🐚 PearlMod: The Moderator Command Center

PearlMod is a high-fidelity, proactive moderation suite built for the Reddit Developer Platform. It empowers community leaders with client-side AI, automated enforcement engines, and real-time threat intelligence.

💡 Inspiration

Moderating high-traffic subreddits is often a race against volume. Standard tools are frequently reactive, leaving moderators to "mop up" toxicity after it has already spread. We built PearlMod to transform moderation from a chore into a mission control experience—giving moderators the "superpowers" of semantic intelligence and automated guardians to keep communities healthy around the clock.

🚀 What it does

PearlMod is a unified dashboard that consolidates fragmented moderation tasks into a single, high-fidelity interface:

• AI Mod Queue Assistant: Uses a client-side Universal Sentence Encoder (USE) model to semantically scan reports. It provides "Auto-Analyse" capabilities for locked items, highlighting problematic evidence and providing risk scores without data ever leaving the browser. Also Moderators have lock feature to mitigate ambiguity when resolving the reports.
• Guardian Engine: A dynamic rule-based system. Moderators can create rules with blacklisted keywords and custom "Toleration Limits."
  • Post Actions: Automatically flairs violating posts and sends explanatory Private Messages (PMs) to authors.
  • Comment Actions: Immediately removes violating comments. If a thread exceeds the user-defined toleration limit, the entire post is locked and the owner is notified via PM.
• Raid & Brigade Protection: Real-time detection of rapid participation spikes, coordinate attacks, and "Hostile Bursts."
• Cases & Escalation: A dedicated workflow for tracking complex moderation cases that require team-wide collaboration and follow-up.
• Live Analytics: Interactive visualizations of moderation trends, solve rates, and community health metrics.

🛠️ How I built it

PearlMod is built on a modern, robust stack designed for the Reddit ecosystem:

• Frontend: React 19 with a custom "Mission Control" dark theme, styled with Tailwind CSS and powered by Lucide-React icons.
• Intelligence: TensorFlow.js running the Universal Sentence Encoder (USE). We optimized the model to run on the CPU backend within the Reddit Webview for maximum compatibility.
• Backend: Hono for lightweight, high-performance API routing.
• State & Storage: Redis handles everything from real-time violation counters and rule storage to escalated case snapshots.
• Platform: Built exclusively for Devvit, utilizing the full power of Triggers (Content Submit, Reports, Mod Actions) and the Webview API.

🚧 Challenges I ran into

• Asset Delivery: Integrating a 30MB+ ML model into a Reddit Webview was challenging. We had to move the model shards to the public directory and manage static asset fetching manually to avoid 404 errors during the build process.
• Trigger Payload Nuances: Reddit's trigger payloads can be highly inconsistent. We had to build a robust data extraction layer to handle variations in how author and content fields are populated across different event types.
• Real-time Sync: Coordinating between background triggers (Automod removals) and the moderator's live dashboard required a carefully designed Redis state management system to ensure data consistency.

✨ Accomplishments that I'm proud of

• Zero-Latency Enforcement: Achieving immediate comment removal and thread locking based on semantic matches before a moderator even sees the item.
• Privacy-First AI: Running sophisticated semantic analysis entirely on the client side, ensuring moderator activity and community data remain private.
• Unified UX: Creating a professional-grade dashboard that feels like a premium SaaS product but is available for free to community volunteers.

🧠 What I learned

• ML in Constraints: Learned how to effectively deploy and optimize machine learning models in restricted web environments like the Reddit Webview.
• Event-Driven Moderation: Gained deep insights into Reddit's trigger-based architecture and how to build resilient automated systems around it.
• Mission-Critical Design: Focused on how to design UIs for high-stress environments where clarity and speed are paramount.

🔮 What's next for PearlMod

• Improvement in UI: UI will be improved much better.
• Multilingual Support: Integrating lightweight multilingual models to support non-English speaking communities.
• Mobile Command Center: Further optimizing the dashboard for a seamless "on-the-go" moderation experience in the Reddit mobile app.

Built With

• devvit
• typescript