Open-source LLM safety guardrails: prompt injection protection, PII redaction, toxicity filtering, and OpenAI-compatible AI proxy
DeepEval — independent third-party profile of a public API surface, by API Evangelist. DeepEval is an open-source LLM evaluation framework — built and maintained by Confident AI — for testing and benchmarking large language model applications. It is structured like Pytest but specialized for LLM systems, providing 40+ research-backed metrics (G-Eva
A comprehensive, agent-agnostic evaluation framework that tests any AI agent across 41 structured test cases using automated 5-layer evaluation and adversarial robustness testing. Features multi-dimensional scoring with rule-based safety checks, LLM-as-judge assessment, semantic similarity analysis, toxicity detection, and RAG grounding evaluation
Fast, drop-in multilingual profanity and toxicity screener for Node.js, powered by System 1 models like TypeSafe AI Jev and Laya. Catches leetspeak, character spacing, and romanized profanity across languages including Kannada, Telugu, Tamil, Hindi, and Bengali. ~50-500ms latency.
TypeScript toolkit for testing LLM outputs: hallucination, quality, regression and toxicity checks (OpenAI, Anthropic)
A GitHub Action that hides abusive issue, pull request and discussion comments with the Unprofane profanity and toxicity API
Open-source ML-powered profanity filter with TensorFlow.js toxicity detection, leetspeak & Unicode obfuscation resistance. 21M+ ops/sec, 23 languages, React hooks, LRU caching. npm & PyPI.
The Scunthorpe Test: a profanity-filter benchmark where blocking Scunthorpe counts against you
Self-hosted content moderation API that outperforms Amazon Comprehend. 100% offline, your data never leaves your server. Text + Image moderation.
JavaScript and TypeScript client for the Unprofane profanity and toxicity API (@unprofane/sdk)
Unified AI Trust & Safety Workbench - Classical ML for evidence verification (FEVER), misinformation detection (LIAR), toxicity analysis (Jigsaw), and length-aware algorithmic fairness.
Multilingual profanity filter with a Rust core — Python, JavaScript/WASM and Rust, with optional multi-label toxicity scoring
Toxicity detection & moderation: fine-tuned DistilBERT (6 Jigsaw labels, macro ROC-AUC 0.989), word-level explanations, review queue, FAISS similar-case search, identity-bias mitigation. FastAPI + vanilla JS.
Python client for the Unprofane profanity and toxicity API (pip install unprofane)
Real-time toxic chat message filter using TF-IDF + Ridge Regression, with warn/block thresholds and auto-muting. FastAPI + Streamlit.
Zero-shot TypeSafe JEV versus logistic-regression and DistilBERT baselines on Jigsaw toxic-comment classification, with a three-way moderation policy and a console. Frozen evaluation protocol.
🤖 API Endpoint for the Redacted Discord Bot
🛡️ Open-source moderation powered by AI
Real-time NLP toxicity classification microservice using TF-IDF & Logistic Regression (90.72% accuracy on 77.8k Turkish comments).
Russian text toxicity detection.
To associate your repository with the toxicity-detection topic, visit your repo's landing page and select "manage topics."