#on-device

Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces

A local-Llm-first coding agent. Runs Qwen 2.5 and Qwen 3

realtime speech ⇄ text, 音を字に — wire mic → STT → translate → speech as an on-device voice graph that spans your devices over WebRTC. Runs in the browser (transformers.js/ONNX/WebGPU); no API keys, nothing leaves the device by default.

SigRank MCP — the yield cascade + live leaderboard as MCP tools any agent can call. On-device token scanner + tabbed TUI dashboard. npm i -g sigrank

A private reasoner for your terminal. Direct conversation or grounded multi-agent research, GPU-native and fully local. No API keys, no inference servers.