions (Unchanged) --- async def _call_gemini(text: str, summary_type: str) -> Tuple[Optional[str], Optional[str]]: """Internal function to call Gemini API. Returns (summary, error_message).""" global GEMINI_MODEL, _gemini_primary_enabled if not _gemini_primary_enabled: logger.error("[Gemini Primary] Called but is disabled."); return None, "Error: Primary AI service (Gemini) not configured/available." logger.info(f"[Gemini Primary] Generating {summary_type} summary using {GEMINI_MODEL}. Input length: {len(text)}") # Define prompts (Keep existing prompts) if summary_type == "paragraph": prompt = ("You are an AI model designed to provide concise summaries using British English spellings. Your output MUST be:\n" "• Clear and simple language suitable for someone unfamiliar with the topic.\n" "• Uses British English spellings throughout.\n" "• Straightforward and understandable vocabulary; avoid complex terms.\n" "• Presented as ONE SINGLE PARAGRAPH.\n" "• No more than 85 words maximum; but does not have to be exactly 85.\n" "• Considers the entire text content equally.\n" "• Uses semicolons (;) instead of em dashes (– or —).\n" "• **Focus ONLY on the main content; strictly EXCLUDE information about website features, subscriptions, ads, cookie notices, or navigation elements. Do not include things like free/paid tiers; basic/premium memberships. Especially for ACS membership.**\n\n" # Added instruction "Here is the text to summarise:") else: # points summary prompt = ("You are an AI model designed to provide concise summaries using British English spellings. Your output MUST strictly follow this format:\n\n" "• For each distinct topic or section identified in the text, create a heading.\n" "• Each heading MUST be plain text without any formatting (e.g., Section Title).\n" "• Immediately following each heading, list the key points as a bulleted list.\n" "• Each bullet point MUST start with a hyphen and a space (- ) on a new line.\n" "• The text within each bullet point should NOT contain any bold formatting.\n" "• IMPORTANT: Never use bold formatting (double asterisks) within the text of the bullet points themselves.\n" "• Use clear, simple, and straightforward language suitable for someone unfamiliar with the topic.\n" "• Use British English spellings throughout.\n" "• Avoid overly complex or advanced vocabulary.\n" "• Keep bullet points concise.\n" "• Ensure the entire summary takes no more than two minutes to read.\n" "• Consider the entire text's content, not just the beginning or a few topics.\n" "• Use semicolons (;) instead of em dashes (– or —).\n" "• **Focus ONLY on the main content; strictly EXCLUDE information about website features, subscriptions, ads, cookie notices, or navigation elements. Do not include things like free/paid tiers; basic/premium memberships. Especially for ACS membership.**\n\n" # Added instruction "Here is the text to summarise:") # Input Length Check (Gemini-specific limits if known, otherwise use a large default) MAX_INPUT_LENGTH_GEMINI = 1000000 # Default limit for Gemini if len(text) > MAX_INPUT_LENGTH_GEMINI: logger.warning(f"[Gemini Primary] Input length ({len(text)}) exceeds limit ({MAX_INPUT_LENGTH_GEMINI}). Truncating."); text = text[:MAX_INPUT_LENGTH_GEMINI] + "... (Content truncated)" full_prompt = f"{prompt}\n\n{text}" # Safety Settings (Block None) safety_settings = { HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE, HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE, HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE, HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE, getattr(HarmCategory, 'HARM_CATEGORY_CIVIC_INTEGRITY', None): HarmBlockThreshold.BLOCK_NONE } safety_settings = {k: v for k, v in safety_settings.items() if k is not None} logger.debug(f"[Gemini Primary] Using safety settings: {safety_settings}") try: logger.debug(f"[Gemini Primary] Initializing model {GEMINI_MODEL}") model = genai.GenerativeModel(GEMINI_MODEL) logger.info(f"[Gemini Primary] Sending request to Gemini ({GEMINI_MODEL})...") request_options = {"timeout": 120} # Generous timeout for Gemini response = await model.generate_content_async( full_prompt, safety_settings=safety_settings, request_options=request_options ) logger.info("[Gemini Primary] Received response from Gemini.") # Check for immediate blocking reasons if response.prompt_feedback.block_reason: logger.warning(f"[Gemini Primary] Request blocked by API. Reason: {response.prompt_feedback.block_reason}"); return None, f"Sorry, the primary AI model ({GEMINI_MODEL}) blocked the request (Reason: {response.prompt_feedback.block_reason})." # Check candidate-level blocking for cand in response.candidates: if cand.finish_reason == 'SAFETY': logger.warning(f"[Gemini Primary] Candidate blocked due to SAFETY. Ratings: {cand.safety_ratings}") return None, f"Sorry, the