Liuying is a local-first AI knowledge base for Windows. Your notes, documents and chat in one place — with an AI that runs on your own hardware and doesn't need the internet.
Not a screenshot, just the point: everything above keeps working with the cable unplugged.
Liuying is in open beta, and every feature is free: the built-in local AI, search over your own archive, bring-your-own-key cloud models, packs and plugins. No seat caps, no card, no expiry date on your download.
A paid v1.0 will come later — pricing and timing aren't decided yet, and we'll say it plainly here when they are. What we ask of beta users is simpler: point it at your real archive, hit the feedback button when something breaks, and help us figure out what v1.0 should be.
Open beta terms, October 2026. When v1.0 pricing exists it will be published plainly — no countdown timers, and nothing you use today gets locked in the meantime.
One Windows app. Your files stay your files.
Point Liuying at your notes and documents. Ask in plain language; get answers that quote the source file, so you can check every claim against the original. The index lives on your disk, not in our cloud — we don't have one.
A local AI model ships with the app — it's fetched on first launch, and from then on it runs on your own hardware, offline. A regular Windows PC with 16 GB of RAM is enough; no dedicated GPU, no account, no upload, no round trip. Pull the network cable and it all still runs.
Want a frontier model for a hard question? Paste your own API key — OpenAI, Anthropic, Google Gemini, or a custom endpoint — and use the cloud only for the conversations you choose. Keys are encrypted at rest, and the local model remains the default.
Share curated knowledge as encrypted .lypk packs: readers can search the content but cannot export the source. A plugin system extends the rest.
Hit the feedback button and your note goes straight to the team with an optional diagnostic bundle — sent only when you choose to send it. No community to join, no email address to hand over.
Liuying is local-first: your notes, your documents and the AI index all live on your own disk — there is no server of ours for them to live on. There's no account to create and no telemetry running in the background; nothing counts your keystrokes or phones home. By default, none of your data leaves your machine. The only things that ever leave are the ones you explicitly send: feedback, when you press the button and choose to send it, and the conversations you deliberately route to an optional cloud model — your API key stays on your machine, encrypted at rest.
Everything local keeps working: indexing, search, chat with the local model, reading and editing notes. Only the optional bring-your-own-key cloud models need a connection — and only in the conversations where you picked one.
On your machine, encrypted at rest. They are never printed to logs, never shown back in full — the settings page only ever displays a masked version like sk-****c123.
Yes. The local side runs models through its bundled Ollama engine, and the cloud side is bring-your-own-key with presets for OpenAI, Anthropic and Google Gemini — or plug in a custom API endpoint.
A Windows PC with 16 GB of RAM. No dedicated graphics card required — the built-in local model runs on ordinary hardware, and everything keeps working offline.
During the open beta, every feature is free and nothing is locked. A paid v1.0 will come later — we haven't set a price or a date, and when we do it will be announced plainly. Your beta download doesn't expire in the meantime.