• 

  Problem Statement

• 

  Architecture

RuleForge — Devpost Story Sections

Inspiration

A moderator removes a post for breaking Rule 5. Another moderator looks at the same post and approves it back. This happens thousands of times daily across Reddit — and nobody tracks it.

We dug into the research. Academic studies found that 87% of gray-area mod decisions trace back to poorly-written rules, and 13.5% of all mod actions are disagreed upon by fellow moderators. That's not a communication problem — it's a visibility problem. Mod teams have zero data connecting their rules to how those rules are actually enforced.

We analyzed 50+ hackathon submissions and found exactly zero tools in this space. RuleForge was born to fill that gap.

What it does

RuleForge gives every subreddit an enforcement intelligence layer. Install it, and you immediately get:

• Health Score (0-100) telling you how well your rules are working
• 6-tab dashboard — Overview, Heatmap, Consistency, Problems, Trends, Workload
• Rule Autopsy — click any rule for a full-page breakdown: who enforces it, how often it's overridden, what mods think about it, and what to do about it
• Auto-tagging — mod actions are silently matched to rules in the background, zero extra work
• Burnout detection — see if one mod is handling 40% of all removals while others are idle
• One-click action — start a mod discussion post pre-filled with stats and annotations
• Weekly modmail digest — enforcement summary delivered every Monday

The dashboard comes pre-populated with sample data on install, so the value is visible in seconds.

How we built it

Entirely on Devvit — no external APIs, no servers, no dependencies.

Frontend: React 19 + Tailwind CSS 4 + Chart.js inside a Devvit custom post webview Backend: Hono server with 9 API endpoints, running on Reddit's infrastructure Storage: Redis sorted sets for enforcement counts, hashes for mod-rule matrices, TTL keys for deduplication and override detection Intelligence: Template-based recommendation engine analyzing override rates, annotation keywords, and consistency scores

We use 7 Devvit APIs — Custom Posts, Menu Actions, Forms, Triggers, Scheduler, Redis, and Reddit API. The ModAction trigger fires on every removal and approval with 24-hour deduplication. A daily scheduler at 3 AM UTC aggregates stats, flags problem rules, and detects health score improvements.

Challenges we ran into

The auto-tag problem. Reddit's ModAction events don't always include why a post was removed. We built a keyword matching engine that tokenizes rule names and action descriptions to find the best match — and an "untagged counter" as a fallback so mods know what was missed.

The seed data problem. An empty dashboard doesn't impress judges. But hardcoded sample data breaks when installed on a different subreddit with different rules. Our solution: the seed function fetches the subreddit's actual rules via the Reddit API and generates proportional data using those exact rule names. Install on any subreddit — the dashboard adapts.

The override detection problem. How do you know a removal was reversed? We store a temporary Redis key when content is removed, and check for it when an approval event fires on the same content. If found — that's an override, and it gets tracked.

Accomplishments that we're proud of

11 features that work together as a product, not a prototype. Auto-tag, manual tag, override detection, 6-tab dashboard, Rule Autopsy, smart recommendations, workload balance, impact ranking, one-click discussions, weekly modmail, health milestones.

The Rule Autopsy. Click any rule name anywhere in the dashboard and get the full story — per-mod enforcement chart, 12-week trend, all annotations, consistency score, and actionable recommendations. No tool on Reddit has ever shown this view.

Zero behavior change required. Mods don't need to learn anything new. The auto-tag engine works silently. The weekly digest arrives automatically. The dashboard is pre-populated on install. Value is immediate.

What we learned

Devvit is production-ready. The combination of Redis, triggers, scheduler, and webview is powerful enough to build real analytics tools — not just simple bots.

Seed data is critical for demos. Pre-populated dashboards communicate 10x more value than empty ones. But they must adapt to the subreddit's actual rules to feel authentic.

The research shaped the product. The 87% gray-area statistic from academic studies isn't just a talking point — it's the reason Rule Autopsy exists. If most enforcement problems come from poorly-written rules, the tool should help you find and fix those rules.

What's next for RuleForge — Rule Enforcement Analytics

• Real-time spike alerts — notification when a rule's override rate suddenly increases
• Mod training mode — show new moderators how experienced mods enforce each rule
• Cross-subreddit benchmarking — compare your enforcement health against similar communities
• AI-powered rule rewrites — analyze annotations and suggest improved rule text
• Configurable settings — adjustable thresholds, report frequency, tagging preferences

Long-term, RuleForge aims to become the standard enforcement intelligence layer for every subreddit on Reddit. Rules should be living documents — not dead wiki pages.

Built With

• chart.js
• devvit
• hono
• react-19
• redis
• tailwind-css-4
• typescript
• vite