primary AI model ({GEMINI_MODEL}) blocked the response due to safety filters." # Try to get text, handle potential ValueError if blocked try: summary = response.text except ValueError as e: logger.warning(f"[Gemini Primary] Error accessing response text (likely blocked content): {e}"); summary = None if summary: logger.info(f"[Gemini Primary] Success generating summary. Output len: {len(summary)}"); return summary.strip(), None else: finish_reason = response.candidates[0].finish_reason if response.candidates else 'N/A' logger.warning(f"[Gemini Primary] Gemini returned empty summary or content was blocked. Finish reason: {finish_reason}"); return None, f"Sorry, the primary AI model ({GEMINI_MODEL}) did not provide a summary (Finish Reason: {finish_reason})." except Exception as e: logger.error(f"[Gemini Primary] Unexpected error during Gemini API call: {e}", exc_info=True); return None, f"Sorry, an unexpected error occurred while using the primary AI service ({GEMINI_MODEL})." async def _call_openrouter(text: str, summary_type: str) -> Tuple[Optional[str], Optional[str]]: """Internal function to call OpenRouter API (Fallback). Returns (summary, error_message).""" global OPENROUTER_API_KEY, OPENROUTER_MODEL, _openrouter_fallback_enabled if not _openrouter_fallback_enabled: logger.error("[OpenRouter Fallback] Called but is disabled."); return None, "Error: Fallback AI service (OpenRouter) not configured/available." logger.info(f"[OpenRouter Fallback] Generating {summary_type} summary using {OPENROUTER_MODEL}. Input length: {len(text)}") # Define prompts (Keep existing prompts) if summary_type == "paragraph": prompt = ("You are an AI model designed to provide concise summaries using British English spellings. Your output MUST be:\n" "• Clear and simple language suitable for someone unfamiliar with the topic.\n" "• Uses British English spellings throughout.\n" "• Straightforward and understandable vocabulary; avoid complex terms.\n" "• Presented as ONE SINGLE PARAGRAPH.\n" "• No more than 85 words maximum; but does not have to be exactly 85.\n" "• Considers the entire text content equally.\n" "• Uses semicolons (;) instead of em dashes (– or —).\n" "• **Focus ONLY on the main content; strictly EXCLUDE information about website features, subscriptions, ads, cookie notices, or navigation elements. Do not include things like free/paid tiers; basic/premium memberships. Especially for ACS membership.**\n\n" # Added instruction "Here is the text to summarise:") else: # points summary prompt = ("You are an AI model designed to provide concise summaries using British English spellings. Your output MUST strictly follow this format:\n\n" "• For each distinct topic or section identified in the text, create a heading.\n" "• Each heading MUST be plain text without any formatting (e.g., Section Title).\n" "• Immediately following each heading, list the key points as a bulleted list.\n" "• Each bullet point MUST start with a hyphen and a space (- ) on a new line.\n" "• The text within each bullet point should NOT contain any bold formatting.\n" "• IMPORTANT: Never use bold formatting (double asterisks) within the text of the bullet points themselves.\n" "• Use clear, simple, and straightforward language suitable for someone unfamiliar with the topic.\n" "• Use British English spellings throughout.\n" "• Avoid overly complex or advanced vocabulary.\n" "• Keep bullet points concise.\n" "• Ensure the entire summary takes no more than two minutes to read.\n" "• Consider the entire text's content, not just the beginning or a few topics.\n" "• Use semicolons (;) instead of em dashes (– or —).\n" "• **Focus ONLY on the main content; strictly EXCLUDE information about website features, subscriptions, ads, cookie notices, or navigation elements. Do not include things like free/paid tiers; basic/premium memberships. Especially for ACS membership.**\n\n" # Added instruction "Here is the text to summarise:") # Input Length Check (OpenRouter-specific limit) MAX_INPUT_LENGTH_OR = 500000 if len(text) > MAX_INPUT_LENGTH_OR: logger.warning(f"[OpenRouter Fallback] Input length ({len(text)}) exceeds limit ({MAX_INPUT_LENGTH_OR}). Truncating."); text = text[:MAX_INPUT_LENGTH_OR] + "... (Content truncated)" full_prompt = f"{prompt}\n\n{text}" headers = { "Authorization": f"Bearer {OPENROUTER_API_KEY}", "Content-Type": "application/json" } payload = { "model": OPENROUTER_MODEL, "messages": [{"role": "user", "content": full_prompt}] } openrouter_api_endpoint = "https://openrouter.ai/api/v1/chat/completions" # Use reasonable timeouts for fallback api_timeouts = httpx.Timeout(25.0, read=20.0, write=25.0, pool=60.0) response = None try: async with httpx.AsyncClient(timeout=api_timeouts) as client: logger.info(f"[OpenRouter Fallback] Sending request to OpenRouter ({OPENROUTER_MODEL}) with