clearchat.tsx

ClearChat

2025live

AI-powered Slack moderation bot — a three-stage pipeline that combines keyword caching with GPT-4o-mini review to catch problematic messages without lighting money on fire.

Accuracy

~93%

False positives

~2%

Inference cost reduction

>95%

Story

The straightforward way to do AI moderation is to send every message to an LLM. The bill arrives in week two and you cancel the experiment. ClearChat keeps the accuracy and kills the bill: a three-stage pipeline routes most messages through cheap pattern matching and a cached classification layer, sending only the genuinely ambiguous cases to GPT-4o-mini for nuanced review.

On test scenarios the staged pipeline holds ~93% accuracy with ~2% false positives — and inference cost drops by more than 95% compared to calling the LLM on every message.

What I built

1Designed the three-stage pipeline: keyword filters → cached classification → GPT-4o-mini review for ambiguous cases
2Implemented Prisma-backed persistence for caching and audit trails on Postgres
3Tuned thresholds against a labeled test corpus to balance recall and cost

Tech

TypeScriptNode.jsPrismaPostgreSQLOpenAI APIGPT-4o-mini

Up Next

Story

On test scenarios the staged pipeline holds ~93% accuracy with ~2% false positives — and inference cost drops by more than 95% compared to calling the LLM on every message.