AI Morning Briefing — Automated News Digest
An n8n automation that collects news from multiple web sources and YouTube channels, processes articles through a two-pass LLM pipeline for summarization and relevance scoring, sends candidates to Telegram for manual approval, then compiles and delivers a formatted HTML email briefing.
Key features
- Multi-source scraping: news websites, YouTube channels, and weather APIs, all running on a scheduled trigger
- Two-pass LLM pipeline: first pass extracts and summarizes, second pass scores relevance and removes noise
- Deduplication engine that checks new articles against previously sent briefings
- Telegram approval flow: each candidate story gets sent for quick approve/reject before inclusion
- Webhook-based approval handling that captures responses and updates session state in real time
- Styled HTML email generation with merged approved stories, formatted and ready to read
- Briefing history storage so the system learns what's already been covered
- Fault-tolerant design: skips unavailable sources without breaking the pipeline
Technologies
About the project
Every morning, the client receives a curated briefing in their inbox without ever needing to open a news site. This n8n workflow scrapes a dozen sources overnight—including industry news, YouTube channels, and weather APIs—and runs each article through an LLM to extract what actually matters. It deduplicates content against previous briefings and sends a shortlist to my Telegram for a quick 'approve' or 'reject' before anything goes out. Once I tap approve, the workflow merges the stories, generates a styled HTML email, and delivers it. The whole pipeline is autonomous: if a source is down, it's skipped; if an article is a duplicate, it's dropped. This ensures nothing lands in the client's inbox that I haven't vetted first. I built this because the client was wasting three hours every morning skimming the same sites; now, it takes just two minutes on Telegram.