read timeout {api_timeouts.read}s...") response = await client.post(openrouter_api_endpoint, headers=headers, json=payload) if response: logger.info(f"[OpenRouter Fallback] Received response from OpenRouter. Status code: {response.status_code}") else: logger.error("[OpenRouter Fallback] No response from OpenRouter (unexpected)."); return None, "Sorry, fallback AI service failed unexpectedly." if response.status_code == 200: try: data = response.json() if data.get("choices") and isinstance(data["choices"], list) and len(data["choices"]) > 0: message = data["choices"][0].get("message") if message and isinstance(message, dict): summary = message.get("content") if summary: logger.info(f"[OpenRouter Fallback] Success via OpenRouter. Output len: {len(summary)}") return summary.strip(), None else: logger.warning(f"[OpenRouter Fallback] OpenRouter success but content empty. Resp: {data}") return None, "Sorry, the fallback AI model returned an empty summary." else: logger.error(f"[OpenRouter Fallback] Unexpected message structure: {message}. Full: {data}") return None, "Sorry, could not parse fallback AI response (format)." else: logger.error(f"[OpenRouter Fallback] Unexpected choices structure: {data.get('choices')}. Full: {data}") return None, "Sorry, could not parse fallback AI response (choices)." except json.JSONDecodeError: logger.error(f"[OpenRouter Fallback] Failed JSON decode OpenRouter. Status:{response.status_code}. Resp:{response.text[:500]}") return None, "Sorry, failed to understand fallback AI response." except Exception as e: logger.error(f"[OpenRouter Fallback] Error processing OpenRouter success response: {e}", exc_info=True) return None, "Sorry, error processing fallback AI response." # Handle specific error codes elif response.status_code == 401: logger.error("[OpenRouter Fallback] API key invalid (401)."); return None, "Error: Fallback AI model configuration key is invalid." elif response.status_code == 402: logger.error("[OpenRouter Fallback] Payment Required (402)."); return None, "Sorry, fallback AI service limits/payment issue." elif response.status_code == 429: logger.warning("[OpenRouter Fallback] Rate Limit Exceeded (429)."); return None, "Sorry, fallback AI model is busy. Try again." elif response.status_code == 500: logger.error(f"[OpenRouter Fallback] Internal Server Error (500). Resp:{response.text[:500]}"); return None, "Sorry, fallback AI service internal error." else: error_info = ""; try: error_info = response.json().get("error", {}).get("message", "") except Exception: pass logger.error(f"[OpenRouter Fallback] Unexpected status {response.status_code}. Error: '{error_info}' Resp:{response.text[:500]}"); return None, f"Sorry, fallback AI service returned unexpected status ({response.status_code})." except httpx.TimeoutException as e: logger.error(f"[OpenRouter Fallback] Timeout error ({type(e)}) connecting/reading from OpenRouter API: {e}") return None, f"Sorry, the fallback AI service ({OPENROUTER_MODEL}) timed out." except httpx.RequestError as e: logger.error(f"[OpenRouter Fallback] Request error connecting to OpenRouter API: {e}") return None, "Sorry, there was an error connecting to the fallback AI model service." except Exception as e: logger.error(f"[OpenRouter Fallback] Unexpected error during OpenRouter call: {e}", exc_info=True) return None, "Sorry, an unexpected error occurred while using the fallback AI service." async def generate_summary(text: str, summary_type: str) -> str: """ Generates summary using Gemini (Primary) and falls back to OpenRouter if needed. Ensures point-based summaries have plain text headings. """ global _gemini_primary_enabled, _openrouter_fallback_enabled, GEMINI_MODEL, OPENROUTER_MODEL logger.info(f"[Summary Generation] Starting process. Primary: Gemini ({GEMINI_MODEL}), Fallback: OpenRouter ({OPENROUTER_MODEL})") final_summary: Optional[str] = None error_message: Optional[str] = None # --- Attempt 1: Gemini (Primary) --- if _gemini_primary_enabled: logger.info(f"[Summary Generation] Attempting primary AI: Gemini ({GEMINI_MODEL})") final_summary, error_message = await _call_gemini(text, summary_type) if final_summary: logger.info(f"[Summary Generation] Success with primary AI (Gemini).") # Return directly - prompt handles formatting return final_summary else: logger.warning(f"[Summary Generation] Primary AI (Gemini) failed or returned unusable result. Error: {error_message}. Proceeding to fallback.") else: logger.warning("[Summary Generation] Primary AI (Gemini) is disabled or unavailable. Proceeding directly to fallback.") error_message = "Primary AI (Gemini) unavailable." # Set initial error message # --- Attempt 2: OpenRouter (Fallback) --- if _openrouter_fallback_enabled: logger.info(f"[Summary Generation] Attempting fallback AI: OpenRouter ({OPENROUTER_MODEL})") fallback_summary, fallback_error = await _call_openrouter(text, summary_type) if fallback_summary: logger.info(f"[Summary Generation] Success with fallback AI (OpenRouter).") # Return directly - prompt handles formatting return fallback_summary else: logger.error(f"[Summary Generation] Fallback AI (OpenRouter) also failed. Error: {fallback_error}") # Combine error messages if possible if error_message: # Keep the primary error if it exists return f"{error_message} Fallback AI ({OPENROUTER_MODEL}) also failed: {fallback_error}" else: # If primary was skipped return f"Fallback AI ({OPENROUTER_MODEL}) failed: {fallback_error}" else: logger.error("[Summary Generation] Fallback AI (OpenRouter) is disabled or unavailable. Cannot proceed.") if error_message: # Return the primary error return f"{error_message} Fallback AI is also unavailable." else: # Should not happen if logic is correct, but safeguard return "Error: Both primary and fallback AI services are unavailable." # Should ideally not be reached if logic above is sound, but as a final catch logger.error("[Summary Generation] Reached end of function without returning a summary or specific error.") return "Sorry, an unknown error occurred during summary generation after trying all available models." # --- Main Task Processing (Modified Web Fetching Logic) --- async def process_summary_task( user_id: int, chat_id: int, message_id_to_edit: Optional[int], url: str, summary_type: str, bot_token: str ) -> None: task_id = f"{user_id}-{message_id_to_edit or 'new'}"; logger.info(f"[Task {task_id}] Starting processing for URL: {url}") background_request: Optional[BaseRequest] = None; bot: Optional[Bot] = None try: background_request = HTTPXRequest( connect_timeout=15.0, read_timeout=60.0, write_timeout=60.0, pool_timeout=60.0 ); bot = Bot(token=bot_token, request=background_request) except Exception as e: logger.critical(f"[Task {task_id}] Failed to create background bot: {e}", exc_info=True); return content: Optional[str] = None user_feedback_message: Optional[str] = None success = False status_message_id = message_id_to_edit message_to_delete_later_id : Optional[int] = None try: # Send initial "Processing..." message (or edit existing) processing_message_text = f"Got it! Generating '{summary_type}' summary for:\n`{url}`\n\nThis might take a moment..." if status_message_id: try: await retry_bot_operation( bot.edit_message_text, chat_id=chat_id, message_id=status_message_id, text=processing_message_text, parse_mode=ParseMode.MARKDOWN, reply_markup=None ); logger.debug(f"[Task {task_id}] Edited message {status_message_id} to 'Processing'") except Exception as e: logger.warning(f"[Task {task_id}] Could not edit original message {status_message_id}: {e}. Sending new."); status_message_id = None if not status_message_id: try: status_message = await retry_bot_operation( bot.send_message, chat_id=chat_id, text=processing_message_text, parse_mode=ParseMode.MARKDOWN ) if status_message: message_to_delete_later_id = status_message.message_id; logger.debug(f"[Task {task_id}] Sent new status message {message_to_delete_later_id}") else: raise RuntimeError("Failed to send status message after retries.") except Exception as e: logger.error(f"[Task {task_id}] CRITICAL: Failed to send new status message: {e}"); raise try: await retry_bot_operation(bot.send_chat_action, chat_id=chat_id, action='typing') is_youtube = is_youtube_url(url) logger.debug(f"[Task {task_id}] URL type: {'YouTube' if is_youtube else 'Website'}") # --- YouTube Processing (Unchanged) --- if is_youtube: video_id = extract_youtube_id(url) if video_id: content = await get_youtube_transcript(video_id, url) # Tries lib -> Supadata -> Apify else: user_feedback_message = "Sorry, I couldn't understand that YouTube URL format." if not content and not user_feedback_message: user_feedback_message = "Sorry, I couldn't get the transcript for that YouTube video using any available method (unavailable/private/no captions?)." # --- Website Processing (NEW Logic) --- else: # Method 1: Crawl4AI (Primary) if _crawl4ai_primary_web_enabled: logger.info(f"[Task {task_id}] Trying primary web method: Crawl4AI") await retry_bot_operation(bot.send_chat_action, chat_id=chat_id, action='typing') content = await get_website_content_via_crawl4ai(url) if content: logger.info(f"[Task {task_id}] Success via Crawl4AI for {url} (len: {len(content)})") else: logger.warning(f"[Task {task_id}] Crawl4AI failed or returned empty for {url}.") else: logger.warning(f"[Task {task_id}] Crawl4AI is disabled. Skipping.") # Method 2: BeautifulSoup (Fallback 1) if not content and _bs4_fallback_web_enabled: logger.warning(f"[Task {task_id}] Trying fallback web method 1: BeautifulSoup") await retry_bot_operation(bot.send_chat_action, chat_id=chat_id, action='typing') content = await get_website_content_via_bs4(url) # Use