ModMindUltra

AI-Powered Mod Queue Triage Command Center

Comment

ModMind — AI-Powered Mod Queue Triage Command Center

🛡️ Inspiration

Every active subreddit’s mod queue is a firehose. Reported content, AutoModerator holds, and unreviewed posts pile up faster than volunteer mod teams can process them. The result: high-risk content stays up too long while low-risk content wastes valuable moderator time.

Existing tools (AutoModerator, Toolbox) help, but they rely on pattern matching, not context comprehension. We wanted to build a tool that actually reads, understands, and prioritizes queue items based on your subreddit's unique rules.

📊 What it Does

ModMind is an AI-powered moderation command center built directly on Reddit’s Developer Platform (Devvit). It features:

  1. Subreddit Rules-Aware AI Triage: ModMind fetches your subreddit’s official rules dynamically via the Reddit API and triages posts/comments against your community standards. Hovering over violation badges shows a tooltip with the rule's full description.
  2. Interactive SVG Analytics Dashboard: Rendered natively in pure React SVG (no heavy charting libraries), moderators get an immediate visual health report, including a gamified Subreddit Health Score, a Donut Chart of risk levels, and a Bar Chart of top rule violations.
  3. AI-Drafted Removal Comments: When a mod removes an item, ModMind opens a drawer with a custom, polite removal notice written by the AI, explaining which rules were violated. Mods can edit the text and toggle an option to post it as a stickied, distinguished comment.
  4. Bulk Action Automation: Clear queues instantly with one-click actions: "Approve all Low-Risk items" or "Remove all Critical-Risk items".
  5. Persistent Webview Configuration: Adjust the AI strictness (Lenient, Moderate, Strict), LLM model (Llama 3.1 8B vs Llama 3.3 70B), and comment tone (Polite, Firm, Direct) directly inside the app, saved to the Devvit KV Store.
  6. Mod Audit Log: Logs all actions taken via the dashboard in the KV Store, displaying a complete history in the "Mod Action Logs" tab.

🛠️ How We Built It

  • Platform: Reddit Devvit (Developer Platform)
  • Backend: TypeScript, Devvit Public API (using KV Store, getRules, submitComment, and getModQueue).
  • Frontend: React 18, Vite, and custom CSS (pure Vanilla CSS for modular design).
  • AI Triage: Llama 3.1 8B (fast/free) & Llama 3.3 70B (high reasoning) via Groq API.
  • Charts: Custom SVGs drawn dynamically in React.

🚀 Accomplishments we're proud of

  • Eliminating standard iframe loading race conditions in Devvit by implementing a robust state-handshake retry mechanism.
  • Building custom SVG charts natively without using heavy external canvas libraries, making the dashboard load instantly.
  • Automating the tedious task of drafting personalized removal reasons, which moderators normally spend hours writing.

Built With

Share this project:

Updates