Orbsiss

Inspiration

Moderating active subreddits is a balancing act. AutoModerator is powerful but lacks a real-time, user-friendly visual interface, while standard Reddit mod queues can feel slow during high-traffic spikes. We wanted to build a modern, visual control center for subreddit moderators.

My goal was to create Orbsiss—a premium real-time dashboard that combines automated keyword-based moderation (the "Shield"), active queue management, warnings logs, scheduled megathreads, and custom rules in a single unified dashboard, directly integrated into Reddit's Devvit platform.

What it does

Orbsiss serves as the command center for your subreddit's moderation team:

Command Center Dashboard: Visualizes subreddit health metrics, recent moderation actions, and keyword trigger statistics.

Shield Auto-Moderator: A custom keyword scanner. When users post or comment, the Shield scans the content instantly. If a banned keyword is found, the content is automatically flagged or removed, a moderation warning is issued to the user, and an alert is broadcasted to the dashboard.

Custom Rules Manager: Directly integrates with the subreddit's rule list. Moderators can pull rules live from Reddit and add new subreddit rules dynamically from the app.

Interactive Moderator Queue: An optimized queue layout showing recent violations, flagging reasons, and moderator quick actions.

Warnings & Profile logs: Track user profiles, warnings histories, bans, and moderation notes.

Megathread Organizer: Schedules and publishes recurring megathreads with template-based post content.

Response Templates: Manage custom template responses for warning messages and ban notices.

How we built it

Orbsiss is designed for maximum speed and robustness under Devvit's serverless environment:

Algorithms: We implemented a Trie data structure to search for multiple keyword patterns in a single pass. To avoid false positives, each match is verified using the Knuth–Morris–Pratt (KMP) exact-match algorithm.

$$ \text{Trie Multi-Scan: } O(N \cdot K) $$

$$ \text{KMP Verification: } O(N + M) $$

Where:

(N) is the length of the post content
(K) is the height of the keyword trie
(M) is the length of the matched pattern

Backend: Built using Hono running in Devvit's serverless Node.js environment. We leveraged Redis for rules caching, user mapping, and moderation queues.

Frontend: Built with React 19 and Vite running inside a Reddit Web-iFrame entrypoint.

Design & Styling: Styled using Tailwind CSS 4 and custom vanilla CSS variables for a premium, fast-loading, glassmorphic dark-mode interface without heavy external component libraries.

Challenges I ran into

Serverless Execution Limits: Doing complex string matches on long self-posts within Devvit's low-latency serverless limits required highly optimized matching. We solved this by compiling our keywords into a Trie to scan entire posts in a single pass instead of looping over every keyword individually.
React 19 Transitions: Adapting to React 19 typings meant overcoming deprecations of global types like React.FormEvent in favor of importing clean SyntheticEvent instances, while maintaining strict TypeScript rules.
App Versioning & CLI Conflicts: When testing and deploying the app using the Devvit CLI, multiple active daemons would conflict on AppVersion locks. Terminating background services and configuring clean type-check/lint build pipelines resolved this.

Accomplishments that we're proud of

Trie-KMP Search Engine: Highly performant content scanner that reliably filters out malicious keywords with zero lag. Beautiful UI/UX: Created a high-fidelity dashboard that matches modern dark-mode aesthetic standards while loading in under a second. Clean Devvit Integration: Successfully connected Hono routing, Redis caching, real-time dashboard pub-sub sync, and official Reddit Subreddit API actions under one hood.

What I learned

How to build fully-featured React apps inside Devvit web iFrames. Best practices for state synchronization and real-time pub-sub broadcasts between serverless backend triggers and frontend clients. The performance difference of single-pass multi-pattern searching in high-throughput social applications.