the renamed function if content: logger.info(f"[Task {task_id}] Success via BS4 scrape for {url} (len: {len(content)})") else: logger.warning(f"[Task {task_id}] BS4 scrape failed or returned empty for {url}.") # Method 3: urltotext.com API (Fallback 2) if not content and _urltotext_fallback_web_enabled: logger.warning(f"[Task {task_id}] Trying fallback web method 2: urltotext.com API") await retry_bot_operation(bot.send_chat_action, chat_id=chat_id, action='typing') content = await get_website_content_via_api(url, URLTOTEXT_API_KEY) # API key checked inside function if content: logger.info(f"[Task {task_id}] Success via urltotext.com API for {url} (len: {len(content)})") else: logger.warning(f"[Task {task_id}] urltotext.com API failed or returned empty for {url}.") # Final check if any website method succeeded if not content: methods_tried = [] if _crawl4ai_primary_web_enabled: methods_tried.append("Crawl4AI") if _bs4_fallback_web_enabled: methods_tried.append("BS4") if _urltotext_fallback_web_enabled: methods_tried.append("API") tried_str = ", ".join(methods_tried) if methods_tried else "configured methods" user_feedback_message = f"Sorry, I couldn't fetch content from that website using any available method ({tried_str}). It might be blocked, inaccessible, or empty." # --- Summarization --- if content: logger.info(f"[Task {task_id}] Content fetched (len:{len(content)}). Generating summary.") await retry_bot_operation(bot.send_chat_action, chat_id=chat_id, action='typing') # NOTE: `content` might be Markdown (from Crawl4AI) or plain text (from others). # The LLM prompts should handle this reasonably well. final_summary = await generate_summary(content, summary_type) # Calls Gemini -> OpenRouter if final_summary.startswith("Error:") or final_summary.startswith("Sorry,"): user_feedback_message = final_summary # Pass AI error message to user logger.warning(f"[Task {task_id}] Summary generation failed: {final_summary}") else: max_length = 4096; summary_parts = [final_summary[i:i+max_length] for i in range(0, len(final_summary), max_length)] # Send summary without explicit Markdown parsing, assuming LLM followed instructions # for plain headings and standard bullet points. Using parse_mode=None. await retry_bot_operation( bot.send_message, chat_id=chat_id, text=summary_parts[0], parse_mode=None, link_preview_options={'is_disabled': True} ) for part in summary_parts[1:]: await asyncio.sleep(0.5) await retry_bot_operation( bot.send_message, chat_id=chat_id, text=part, parse_mode=None, link_preview_options={'is_disabled': True} ) success = True logger.info(f"[Task {task_id}] Successfully sent summary ({len(summary_parts)} parts).") user_feedback_message = None # Clear any previous fetch error if summary succeeded # --- Send Feedback if Fetching Failed --- elif user_feedback_message: logger.warning(f"[Task {task_id}] Sending failure feedback: {user_feedback_message}") await retry_bot_operation( bot.send_message, chat_id=chat_id, text=user_feedback_message, link_preview_options={'is_disabled': True} ) except Exception as e: logger.error(f"[Task {task_id}] Unexpected error during processing: {e}", exc_info=True) user_feedback_message = "Oops! Something went really wrong while processing your request. Please try again later." try: await retry_bot_operation( bot.send_message, chat_id=chat_id, text=user_feedback_message ) except Exception: logger.error(f"[Task {task_id}] Failed to send unexpected error feedback.") except Exception as outer_e: logger.critical(f"[Task {task_id}] Critical outer error: {outer_e}", exc_info=True) try: if bot: await retry_bot_operation( bot.send_message, chat_id=chat_id, text="❌ Critical internal error occurred processing your request." ) except Exception: logger.exception(f"[Task {task_id}] Failed even to send critical error message.") finally: # Delete the "Processing..." or button message delete_target_id = message_to_delete_later_id if message_to_delete_later_id else status_message_id if delete_target_id and bot: try: await retry_bot_operation(bot.delete_message, chat_id=chat_id, message_id=delete_target_id); logger.debug(f"[Task {task_id}] Deleted status/button message {delete_target_id}") except Exception as del_e: logger.warning(f"[Task {task_id}] Failed to delete status/button message {delete_target_id}: {del_e}") # Close the background bot's HTTP client if background_request and hasattr(background_request, '_client') and background_request._client: try: await background_request._client.aclose(); logger.debug(f"[Task {task_id}] Background bot's HTTPX client closed.") except Exception as e: logger.warning(f"[Task {task_id}] Error closing background bot's client: {e}") logger.info(f"[Task {task_id}] Task completed. Success: {success}") # --- Bot Handlers (Unchanged) --- async def start(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None: user = update.effective_user; mention = user.mention_html() if not user or not update.message: return logger.info(f"User {user.id} used /start.") await update.message.reply_html( f"👋 Hello {mention}! I can summarise YouTube links or website URLs.