A high-performance NLP pipeline — tokenization, sentence boundary detection, negation handling, sentiment analysis, POS tagging, NER, and custom entities — in a single zero-dependency package with full TypeScript support. The entire pipeline runs at 1M tokens per second, identically in Node.js and browsers. Run the live benchmark on Observable. Explore the documentation and recipes.

From IIoT streams to actionable insight — in real time. Composable blocks for signal conditioning, statistics, anomaly detection, and flow control — from Kalman filters to spiking neural networks. Express what you want, not how to build it; AI-native via MCP Server. A change point detection pipeline runs at 1M messages per second on a modern server, 100K on a Raspberry Pi — same code, edge to cloud. Run the live benchmark in your browser. Explore the documentation and use cases.

Our thanks to Zulip and Coveralls for sponsoring winkJS — their support helps us keep the work open and the quality uncompromised.

A star on winkNLP or winkComposer helps more developers find the work, and genuinely encourages a small team shipping in the open.

_{Focused utility packages for text processing, search, similarity, and statistics — see winkjs.org/packages.}