Monitor automatically
RSS feeds, Google News and other sources are queried regularly and robustly – with timeouts, retries and error handling.
Technical book · Python · July 2026
The information comes to you.
Watch & Alert shows you step by step how to build your own news monitor: query RSS feeds, filter relevant reports, optionally assess them with an LLM, and receive immediate notifications on Telegram or by email – using real, tested Python code.
No more searching yourself. Define what interests you once – relevant news comes to you automatically.
Order
Offer — save €2
E-book available right now
Buy the PDF directly and download it instantly — with a discount code just €7.99 instead of €9.99
Buy the e-bookThe paperback is coming soon. Add yourself to the waiting list – you will be notified as soon as order links and the ISBN are confirmed. Registration is free and non-binding.
You will receive at most one brief message when the book is published. No advertising, no tracking. Your data is used only for notification and never shared with third parties.
What it is about
You are waiting for important news – a tender, an IPO, a product recall, or a report on your subject. With Watch & Alert, you can stop searching every day.
Filter for relevance
Context-sensitive keyword filters, duplicate detection and SQLite as memory. No duplicate alerts, no noise.
Assess with an LLM
An on-device or cloud LLM can classify each report and decide whether it genuinely matches your subject.
Notify immediately
Matches arrive as formatted messages on your Telegram bot, by email or via ntfy.sh – as configured.
What the book covers
15 chapters that build logically on one another – from your first Python script to a production-ready alert system.
1 · Install and configure Python
Installation, venv, pip and your first script.
2 · Linux fundamentals & Bash
Terminal, navigation, processes, permission model, Cron and a Git excursion.
3 · Architecture and modules
The bot’s five classes, modules vs scripts, and your first RSS script.
4 · Configuration and secrets
.env, YAML, TOML – and where each one lives. No tokens in source code.
5 · Security and hygiene
HTTPS, Token rotieren, Logging ohne Geheimnisse zu leaken.
6 · The first watcher
From script to reusable module – clean, testable, maintainable.
7 · Query RSS feeds robustly
Timeouts, retries, backoff and error handling in production.
8 · Databases with SQLite
Schema, migrations, duplicate detection and a clean lifecycle.
9 · Using LLMs effectively
Classification, prompt design, cost, latency and fallbacks.
10 · Filter-Pipeline
Sources → pre-filter → LLM → delivery, cleanly orchestrated.
11 · Notification channels
Implement robust delivery via Telegram bot, ntfy.sh and email.
12 · Automating with Cron, systemd and anacron
Run the watcher reliably without having to start it manually.
13 · Operations, heartbeats, logs and maintenance
Detect outages, analyse logs, rotate tokens and perform audits.
14 · Testing watch systems
pytest, mocks, temporary test databases and continuous integration.
15 · From prototype to production-ready operation
Docker, separate environments, secrets and operational maturity levels.
Appendix
Further guidance and supplementary material.
What you will learn
Six concrete results you will have in hand after reading.
- Complete watch and alert system with Python, RSS and SQLite
- LLM-assisted classification with OpenAI and Ollama
- Notifications via Telegram, ntfy and email
- Automation with Cron, systemd timers and anacron
- Security: secrets, token rotation and auditing
- Testing with pytest and a maturity model
Excerpt
Three pages from the book: the complete foreword and the beginning of Chapter 1 (Install and configure Python, including a WSL2 note for Windows users). This gives you a genuine feel for the tone, depth and structure before you order.
3 pages · A4 · PDF
- Preface — what the book is, who it is for and how it is structured
- The case study behind the EngineAI IPO monitor
- Chapter 1: Installing and Setting Up Python — introduction and learning objective, WSL2 excursion, 1.1 Aside: Windows users — setting up WSL2, 1.2 Why Python?
Open directly in the browser: assets/leseprobe-en.pdf
A real reference project
Not a toy project. The book’s end-to-end example is a productive watcher that monitors an IPO, for instance, collects and filters reports, and, if desired, delivers them via Telegram alert.
- Complete source code with modules, tests and configuration
- Clean architecture in five clearly separated classes
- Real test runs – including a genuine alert on a genuine Telegram bot
- Reproducible: scripts, database and caches are documented in the book
Example output
10 articles found
7 relevant reports sent
0 duplicates on the second run
1 Telegram alert deliveredDescribed in the book. Tested with another subject. It works.
From the blog
Short, practical articles around the book’s topic — to read, try out and riff on.
Python RSS bot RSS filtering LLM classification
Python RSS Bot: Filtering noise with SQLite & LLM (without cost explosion)
How to filter noise from RSS feeds with a Python bot, SQLite deduplication, and LLM classification - without exploding API costs.
Read blog post
ntfy.sh telegram bot alternative push notifications
ntfy.sh over Telegram: Push without a bot
Telegram bots hit 30 msg/s limits and BotFather overhead. ntfy.sh needs neither — a migration from a 12-source watch setup, with code.
Read blog postcron systemd timer monitoring
Cron vs. systemd Timers: Which One to Use in Your
Cron vs. systemd timers in your own monitoring setup: we compare both worlds, show a concrete migration, and call out the trade-offs. A conc
Read blog post
Secret Rotation Watch & Alert SQLite
Zero-Downtime Secret Rotation in Our Watch & Alert Setup
Zero-downtime secret rotation in our Watch & Alert setup: how we safely rotate API keys using SQLite, Ollama, and cron jobs — no external tools.
Read blog postpytest data pipeline SQLite testing
pytest strategies for data pipelines in our Watch-&-Alert
pytest strategies for data pipelines in our Watch-&-Alert project: SQLite in-memory, LLM stubs, property tests, and strict markers explained in detail.
Read blog postWatch & Alert Self-Hosted LLM Agent Ollama
Watch & Alert: Local AI Agent Replaces Cloud API — Open
Watch & Alert with a self-hosted LLM agent: open source, MIT-licensed, no API keys. Ollama setup, fallback strategy, and heartbeat pattern explained.
Read blog postRSS parsing without library Python XML Parser Feed Monitoring
Parsing RSS feeds robustly without a library: How our
Parse RSS feeds robustly without a library: tame the XML storm in your own monitoring system, survive broken feeds, clear Python patterns. A
Read blog postTelegram Bot Rate Limit Watch and Alert Python Queue
Working around Telegram bot rate limits: When your
How to reliably handle Telegram bot rate limits in your home monitoring setup: queue, retry, jitter and coalescing for Watch & Alert, with code.
Read blog postLiftTrace Self-hosted Weightlifting Tracker
LiftTrace as a Watch & Alert source: Self-hosted strength
LiftTrace is a self-hosted weightlifting tracker under AGPL. A concrete use case with Python code and real trade-offs.......................
Read blog postThe blog collects excerpts, practical tips and updates around Watch & Alert. Check back regularly.
Open Source: the complete code
All scripts, modules and configurations from the book are available publicly on GitHub. You can clone them, try them out and adapt them to your own subjects.
watch-alert-book / engineai-monitoring Python news & market monitor: RSS → LLM → Telegram. Reference project for the book „Watch & Alert“.
Clone the repository, create your own .env with your Telegram token, and get started.
You will find step-by-step instructions in Chapters 3 and 6.
About the author
Stephan Paul is very interested in developments in the world of IT.
This book grew from the desire not to miss the IPO of a Chinese robotics company, and from the realisation that the resulting tool is useful for many other monitoring problems.