\n\nJust send me a link anytime!" ) async def help_command(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None: user = update.effective_user if not user or not update.message: return logger.info(f"User {user.id} used /help.") help_text = ( "🔍 How to use this bot:\n\n" "1. Send me any YouTube video link or website URL.\n" "2. I'll ask you how you want it summarised (paragraph or points).\n" "3. Click the button for your choice.\n" "4. Wait for the summary!\n\n" "I'll try multiple methods to get content if the first one fails (especially for YouTube transcripts).\n\n" "Commands:\n" "`/start` - Display welcome message\n" "`/help` - Show this help message" ) await update.message.reply_text(help_text, parse_mode=ParseMode.MARKDOWN) async def handle_potential_url(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None: if not update.message or not update.message.text: return url = update.message.text.strip(); user = update.effective_user if not user: return # Basic URL validation if not (url.startswith('http://') or url.startswith('https://')) or '.' not in url[8:]: logger.debug(f"Ignoring non-URL message from {user.id}") # Optional: Reply if it looks like they tried to send something else? # await update.message.reply_text("Please send a valid website URL (starting with http:// or https://) or a YouTube link.") return logger.info(f"User {user.id} sent potential URL: {url}") context.user_data['url_to_summarize'] = url context.user_data['original_message_id'] = update.message.message_id # Store original message ID if needed later keyboard = [[ InlineKeyboardButton("Paragraph Summary", callback_data="paragraph"), InlineKeyboardButton("Points Summary", callback_data="points") ]] reply_markup = InlineKeyboardMarkup(keyboard) await update.message.reply_text( f"Okay, I see this link:\n{url}\n\nHow would you like it summarised?", reply_markup=reply_markup, disable_web_page_preview=True ) async def handle_summary_type_callback(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None: query = update.callback_query if not query or not query.message or not query.from_user: logger.warning("Callback query missing data."); return user = query.from_user; summary_type = query.data; query_id = query.id try: await query.answer(); logger.debug(f"Ack callback {query_id} from {user.id}") except Exception as e: logger.error(f"Error answering callback {query_id}: {e}", exc_info=True) # Log but continue url = context.user_data.get('url_to_summarize') message_id_to_edit = query.message.message_id # The message with the buttons logger.info(f"User {user.id} chose '{summary_type}' for msg {message_id_to_edit}. URL in context: {'Yes' if url else 'No'}") if not url: logger.warning(f"No URL in context for user {user.id} (cb {query_id}). Might be an old button.") try: await query.edit_message_text(text="Sorry, I couldn't find the URL associated with this request. Please send the link again.") except Exception as e: logger.error(f"Failed to edit 'URL not found' message: {e}") # Attempt to send a new message as a fallback try: await context.bot.send_message(chat_id=user.id, text="Sorry, the context for your previous request seems to have expired. Please send the link again.") except Exception as send_e: logger.error(f"Failed even to send new message about lost context: {send_e}") return # Clear context *after* checking it exists context.user_data.pop('url_to_summarize', None) context.user_data.pop('original_message_id', None) # Clear original ID too logger.debug(f"Cleared URL context for user {user.id}") # Check critical configurations before scheduling task global TELEGRAM_TOKEN, _gemini_primary_enabled, _openrouter_fallback_enabled if not TELEGRAM_TOKEN: logger.critical("TELEGRAM_TOKEN missing! Cannot schedule task.") try: await query.edit_message_text(text="❌ Bot configuration error (Token). Task cannot be started.") except Exception: pass # Ignore if edit fails return if not _gemini_primary_enabled and not _openrouter_fallback_enabled: logger.critical("Neither Gemini nor OpenRouter API keys are configured/valid! Cannot schedule task.") try: await query.edit_message_text(text="❌ AI configuration error: No summarization models available. Task cannot be started.") except Exception: pass # Ignore if edit fails return elif not _gemini_primary_enabled: logger.warning("Primary AI (Gemini) is unavailable, will rely on fallback for summarization.") # User will be notified by generate_summary if fallback also fails elif not _openrouter_fallback_enabled: logger.warning("Fallback AI (OpenRouter) is unavailable