Tesla on Friday recalled roughly 1.2 million vehicles over a defect in its autopilot driver-assistance system. The recall covers most Model 3 and Model Y units built since 2023.
", "language": "en", "author": { "id": 4521, "name": "Jane Doe" }, "image": "https://example.com/images/tesla-recall.jpg", "categories": [ { "id": 211, "name": "automotive", "score": 0.91, "taxonomy": "iptc_mediatopics", "links": { "self": "https://api.apitube.io/v1/news/category/iptc_mediatopics/medtop:20000266" } } ], "topics": [ { "id": "automotive.recalls", "name": "Vehicle Recalls", "score": 0.34, "links": { "self": "https://api.apitube.io/v1/news/topic/automotive.recalls" } } ], "industries": [ { "id": 88, "name": "Automotive", "links": { "self": "https://api.apitube.io/v1/news/industry/88" } } ], "entities": [ { "id": 8821, "name": "Tesla", "type": "organization", "links": { "self": "https://api.apitube.io/v1/news/entity/8821", "wikipedia": "https://en.wikipedia.org/wiki/Tesla,_Inc.", "wikidata": "https://www.wikidata.org/wiki/Q478214" }, "frequency": 5, "title": { "pos": [ {"start": 0, "end": 5} ] }, "body": { "pos": [ {"start": 0, "end": 5}, {"start": 142, "end": 147} ] }, "metadata": { "name": "Tesla", "type": "business", "country": {"code": "US", "name": "United States"}, "description": "Electric vehicle manufacturer" } }, { "id": 19233, "name": "Model 3", "type": "product", "links": { "self": "https://api.apitube.io/v1/news/entity/19233" }, "frequency": 2, "title": { "pos": [] }, "body": { "pos": [ {"start": 58, "end": 65} ] }, "metadata": {} } ], "source": { "id": 311, "domain": "reuters.com", "home_page_url": "https://www.reuters.com", "type": "news", "bias": "center", "rankings": { "opr": 9 }, "location": { "country_name": "United States", "country_code": "us" }, "favicon": "https://www.reuters.com/favicon.ico" }, "sentiment": { "overall": {"score": -0.62, "polarity": "negative"}, "title": {"score": -0.55, "polarity": "negative"}, "body": {"score": -0.64, "polarity": "negative"} }, "summary": [ { "sentence": "Tesla recalled roughly 1.2 million vehicles over an autopilot defect.", "sentiment": {"score": -0.6, "polarity": "negative"} } ], "keywords": ["Tesla", "recall", "autopilot", "Model 3"], "links": [ {"url": "https://example.com/tesla-recall-followup", "type": "link"} ], "media": [ {"url": "https://example.com/images/tesla-recall.jpg", "type": "image"} ], "story": { "id": 3019419820, "uri": "https://api.apitube.io/v1/news/story/3019419820" }, "shares": { "total": 412, "facebook": 220, "twitter": 150, "reddit": 42 }, "is_duplicate": false, "is_free": true, "is_breaking": true, "read_time": 2, "sentences_count": 8, "paragraphs_count": 4, "words_count": 320, "characters_count": 1820 } ] } ``` The bot only touches a handful of these — `title`, `description`, `href`, `source.domain`, `sentiment.overall.polarity`, and `entities[].name` — but everything else (topics, industries, source bias, per-sentence summary, share counts) is there if you want richer posts later. In Python, async `httpx`: ```python import httpx, os API_URL = "https://api.apitube.io/v1/news/everything" async def fetch_news(query: str) -> list[dict]: params = { "title": query, "language.code": "en", "per_page": 20, "sort.by": "published_at", "sort.order": "desc", } headers = {"X-API-Key": os.environ["APITUBE_KEY"]} async with httpx.AsyncClient(timeout=15) as client: r = await client.get(API_URL, params=params, headers=headers) r.raise_for_status() return r.json().get("results", []) ``` ### Step 4 — Filter by sentiment and entities This is the step that turns a firehose into a curated channel. Drop articles that don't match the configured polarity, and keep only ones whose entities include at least one watched name. None of the popular tutorials do this — they post everything they fetch. ```python WATCHED_ENTITIES = {"Tesla", "SpaceX", "OpenAI"} WANTED_POLARITY = {"negative"} # use {"positive", "negative", "neutral"} to allow all def passes_filter(article: dict) -> bool: polarity = article.get("sentiment", {}).get("overall", {}).get("polarity") if polarity not in WANTED_POLARITY: return False names = {e.get("name") for e in article.get("entities", [])} return bool(names & WATCHED_ENTITIES) ``` A 30-second filter, but it's the difference between a bot that posts 200 articles a day and one that posts 6 useful ones. ### Step 5 — Schedule channel posts with APScheduler `AsyncIOScheduler` runs inside the same event loop aiogram is using, so the whole bot is one process. Dedup posted IDs in an in-memory `set`; for persistence across restarts, swap in Redis or SQLite. ```python from apscheduler.schedulers.asyncio import AsyncIOScheduler from aiogram.exceptions import TelegramRetryAfter import asyncio, html posted_ids: set[int] = set() async def post_news(bot, channel_id: int): articles = await fetch_news("Tesla OR SpaceX OR OpenAI") for art in articles: if art["id"] in posted_ids or not passes_filter(art): continue text = ( f"{html.escape(art['title'])}\n" f"{html.escape(art['description'])}\n\n" f"{art['source']['domain']} · " f"sentiment: {art['sentiment']['overall']['polarity']}\n" f"Read more" ) try: await bot.send_message(channel_id, text, disable_web_page_preview=False) posted_ids.add(art["id"]) except TelegramRetryAfter as e: await asyncio.sleep(e.retry_after) ``` The `TelegramRetryAfter` catch is what keeps the bot alive when Telegram throttles you for posting too fast — it sleeps the exact requested interval, then the next scheduler tick retries. ### Step 6 — Commands and main loop Three small commands so subscribers can interact with the bot directly: - `/start` — welcome message - `/subscribe` — explain that the bot auto-posts to the channel - `/filter` — show current filter config The dispatcher wires them up, and `start_polling` runs alongside the scheduler. ## Complete code (`bot.py`) ```python import asyncio, html, logging, os from aiogram import Bot, Dispatcher from aiogram.client.default import DefaultBotProperties from aiogram.enums import ParseMode from aiogram.exceptions import TelegramRetryAfter from aiogram.filters import Command, CommandStart from aiogram.types import Message from apscheduler.schedulers.asyncio import AsyncIOScheduler from apscheduler.triggers.interval import IntervalTrigger from dotenv import load_dotenv import httpx load_dotenv() TOKEN = os.environ["BOT_TOKEN"] CHANNEL_ID = int(os.environ["CHANNEL_ID"]) APITUBE_KEY = os.environ["APITUBE_KEY"] API_URL = "https://api.apitube.io/v1/news/everything" WATCHED_ENTITIES = {"Tesla", "SpaceX", "OpenAI"} WANTED_POLARITY = {"negative"} posted_ids: set[int] = set() dp = Dispatcher() async def fetch_news(query: str) -> list[dict]: params = { "title": query, "language.code": "en", "per_page": 20, "sort.by": "published_at", "sort.order": "desc", } headers = {"X-API-Key": APITUBE_KEY} async with httpx.AsyncClient(timeout=15) as client: r = await client.get(API_URL, params=params, headers=headers) r.raise_for_status() return r.json().get("results", []) def passes_filter(article: dict) -> bool: polarity = article.get("sentiment", {}).get("overall", {}).get("polarity") if polarity not in WANTED_POLARITY: return False names = {e.get("name") for e in article.get("entities", [])} return bool(names & WATCHED_ENTITIES) async def post_news(bot: Bot): articles = await fetch_news("Tesla OR SpaceX OR OpenAI") for art in articles: if art["id"] in posted_ids or not passes_filter(art): continue text = ( f"{html.escape(art['title'])}\n" f"{html.escape(art.get('description') or '')}\n\n" f"{art['source']['domain']} · " f"sentiment: {art['sentiment']['overall']['polarity']}\n" f"Read more" ) try: await bot.send_message(CHANNEL_ID, text) posted_ids.add(art["id"]) await asyncio.sleep(1) except TelegramRetryAfter as e: await asyncio.sleep(e.retry_after) except Exception: logging.exception("send_message failed for %s", art["id"]) @dp.message(CommandStart()) async def cmd_start(m: Message): await m.answer("News bot online. Curated negative-sentiment news on Tesla, SpaceX, OpenAI.") @dp.message(Command("subscribe")) async def cmd_subscribe(m: Message): await m.answer("Posts go to the channel automatically every 15 minutes.") @dp.message(Command("filter")) async def cmd_filter(m: Message): await m.answer( f"Polarity: {sorted(WANTED_POLARITY)}\nEntities: {sorted(WATCHED_ENTITIES)}" ) async def main(): logging.basicConfig(level=logging.INFO) bot = Bot(TOKEN, default=DefaultBotProperties(parse_mode=ParseMode.HTML)) scheduler = AsyncIOScheduler() scheduler.add_job(post_news, IntervalTrigger(minutes=15), args=[bot]) scheduler.start() await dp.start_polling(bot) if __name__ == "__main__": asyncio.run(main()) ``` Run it: ```bash python bot.py ``` The first scheduler tick fires 15 minutes after start. To trigger immediately, add `next_run_time=datetime.utcnow()` to `add_job`. ## Deploy with Docker A multi-stage Python 3.12 slim image keeps the final container small and the build deterministic. `Dockerfile`: ```dockerfile FROM python:3.12-slim AS builder WORKDIR /app COPY requirements.txt . RUN pip install --user --no-cache-dir -r requirements.txt FROM python:3.12-slim WORKDIR /app COPY --from=builder /root/.local /root/.local ENV PATH=/root/.local/bin:$PATH COPY bot.py . CMD ["python", "-u", "bot.py"] ``` `requirements.txt`: ``` aiogram==3.27.0 apscheduler==3.10.4 httpx==0.27.0 python-dotenv==1.0.1 ``` `docker-compose.yml`: ```yaml services: bot: build: . env_file: .env restart: unless-stopped ``` For free or near-free hosting in 2026: Fly.io's hobby tier handles a single bot container fine; Railway gives you logs and one-click redeploys; or grab a $5/month VPS (Hetzner, Vultr, DigitalOcean) and run `docker compose up -d`. Whatever you pick, `restart: unless-stopped` ensures a crash doesn't take the bot offline overnight. ## FAQ ### How do I make a Telegram news bot in Python? Register a bot with `@BotFather`, install `aiogram` 3.27 and a news client like `httpx`, fetch articles from a news API such as APITube, filter the response, and post each item to a Telegram chat or channel using `bot.send_message`. Wrap the fetch-and-post step in `APScheduler` to make it run on an interval. ### Which library is better, aiogram or python-telegram-bot? Unlike python-telegram-bot, aiogram is async-only since v3, which means it pairs cleanly with `AsyncIOScheduler` for channel auto-posting in a single event loop. For news/channel bots above ~50 concurrent users, pick aiogram 3.27. For command-driven personal bots, sync-codebase teams, or gradual migration paths, python-telegram-bot 21.x remains a solid choice. Both are actively maintained in 2026. ### Can a Telegram bot post to a channel automatically? Yes. Add the bot to the channel as an administrator with the "Post messages" permission, grab the channel's numeric chat ID, and call `bot.send_message(chat_id, text)` from a scheduled task. `APScheduler`'s `AsyncIOScheduler` runs inside the bot's event loop, so one process handles both polling and scheduled posting. ### Is python-telegram-bot still maintained in 2026? Yes — python-telegram-bot is still actively maintained. Version 21.x ships an async API and ongoing fixes; the project sees regular releases on GitHub. The "is it maintained" question gets asked because the v13 sync API is deprecated, but v20 onwards is async and current. ## Try it Clone the snippets above, drop them in a folder, fill `.env`, `docker compose up -d`. APITube's free tier returns sentiment and entities on every article — get a key at apitube.io and replace the placeholder in step 3. If you want to filter on category instead of entities, read `article["categories"]` and match on each category's `name` (e.g. keep only articles whose category name contains `"technology"`).