ModQueue Copilot

• 

  Modqueue-copilot

Inspiration

Moderating active subreddits can feel like drinking from a firehose. The standard moderation queue is a chaotic, endless stream of reports, spanning everything from harmless banter to coordinated brigading or toxic speech. For moderators, reading long threads, checking report histories, and making decisions for every single item takes massive amounts of cognitive load and between 2–5 minutes per item. We wanted to build a co-pilot that acts as a moderator's second brain—instantly classifying severity, highlighting exactly why a post or comment is flagged, and offering a one-click decision engine, all built natively on top of Reddit’s Devvit platform.

What it does

ModQueue Copilot is an intelligent assistant for Reddit moderators that streamlines the review process down to seconds. Key capabilities include:

• AI-Powered Triage Engine: Automatically analyzes content (toxicity, spam, brigading, NSFW, and high report volumes) and categorizes reports into three severity tiers: Urgent (🔴), Review (🟡), or Safe (🟢).
• Interactive Glassmorphism Dashboard: A beautifully designed React-based dashboard embedded directly inside Reddit. It provides real-time counts of pending items, time saved, and interactive cards for queue filtering.
• One-Click Actions: Approve, Remove, Lock, or Escalate items directly from the queue cards with automated rule citations and custom notes.
• Bulk Mop Operations: Clean up entire comment threads or post comments instantly by removing/locking comments with optional filters to skip distinguished mod comments.
• Analytics Hub: Track moderation speed, total time saved, and distribution of flags.
• Customizable AI Rules: Subreddit-specific settings to fine-tune toxicity thresholds, toggle spam/brigading rules, and adjust detection sensitivity.

How we built it

We built ModQueue Copilot completely from scratch utilizing modern web technologies and the Devvit API:

• Backend & Routing: Built on Node.js and Devvit CLI using Hono as an lightweight, high-performance API router to manage custom triggers, forms, and menu endpoints.
• Frontend / Client: A highly interactive, responsive React application built with Vite and TypeScript. We crafted a premium, dark-mode glassmorphism interface with custom CSS variables, skeleton shimmers, and micro-animations for exceptional aesthetics and a premium feel.
• State Management & Persistence: Leveraged @devvit/redis to persist queue items, settings, and analytics metrics safely across server restarts.
• AI Triage Logic: A robust rule-based NLP classifier that extracts and matches flags, adjusts confidence scores dynamically, and incorporates external factors like report counts.

Challenges we ran into

• Webview Integration: Developing a smooth, fast React single-page app that runs inside Devvit's sandbox webview required optimizing assets and bundling setup using Vite.
• Dynamic Context Passing: Synchronizing data between Reddit's native context menus (right-clicking comment/post to trigger forms) and the independent interactive dashboard post.
• Minimizing Bundle Size: Keeping dependencies minimal to fit within Devvit's execution thresholds while maintaining a feature-rich analytics dashboard.

Accomplishments that we're proud of

• Seamless Native Feel: Creating a dashboard that looks like a native, premium extension of Reddit, using modern glassmorphic designs, CSS animations, and smooth transitions.
• Zero-Configuration Installation: Setting up automatic data seeding so that when the app is installed, moderators have demo data in their queue immediately to start testing.
• High Utility: Realizing a tool that drastically improves a moderator's day-to-day workflow, saving an estimated 3.2 minutes per queue item.

What we learned

• Power of Reddit Devvit: Devvit's custom posts and webview framework allow building incredibly rich user interfaces that go far beyond standard text-based bots.
• Moderator Needs: Speed and visual clarity are crucial. Designing a tool to reduce cognitive load requires highlighting the "why" (rules breached, confidence scores) as much as the action.

What's next for ModQueue Copilot

• LLM Integration: Moving from rule-based NLP matching to live, serverless LLM calls for semantic reasoning and context awareness.
• Coordinated Attack Spotter: Incorporating time-series analytics to flag coordinated brigading attacks across the subreddit in real-time.
• Multi-Subreddit Syncing: Letting moderators of multiple subreddits manage all their queues from a single consolidated dashboard.

Built With

• css3
• devvit-sdk
• esbuild
• hono
• node.js
• react
• redis
• typescript
• vite