for summarization.") # User will be notified by generate_summary if primary fails # Schedule the background task logger.info(f"Scheduling task for user {user.id}, chat {query.message.chat_id}, msg {message_id_to_edit}") asyncio.create_task( process_summary_task( user_id=user.id, chat_id=query.message.chat_id, message_id_to_edit=message_id_to_edit, # Pass the button message ID to edit/delete url=url, summary_type=summary_type, bot_token=TELEGRAM_TOKEN ), name=f"SummaryTask-{user.id}-{message_id_to_edit}" ) # Don't edit the message here; the task will handle it immediately. async def error_handler(update: object, context: ContextTypes.DEFAULT_TYPE) -> None: # Ignore specific, known errors if necessary ignore_errors = (AttributeError, ) # Example if isinstance(context.error, ignore_errors) and "object has no attribute" in str(context.error): logger.debug(f"Ignoring known/handled error in error_handler: {context.error}") return logger.error("Exception while handling an update:", exc_info=context.error) # Optionally, try to inform the user if it's a direct message context # if isinstance(update, Update) and update.effective_chat: # try: await context.bot.send_message(chat_id=update.effective_chat.id, text="An internal error occurred.") # except Exception: logger.error("Failed to send error message to user.") # --- Application Setup (Unchanged) --- async def setup_bot_config() -> Application: logger.info("Configuring Telegram Application..."); global TELEGRAM_TOKEN if not TELEGRAM_TOKEN: raise ValueError("TELEGRAM_TOKEN missing.") # Configure HTTPX client for PTB custom_request = HTTPXRequest( connect_timeout=10.0, read_timeout=30.0, write_timeout=30.0, pool_timeout=60.0 ) application = ( Application.builder() .token(TELEGRAM_TOKEN) .request(custom_request) .build() ) # Add handlers application.add_handler(CommandHandler("start", start)) application.add_handler(CommandHandler("help", help_command)) # Message handler for potential URLs (non-command text) application.add_handler(MessageHandler(filters.TEXT & ~filters.COMMAND, handle_potential_url)) # Callback handler for summary type buttons application.add_handler(CallbackQueryHandler(handle_summary_type_callback)) # Error handler application.add_error_handler(error_handler) logger.info("Telegram application handlers configured."); return application # --- ASGI Lifespan & Webhook (Unchanged) --- @contextlib.asynccontextmanager async def lifespan(app: Starlette): global ptb_app, WEBHOOK_SECRET, TELEGRAM_TOKEN logger.info("ASGI Lifespan: Startup initiated..."); if not TELEGRAM_TOKEN: logger.critical("TG TOKEN missing."); raise RuntimeError("Telegram token missing.") try: ptb_app = await setup_bot_config(); await ptb_app.initialize(); bot_info = await ptb_app.bot.get_me(); logger.info(f"Bot initialized: @{bot_info.username} (ID: {bot_info.id})") # Webhook setup logic current_webhook_info = await ptb_app.bot.get_webhook_info() if current_webhook_info and current_webhook_info.url: logger.info(f"Found existing webhook: {current_webhook_info.url}. Deleting...") try: if await ptb_app.bot.delete_webhook(drop_pending_updates=True): logger.info("Webhook deleted.") else: logger.warning("Failed delete webhook (API returned False).") except Exception as e: logger.warning(f"Could not delete webhook: {e}"); await asyncio.sleep(1) # Brief pause before setting new one space_host = os.environ.get("SPACE_HOST") # Provided by Hugging Face Spaces webhook_path = "/webhook" # Matches the Starlette route below full_webhook_url = None if space_host: protocol = "https" host = space_host.split('://')[-1] # Remove potential protocol prefix from env var full_webhook_url = f"{protocol}://{host.rstrip('/')}{webhook_path}" if full_webhook_url: logger.info(f"Setting webhook: {full_webhook_url}") set_webhook_args: Dict[str, Any] = { "url": full_webhook_url, "allowed_updates": Update.ALL_TYPES, "drop_pending_updates": True } if WEBHOOK_SECRET: set_webhook_args["secret_token"] = WEBHOOK_SECRET logger.info("Using webhook secret token.") await asyncio.sleep(1.0) # Short delay before setting webhook try: await ptb_app.bot.set_webhook(**set_webhook_args) webhook_info = await ptb_app.bot.get_webhook_info() # Verify if webhook_info.url == full_webhook_url: logger.info(f"Webhook set successfully: URL='{webhook_info.url}', Secret Configured={bool(WEBHOOK_SECRET)}") else: logger.error(f"Webhook URL mismatch after setting! Expected '{full_webhook_url}', Got '{webhook_info.url}'. Check SPACE_HOST env var.") raise RuntimeError("Webhook URL mismatch after setting.") await ptb_app.start() # Start listening for updates via webhook logger.info("PTB Application started (webhook mode).") except Exception as e: logger.critical(f"FATAL: Failed to set webhook: {e}", exc_info=True) raise RuntimeError(f"Failed to set webhook: {e}") from e else: logger.critical("Could