AI and machine learning

12-month change compares the average of the last 4 weeks of downloads with the first 4 weeks of the 52-week series. Gzip size is for the whole package, as measured by Bundlephobia. "-" means the value has not been fetched.

• gptGUID Partition Table (GPT)
• mlflow-tracingTypeScript implementation of MLflow Tracing SDK for LLM observability
• novel-segmentChinese word segmentation 簡繁中文分词模块 以網路小說為樣本
• gpu-timeA compact neural parser for English schedules
• @joshuaswarren/openclaw-engramDeprecated compatibility shim for Engram installs. Re-exports @remnic/plugin-openclaw and forwards engram-access to @remnic/core.
• @automagik/omniLLM-optimized CLI for Omni
• @desert-ant-labs/clearOn-device speech enhancement for JavaScript: denoise, dereverb, and loudness-normalize to a podcast-ready 48 kHz. Runs in the browser (WebAssembly + LiteRT.js) and server-side in Node (native), from one import.
• @desert-ant-labs/earOn-device spoken language identification for JavaScript: name the language of a recording before you transcribe it. Runs in the browser (WebAssembly + LiteRT.js) and server-side in Node (native), from one import.
• @hamming/hamming-sdkSDK for Hamming Evals Framework
• sherlockjsNatural-language event parser for Javascript.
• @bitfab/sdkBitfab client for provider-based API calls with local BAML execution
• opencode-auto-resumeOpenCode plugin that automatically resumes stalled LLM sessions when thinking/streaming freezes mid-generation.
• pi-provider-litellmLiteLLM proxy provider extension for Pi
• streamdown-svelte<p align="center"> <img src="https://raw.githubusercontent.com/PacificStudio/streamdown-svelte/master/static/favicon.svg" alt="Streamdown Svelte favicon" width="96" height="96" /> </p>
• @blockprotocol/graphImplementation of the Block Protocol Graph service specification for blocks and embedding applications
• fetch-sseAn easy API for making Event Source requests, with all the features of fetch(), Supports browsers and node.
• scanic-mlOptional ML document-corner detector assets for scanic, a channel-slimmed SimCC model (DocCornerNet) and a custom minimal ONNX Runtime Web WASM build (1.5 MB, ~88% smaller than stock ort-web).
• @remnic/capture-screenDesktop screen-activity capture daemon for Remnic — local spool, loopback HTTP API, and replay ingestion for the activity source `screen` (à-la-carte)
• @remnic/capture-audioDesktop audio capture daemon for Remnic — local spool, loopback HTTP API, and replay ingestion for the wearable source `desktop` (à-la-carte)
• @mars-sea/dsh-commandcode-providerUnofficial DeepSeek Harness LLM provider plugin for Command Code, ported from pi-commandcode-provider (MIT). Registers the 'commandcode' provider route with a Models-page card and live model catalog.
• @upstash/semantic-cacheA Semantic cache built on top of Upstash Vector DB
• @ruvector/ruvllm-darwin-arm64RuvLLM native SIMD acceleration - darwin-arm64 (Apple Silicon) platform
• linearisCLI tool for Linear.app with JSON output, smart ID resolution, and optimized GraphQL queries. Designed for LLM agents and humans who prefer structured data.
• @ruvector/ruvllm-win32-x64-msvcRuvLLM native SIMD acceleration - win32-x64-msvc (Windows) platform
• @kenz1117/dsh-ui-usage-billingUsage billing dashboard for DeepSeek Harness: sidebar cost metrics plus a full dashboard modal, priced from a current multi-provider catalog with real usage aggregated from session logs.
• bitfabBitfab client for provider-based API calls with local BAML execution
• ag-mcpMCP server for AG-Grid to help developers integrate AG Grid into their applications
• @antv/gpt-vis-ssrSSR(Server Side Render) for AntV GPT-Vis.
• llama3-tokenizer-jsJS tokenizer for LLaMA 3
• avr-vadA Node.js library for Voice Activity Detection using Silero VAD
• meldMeld: A template language for LLM prompts
• typechatTypeChat is an experimental library that makes it easy to build natural language interfaces using types.
• @composio/claude-agent-sdkComposio provider for Claude Agent SDK
• @kimuson/modular-mcpModular MCP - Modular tool access for reducing context overhead
• @ruvector/wasmHigh-performance Rust vector database for browsers via WASM
• @roychri/mcp-server-asanaMCP Server for Asana
• ucode-agentucode - a terminal coding agent that reads, edits and runs your code, on Google Gemini models.
• @secretlint/secretlint-rule-anthropicA secretlint rule for detecting Anthropic Claude API keys
• docxodusRender, edit, diff and project Word .docx files in the browser — DOCX to HTML, a block editor, Word-grade redlines with tracked changes, and anchor-addressed markdown for LLM agents. WebAssembly, no server.
• @gotgenes/pi-anthropic-authPi extension package for Anthropic OAuth compatibility
• ssajuTiny, blazing-fast saju/manselyeok engine with LLM-ready output. Zero deps, gzip 15KB, 0.04ms per calc.
• @stdlib/ml-incr-binary-classificationIncrementally perform binary classification using stochastic gradient descent (SGD).
• @stdlib/ml-incr-sgd-regressionOnline regression via stochastic gradient descent (SGD).
• @workerdeck/protocolThe WorkerDeck wire protocol: typed session events, commands, and REST shapes shared by server and clients. Dependency-free, browser-safe. This protocol is the product boundary - versioned from day one.
• @decocms/parityE2E parity validator for site migrations. Compares prod vs cand and reports UI, functional, SEO, visual, and Web Vitals deltas with an LLM-ranked HTML report.
• @carderne/sandbox-runtimeAnthropic Sandbox Runtime (ASRT) - A general-purpose tool for wrapping security boundaries around arbitrary processes
• @vuetify/mcpModel Context Protocol server for Vuetify assistance
• ml5A friendly machine learning library for the web.
• byollmBring Your Own LLM — run an app's LLM jobs on your own models.
• @mdgate/iwork-commonmdgate shared Apple iWork (Pages/Numbers/Keynote) IWA parsing

No comparisons for this package yet.