MITHLAK

Bilingual industry news, every morning

Content Pipeline

Internal project · Self-hosted n8n

The problem

Dozens of sources in several languages. Most of it repeats. The useful part needs a human eye.

How it works

  1. 01

    Sources collected on a schedule

    System state

  2. 02

    Duplicates removed — by link and by meaning

    System state

  3. 03

    AI scores and summarises in EN + AR

    System state

  4. 04

    Preview sent for approval

    Human decision

  5. 05

    Approved posts published

    System state

Copper marks the human decision. The rest is tracked by the system.

Under the hood

  • n8n workflows deployed via API
  • Qdrant vector store · intel_hub
  • Text QC before any image work
  • Telegram approval webhook
  • 07:30 daily digest

The short version

AI does the reading. A person makes the call.

Start a projectBack to Work