Toxicity API

Detect toxicity in a sentence via vector similarity against a seeded toxic-term corpus.

• Web: https://toxicity.its-mrinank.com
• API: https://toxicityapi.its-mrinank.com

Layout

apps/
  web/   Next.js frontend
  api/   Fastify service — Postgres + pgvector + Cloudflare Workers AI embeddings

API

curl -X POST https://toxicityapi.its-mrinank.com/ \
  -H "Content-Type: application/json" \
  -d '{"message":"hello there"}'

Response:

{ "isToxic": true,  "score": 0.97, "flaggedFor": "<word>" }
{ "isToxic": false, "score": 0.42 }

Local development

pnpm install

# api
cp apps/api/.env.example apps/api/.env   # fill DATABASE_URL, CF_ACCOUNT_ID, CF_API_TOKEN
pnpm migrate                              # applies scripts/001-init.sql
pnpm seed                                 # embeds training_data.csv
pnpm dev:api                              # Fastify on port 3001

# web
cp apps/web/.env.example apps/web/.env.local
pnpm dev:web                              # Next.js on port 3000

Tests: pnpm test.