Inspiration

Online community growth and engagement kinetics structurally outpace human moderation capacity. Traditional moderation infrastructures, such as AutoModerator, operate on a strictly reactive model—intervening only after manual reports accumulate and the moderation queue overflows. SentinelAI was born from the desire to invert this paradigm: shifting community management from a passive, reactive workflow into a structured, proactive decision-making system driven by artificial intelligence.

What it does

SentinelAI is an AI-driven predictive moderation system natively integrated into Reddit. It continuously monitors semantic discussion dynamics and weak behavioral signals to:

  • Evaluate subreddit stability in real time through a dynamic Community Health Score.
  • Detect the contextual emergence of toxicity and escalating emotional patterns before they reach a critical threshold.
  • Monitor operational moderation pressure and estimate the workload of admin teams.
  • Provide moderators with instant diagnostics and graduated action guidelines (interactive playbooks) directly in-app using the native Devvit action "Scan with SentinelAI".

How we built it

The project is built on a distributed, modular architecture divided into three complementary layers:

  • Frontend (Telemetry Dashboard): Developed using React 19, TypeScript, Vite 6, and Tailwind CSS v4 to provide a dark-mode command center console, leveraging Recharts to map real-time community health data.
  • Backend (Analysis Engine): A robust Node.js and Express API connected to the Google GenAI SDK to execute deep semantic contextual analysis using the Gemini engine.
  • Platform Integration: A native extension built with the official Reddit Devvit SDK, exploiting the Menu Item API and Forms API to capture thread data and display moderation guidelines seamlessly without leaving the platform.

Built With

Share this project:

Updates