Project: Discord Community Bot

Summary

A large gaming community on Discord ran into growing pains as membership scaled. The community issued members roles — both cosmetic and functional — as accolades, but every request had to be reviewed manually by staff before being applied. As membership grew from tens of thousands toward six figures, the backlog became unmanageable.

A second problem compounded this. The game the community was built around required specific in-game unlocks for crafting and progression, so members frequently sought out other players who had those unlocks. This created opportunities for theft and scams. The community existed partly to crowdsource trust, but their reputation system was rudimentary and easily gamed — there was no reliable way to detect bad actors.

Solution

I solely developed a Discord bot and an accompanying web platform to address both problems, along with a suite of quality-of-life utilities for staff.

For roles, the bot integrates directly with the game’s REST API to automatically determine which roles each member is eligible for and assign them without staff involvement. This alone cut the role-review workload by approximately 70%.

For trust, I built a full reputation system backed by a relational database that now holds over 580,000 individual reputation entries. Members can look up another user’s reputation history before handing over valuable materials, and leave reputation afterwards. All incoming reputation is held for staff review in a web interface built with WebSockets so reviewers always see live, conflict-free data.

I also developed a machine learning classifier to automatically categorize reputation submissions into types — crafting services, item transfers, and middleman (escrow) transactions — reducing the manual triage burden on moderators further.

Impact

The bot has been in continuous production since 2019. The automation and trust infrastructure it provides supported the community’s growth from 20,000 members to over 120,000, with membership still growing. Staff time previously spent on manual reviews is now directed toward moderation and community engagement, and members transact with confidence knowing bad actors are surfaced before damage is done.