The infamous Question 8

Kokoro-82M q8 ONNX weights, split into 6 shards so a CDN will serve them

whisper-small ONNX models (transformers.js q8 format) for sts Lab, served via npmmirror

EmbeddingProvider の、外部サービスへ繋がない実装。ruri v3 30m (ONNX/q8) を transformers.js でプロセス内推論する。256次元・対称 prefix（docs/architecture.md §5.5）。

whisper-base ONNX models (transformers.js q8 format) for sts Lab, served via npmmirror

whisper-tiny ONNX models (transformers.js q8 format) for sts Lab, served via npmmirror

High-quality text-to-speech for the web

Batteries-included local multilingual MoE Text-to-Speech with acoustic classifier, audible calibration testing, router-in-the-weights, and calibrated LoRA personas for Pi

Kuwait Prayer Times

High-quality text-to-speech for the web, support Chinese, web播放文本到语音，支持中文语音.

A local-first AI plugin for Joplin that clusters notes semantically, suggests tags and notebook structures.

Wiki knowledge base CLI — terminal search and management

Local AI inference in the browser: embeddings, classification, reranking, semantic search and chat, running in a Web Worker.

zgentic terminal client and local tool bridge (#174). Bun is NOT required to run the published package: the CLI ships as a single pre-bundled Node script (built with Bun at publish time), so plain `node`/`npx` works. Bun is only needed to develop or build

Local, inspectable memory and experience evidence for MCP agents

Private/offline browser-side document chunking and embedding for Knowledge Base using bundled ONNX models via @huggingface/transformers.

Context-aware voice input for DeepSeek Harness with Web Speech, local SenseVoice transcription, model polish, editable Composer drafts, and user-controlled sending

Semantic search over past OpenCode conversations — local embeddings (Transformers.js), SQLite index, native plugin tools. Inspired by obra/episodic-memory, rebuilt natively for OpenCode.

The Unified Model Registry for your local AI apps

High-quality text-to-speech for the web