OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Lime — xushengfeng

LIME is an open-source TypeScript application for Chinese Pinyin input that uses a Qwen3 language model to generate text candidates. A Rime frontend sends keystroke data to the LIME AI server, which returns candidates for the user to select.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Lime — xushengfeng — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use LIME when you want a Rime-based Pinyin input method whose candidate generation is guided by a language model, and can run its server and model in a trusted environment.

Who it is for

  • Rime users who want Qwen3-generated Chinese text candidates while entering Pinyin.
  • Developers interested in extending a Pinyin input method’s candidate-generation algorithms.

Documented uses

  • Generate Chinese text candidates guided by typed Pinyin syllables.
  • Use fuzzy-Pinyin mappings or supported Shuangpin schemes while entering text through Rime.
  • Extend or evaluate candidate-generation algorithms using the project’s shared model sessions, context, and token indexes.
Before you choose

Requirements & limits

  • The server requires Node.js (version 22 or later recommended), pnpm, and Qwen3 model files.
  • Rime is the documented frontend; LIME’s llm schema operates as a separate input method.
  • The Rime frontend enables HiAE by default for candidate and commit requests; the repository advises keeping the server on a trusted local or network environment because keystrokes may contain private information.
  • The project reports that long-sentence input may be less intelligent, fast typing may miss letters, and server restarts lose context because data and new words are not saved.
  • The repository lists the GPL-3.0 license.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up LIME with the Rime frontend

Prepare the LIME server and configure Rime to use its Pinyin input method.

Before you start

  • Node.js runtime; version 22 or later is recommended
  • pnpm
  • The LIME repository and Qwen3 model files
  • A Rime installation and curl

Steps

  1. Install the project dependencies with pnpm install.

  2. Download the Qwen3-0.6B-GGUF model files as described in the repository, and place the model folder alongside the project unless configuration is changed.

  3. Generate a key with pnpm key. In the Rime configuration, enable the llm schema and set the generated key in llm_pinyin.lua.

  4. For the default HiAE mode, run pnpm build:hiae and set hiae_payload in llm_pinyin.lua to the generated file’s absolute path.

  5. Start the server with pnpm start, then switch to the llm Pinyin input method in Rime.

Expected outcome

Rime can send Pinyin keystrokes to the LIME server for candidate generation and return selected text.

Guide limitations

  • The llm schema must be used as a separate Rime input method; it cannot be combined with other Rime input methods.
  • Keystrokes may contain private information. The project advises against exposing the server to the public internet or an untrusted local network.
  • The project notes that long-sentence input may be less intelligent and very fast typing may miss letters.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
5 November 2026
Original website title
xushengfeng/lime · Official homepage title · retained in its source language
How we review sourcesSuggest a correction