openchat-llm

Lightweight OpenAI-compatible chat client based on NextChat.

This is a personal experimental fork for validating a compliant MIT open-source modification and release workflow. It does not include any bundled API key, company endpoint, free quota, or payment flow.

中文说明

中文简介

openchat-llm 是一个基于 NextChat 的轻量实验性开源聊天客户端，主要用于验证 MIT 开源项目复制、轻量修改、依赖检测、密钥检测、Docker 构建和 GitHub 发布的完整合规流程。

本项目保留 OpenAI-compatible API 的配置入口，用户需要自行配置 API Key、Base URL、模型列表和访问码。项目不内置任何 API Key，不绑定公司中转站，不提供免费额度或付费承诺，也不表示获得上游 NextChat 官方背书。

Features

• OpenAI-compatible API configuration
• Custom API key and base URL
• Custom model list and default model
• Local-first chat history
• Markdown, LaTeX, Mermaid, code highlighting
• Prompt templates, masks, plugins, artifacts, and MCP support inherited from upstream
• Responsive web UI, PWA, dark mode, Docker, Vercel, and Windows desktop preview deployment

What Changed From Upstream

• Replaced user-facing product name with openchat-llm.
• Removed upstream SaaS, iOS, sponsor, and promotional sections from this README.
• Replaced visible NextChat SaaS promotion copy in the settings UI with neutral API configuration copy.
• Preserved OpenAI-compatible API settings and user-controlled key/base URL behavior.
• Preserved the original MIT license and copyright notice.

This project is not affiliated with, endorsed by, or sponsored by the original NextChat maintainers.

Quick Start

Create .env.local in the project root:

OPENAI_API_KEY=
BASE_URL=
CODE=
CUSTOM_MODELS=
DEFAULT_MODEL=
HIDE_USER_API_KEY=0
ENABLE_BALANCE_QUERY=0

Install dependencies and start development:

yarn install
yarn dev

Open http://localhost:3000.

Environment Variables

Name	Required	Description
OPENAI_API_KEY	Yes for server-side use	Your OpenAI-compatible API key. Multiple keys can be separated by commas.
BASE_URL	No	OpenAI-compatible API base URL. Default upstream behavior uses https://api.openai.com.
CODE	No	Access password, separated by commas.
CUSTOM_MODELS	No	Add, hide, or rename models. Example: -all,+gpt-4o-mini,+deepseek-chat.
DEFAULT_MODEL	No	Default model shown to users.
HIDE_USER_API_KEY	No	Set to 1 to hide user API key input. Default keeps user configuration available.
ENABLE_BALANCE_QUERY	No	Set to 1 to enable balance query if your endpoint supports it.
DISABLE_FAST_LINK	No	Set to 1 to disable importing settings from URL.
ENABLE_MCP	No	Set to true to enable MCP support before build.

No real API key should be committed to this repository.

Deploy

Windows Desktop Preview

openchat-llm v0.2.0 includes an experimental Windows desktop preview built with Tauri.

Download the Windows installer from the GitHub Release page, install it, then configure your own OpenAI-compatible API key and base URL in Settings. The desktop preview does not include any bundled API key, company endpoint, free quota, or payment flow.

This preview build is not code-signed. Windows SmartScreen may show a warning when installing or launching it. The MSI installs to C:\Program Files\openchat-llm and may require administrator approval on Windows. Verify the SHA256 checksum published in the release notes before installing.

Build locally:

yarn install
yarn export
yarn app:build

The installer artifacts are generated under:

src-tauri/target/release/bundle/

Docker

docker build -t openchat-llm .

docker run -d -p 3000:3000 \
  -e OPENAI_API_KEY=your_api_key \
  -e BASE_URL=https://api.openai.com \
  -e CODE=your_access_code \
  openchat-llm

Vercel

Deploy the repository to Vercel and configure environment variables in the Vercel dashboard:

OPENAI_API_KEY
BASE_URL
CODE
CUSTOM_MODELS
DEFAULT_MODEL

Local Development

yarn install
yarn dev

Build production output:

yarn build

Run tests:

yarn test:ci

Compliance

This project is based on NextChat, which is licensed under the MIT License.

Original copyright:

Copyright (c) 2023-2025 NextChat

The original MIT license text is preserved in LICENSE. Additional attribution and modification notes are documented in NOTICE.md.

Before publishing a release, run:

• dependency license scan
• secret scan
• local build
• README and NOTICE review

Security

Do not commit API keys, access tokens, passwords, or internal service credentials. See SECURITY.md.

Upstream

• Original project: ChatGPTNextWeb/NextChat
• License: MIT