Chaty v0.1.0 — first stable release. A local, private desktop chat app for GGUF models. Load a .gguf and talk to it; everything runs on your machine.
Highlights
- Local GGUF inference (llama.cpp) with fast multi‑turn KV‑cache reuse
- Model hot‑swap from a
models/folder; auto‑loads the last model on launch - Polished chat UI — streaming, foldable
<think>panel + thinking‑mode toggle, KaTeX math, GFM tables, per‑block code copy, and in‑app HTML preview - 🎙️ CPU voice (English) — voice input with automatic silence‑stop, streaming read‑aloud, and a Gemini‑style live mode with an animated orb (runs on CPU, independent of the LLM's VRAM)
- Web search & URL fetch, file/PDF attachments, Latin OCR
- Native shell — system tray, global hotkey (
Ctrl+Shift+Space), custom title bar & right‑click menu, single‑instance - SQLite history with branching; English / 简体中文 UI
Install (Windows x64)
Download Chaty_0.1.0_x64-setup.exe below and run it — installs per‑user (no admin). The voice models (~0.5 GB) download automatically the first time you use a voice feature.
Voice is English‑only (Whisper‑base.en + Kokoro‑82M) and runs on CPU. Build from source: see BUILD.md.