not construct webhook URL. SPACE_HOST environment variable might be missing or invalid.") raise RuntimeError("Webhook URL could not be determined.") logger.info("ASGI Lifespan: Startup complete."); yield # Application runs here except Exception as startup_err: logger.critical(f"Application startup failed: {startup_err}", exc_info=True) # Ensure cleanup even if startup fails midway if ptb_app: if ptb_app.running: await ptb_app.stop() await ptb_app.shutdown() raise # Reraise the exception to stop the ASGI server finally: # Shutdown phase logger.info("ASGI Lifespan: Shutdown initiated...") if ptb_app: if ptb_app.running: logger.info("Stopping PTB application...") await ptb_app.stop() logger.info("Shutting down PTB application...") await ptb_app.shutdown() logger.info("PTB Application shut down.") else: logger.info("PTB application was not initialized or failed during startup.") logger.info("ASGI Lifespan: Shutdown complete.") async def health_check(request: Request) -> PlainTextResponse: global OPENROUTER_MODEL, GEMINI_MODEL, APIFY_ACTOR_ID, _apify_token_exists, _gemini_primary_enabled, _openrouter_fallback_enabled, _crawl4ai_primary_web_enabled, _bs4_fallback_web_enabled, _urltotext_fallback_web_enabled bot_status = "Not Initialized" if ptb_app and ptb_app.bot: try: if ptb_app.running: bot_info = await ptb_app.bot.get_me() bot_status = f"Running (@{bot_info.username})" else: bot_status = "Initialized/Not running" except Exception as e: bot_status = f"Error checking status: {e}" web_status = f"Web Primary: {'Crawl4AI' if _crawl4ai_primary_web_enabled else 'DISABLED'}\n" \ f"Web Fallback 1: {'BS4' if _bs4_fallback_web_enabled else 'DISABLED'}\n" \ f"Web Fallback 2: {'API' if _urltotext_fallback_web_enabled else 'DISABLED'}" summary_status = f"Summarizer Primary: {GEMINI_MODEL if _gemini_primary_enabled else 'DISABLED'}\n" \ f"Summarizer Fallback: {OPENROUTER_MODEL if _openrouter_fallback_enabled else 'DISABLED'}" yt_status = f"YT Fallback 1: {'Supadata' if SUPADATA_API_KEY else 'DISABLED'}\n" \ f"YT Fallback 2: {APIFY_ACTOR_ID if _apify_token_exists else 'DISABLED'}" return PlainTextResponse( f"TG Bot Summariser - Status: {bot_status}\n\n" f"{web_status}\n\n" f"{summary_status}\n\n" f"{yt_status}" ) async def telegram_webhook(request: Request) -> Response: global WEBHOOK_SECRET if not ptb_app: logger.error("Webhook received but PTB application is not initialized.") return PlainTextResponse('Bot not initialized', status_code=503) # Service Unavailable if not ptb_app.running: logger.warning("Webhook received but PTB application is not running.") return PlainTextResponse('Bot not running', status_code=503) # Service Unavailable try: # Validate secret token if configured if WEBHOOK_SECRET: token_header = request.headers.get("X-Telegram-Bot-Api-Secret-Token") if token_header != WEBHOOK_SECRET: logger.warning(f"Webhook received with invalid secret token. Header: '{token_header}'") return Response(content="Invalid secret token", status_code=403) # Forbidden # Process the update update_data = await request.json() update = Update.de_json(data=update_data, bot=ptb_app.bot) logger.debug(f"Processing update_id: {update.update_id} via webhook") await ptb_app.process_update(update) return Response(status_code=200) # OK - Tell Telegram we received it except json.JSONDecodeError: logger.error("Webhook received invalid JSON payload.") return PlainTextResponse('Bad Request: Invalid JSON', status_code=400) except Exception as e: logger.error(f"Error processing webhook update: {e}", exc_info=True) # Still return OK to Telegram to prevent retries for potentially poison-pill updates, # but log the error for debugging. return Response(status_code=200) # --- ASGI App Definition (Unchanged) --- app = Starlette( debug=False, # Keep debug False in production lifespan=lifespan, routes=[ Route("/", endpoint=health_check, methods=["GET"]), Route("/webhook", endpoint=telegram_webhook, methods=["POST"]), ] ) logger.info("Starlette ASGI application created with health check and webhook routes.") # --- Direct Run (for local testing, unchanged) --- if __name__ == '__main__': import uvicorn logger.warning("Running in development mode using Uvicorn directly (not for production)") log_level = os.environ.get("LOGGING_LEVEL", "info").lower() # Use PORT from env var (like HF Spaces provides) or default to 8080 for local dev local_port = int(os.environ.get('PORT', 8080)) # Disable webhook setup for local Uvicorn run if needed (manual polling instead) # You might need to comment out the webhook setting logic in lifespan for local testing # or run with ngrok/similar and set SPACE_HOST manually. # For simplicity, assuming webhook setup will just log errors if SPACE_HOST isn't set locally. uvicorn.run("__main__:app", host='0.0.0.0', port=local_port, log_level=log_level, reload=True)