Why does the Worker ask two questions for usernames?

Look-alike spellings and phonetic puns are hard to catch with one literal question, so kind set to username also asks whether the handle disguises profanity, and flags when either probability clears the threshold.

How do I tune false positives?

Raise or lower PROFANITY_THRESHOLD, or treat a middle band such as 0.3 to 0.7 as send-to-human-review. The README notes rewording question criteria is usually more effective than moving the threshold.

What do I need to deploy profanity-checker?

A Cloudflare account with Workers enabled and AI Gateway credits. `typesafe/jev` is a third-party Workers AI model billed through AI Gateway Unified Billing as prepaid credits, so without them inference fails with `2021: Insufficient AI Gateway credits`. You also need Node.js 18 or newer.

How do I call the Worker from another Worker?

Two ways. A Service Binding is recommended for the same account: add `PROFANITY` to the consumer's `wrangler.jsonc` services and fetch `https://profanity-checker/api/profanity-check` through `env.PROFANITY`. Otherwise a plain fetch to the `*.workers.dev` URL works across accounts, but that endpoint is public.

What does the response contain?

A `success` flag, the echoed `text` and `kind`, a boolean `is_profane`, the `probability` (the max across questions), a `probabilities` map of raw per-question values, and the `threshold` used. Keeping the raw values lets you apply your own policy.

What are the known limits of the profanity check?

Two-hop phonetic gags that resolve to explicit anatomy, like `mike_hunt`, score low and stable around 0.16 even when listed in the criteria; the README calls it a calibration limit, not a wording bug. Look-alike spellings and one-hop puns are caught reliably, and Jev's 32,000-token window means long inputs must be truncated.

Can I reuse this repo to call Jev from my own Worker?

Yes. The README shows an `env.AI.run("typesafe/jev", ...)` call with a typed noul question, and notes that third-party models return `{ state, result, gatewayMetadata }`, so answers live under `envelope.result.answers`. It requires an `ai` binding in `wrangler.jsonc`.

CommunityJev Integrations & Adapters

Profanity Checker

Cloudflare Worker that asks Jev on Workers AI for literal profanity in text or usernames and, for usernames, a second Noul for phonetic or look-alike disguises. The threshold and max policy live in Worker code, with OpenAPI docs generated via chanfana and Hono.

See highlights

Open github.com

Stars: 1
Language: TypeScript
License: MIT
Published by: Community
Last verified: Oct 2026

4rays/profanity-checker

Organization repository on GitHub

View on GitHub

Cloudflare Worker to check for profanity using TypeSafe Jev

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A Cloudflare Worker that checks text and usernames for profanity using TypeSafe's Jev (typesafe/jev on Workers AI). Jev is a structured evaluation model: instead of generating text, it answers typed yes/no questions with calibrated probabilities. This Worker asks it one or two narrow questions per request and turns the probabilities into a boolean verdict in code. OpenAPI docs are generated from the schema via chanfana + Hono.

Quoted from the project's README. Read the full README.

Tagscommunity cloudflare-workers moderation profanity

Works forJev for Moderation Jev for Decision gating

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: Cloudflare API
Capabilities: Open sourceRequires API key

Highlights

Calls typesafe/jev through a Workers AI binding; one question for text, two in parallel for usernames
kind username catches look-alike spellings (a55h0le) and one-hop puns (ben_dover)
Threshold (default 0.5) and max policy live in src/endpoints/profanityCheck.ts; raw probabilities returned
Serves as a minimal reference for calling Jev from any Worker, documenting the envelope.result.answers shape
Callable from other Workers via service bindings with no public URL and no auth needed

Watch out

MIT license. Needs prepaid AI Gateway credits for the third-party typesafe/jev model. Two-hop phonetic gags such as mike_hunt score low (about 0.16) regardless of criteria, and inputs over the 32k token context must be truncated.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

YouTube: Building a contextual brand-safety scorer on Jev from scratch.

X: A website slop detector that checks 35 tells in 243 ms

X: Breaking, golden nugget, or slop

X: A filter for reply-guy comments

More Jev resources like this

Brand Safety Feed

A contextual brand-suitability scorer that judges each feed item against an advertiser's own versioned policy, returning a category, an exclusion flag, and a confidence for routing borderline items to review.

Jev Integrations & Adapters#community#typescript#advertising

Community

LiteLLM

60.9k

LiteLLM's proxy adds TypeSafe guardrail hooks and a Jev complexity classifier for policy judgments and request routing across 100+ model providers.

Jev Integrations & Adapters#community#python#gateway

Community

BerriAI/litellmPython

Inbox Zero

12.4k

Open-source AI email assistant (Next.js, Postgres, 12k+ stars) with a decision-model layer that can run on openrouter:typesafe/jev-1.13: Jev-backed questions cover rule selection, cold-email detection, thread status, recurring patterns, unsubscribe pages, and sender categorization, each with evals.

Jev Integrations & Adapters#community#email#gmail

Community

elie222/inbox-zeroTypeScript

Atomic

856

Verifiable coding agent runtime with 855 stars. Ships a first-class Jev structured-output provider so agent decisions come back typed through the same decision resolver as its other providers, alongside plain-English workflows, approval gates and evidence-backed verification stages.

Jev Integrations & Adapters#community#coding-agents#verification

Community

bastani-inc/atomicTypeScript

GoEventBus

High-performance Go event bus with a bounded MPMC ring buffer, fan-out and ordered handlers, plus an optional routing layer that evaluates deterministic rules first, reuses cached decisions, and asks Jev a typed Choice only for ambiguous cache misses.

Jev Integrations & Adapters#community#go#event-bus

Community

Protocol-Lattice/GoEventBusGo

Kilo Code Auto-Routing

Merged PR in the Kilo Code cloud monorepo that replaces its LLM auto-routing classifier with one OpenRouter /systemone Choice request to typesafe/jev-1.13; the PR reports 0.93-0.94 route accuracy on 72 golden cases at roughly $0.00028 per call, with a 250 ms transport-error retry.

Jev Integrations & Adapters#community#routing#coding-agents

Community

Kilo-Org/cloudTypeScript

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai