You get an in-app notification when a new version is published. Nothing is installed or updated on your computer automatically.
What it does
Tells you whether a text looks appropriate, using two engines. Offline word lists catch strong insults and profanity in 20 languages (leetspeak and missing accents included) and mask them with * in a filtered copy. With your own OpenAI key it also runs OpenAI omni-moderation, which understands meaning: harassment, hate, violence, sexual content, self-harm and illicit content. It can check images too. It is a first pass, not a replacement for human moderators.
🧪 Try it on MLAIEL
Choose the model before you run. “MLAIEL key (OpenAI)” uses the platform key. Your own OpenAI key, Leonardo and Luma are yours: paste them for this request only. MLAIEL does not save them. Pollinations, when it is offered, is free and needs no key. Inputs and results are checked, nothing you send is stored, and you get 8 tries per day across all tools.
Sign in with a verified email to try this tool. Sign in
Downloads come straight from the source repository. MLAIEL does not host files.
🔌 Integrate
Follow these steps in order. If you get stuck, ask in the improvement project.
Use as a library
import { moderateText } from "./multilingual-moderation/src/index.ts";
const result = await moderateText("Your text here");
console.log(result.appropriate, result.filteredText, result.openai?.categories);
Command line
node --env-file=.env examples/cli.ts "Your text here"
Works with
Node.js 22.18+ (runs TypeScript directly, no build step) · Windows, macOS, Linux (tested on Windows 11) · No runtime npm dependencies
License
License: MIT
MIT, declared by the owner: Fahed Mlaiel, who holds all rights to Mlaiel/iacherie, chose the MIT License for this clean TypeScript rewrite on 2026-10-01. The package ships the LICENSE file. This is not legal advice.
This is not legal advice. Check the license yourself before you use or share this tool.
Security check
No secrets found in the checked files.
No risky install commands found.
Checked on Oct 1, 2026
How it was tested
Run test · Passed
Windows 11, Node.js 22.18, package 1.0.0 run from source with a real OpenAI key, 2026-10-01 · Oct 1, 2026
Moderated clean and abusive samples in several languages; abusive terms were caught and masked, clean text passed.
npm test passes (package unit tests)
Local word lists flag German, French and English insults and mask them
OpenAI omni-moderation engine called with a real key and merged with the local verdict
Runs without a key (local engine only) and says so in the result
Run test · Passed
Windows 11 · CPython 3.12.7 · fresh virtual environment with httpx only · all provider keys removed · Oct 1, 2026
Imported in isolation and ran offline. Harmless text was marked appropriate; a word from the English list was flagged and masked with *.
Import with only httpx installed: passed
Harmless sentence → is_appropriate = true
Word from the English list → flagged, masked with *
Moderation languages available: 20 (README claims 644+, not verified)
Versions and changes
0.1.0 → 1.0.0
Oct 1, 2026 · source revision mlaiel packages/marketplace/[email protected]
1.0.0 is a clean TypeScript rewrite that runs on its own. New 20-language word lists, OpenAI omni-moderation for meaning and images, and 18 offline tests. Validated with the real OpenAI API.
Clean TypeScript rewrite of iacherie moderation_multilingual.py.
New word lists for 20 languages, leetspeak and accent folding, substring matching for zh/ja/th.
New OpenAI omni-moderation engine for meaning (text) and moderateImage() for images.
Matched words are never returned in clear, only masked.
0.1.0 · first release
Oct 1, 2026 · source revision 54911be
First listing of the iacherie moderation module, pinned to the analyzed revision. No code changes compared to the source repository.
Updates are never applied silently. You decide if and when to download a new version.