AI Client

An elegant AI client

Acknowledgment: This project is based on NitroRCr/AIaW. Thanks to the original author for the open-source contribution!

Features

Cross-Platform Support

• Multi-platform: Windows, Linux, macOS, Android, Web (PWA)
• Multi-provider: OpenAI, Anthropic, Google, DeepSeek, xAI, Azure, Ollama, etc.
• Custom providers: Support any OpenAI-compatible API

Conversations & Workspaces

• Multiple workspaces: Separate conversations by topics, supports folder nesting
• Assistant system: Workspace assistants and global assistants
• Branching dialogs: Edits and regenerations presented as branches
• Pin conversations & workspaces: Pin important conversations and workspaces to top
• Multi-select delete: Select and batch delete multiple conversations
• Message quoting: Quote previous messages for follow-up questions

Knowledge Base (RAG)

• Multiple knowledge bases: Create and manage multiple knowledge bases
• Multi-format import: TXT, Markdown, PDF, Word, Excel, etc.
• Smart retrieval: Document chunking, vector storage, similarity search
• Dialog integration: Link knowledge bases for retrieval-augmented generation

MCP Protocol

• Full support: Tools, Prompts, Resources
• Multiple transports: STDIO (desktop), HTTP, SSE
• MCP Proxy: Web version bypasses CORS via proxy
• Plugin marketplace: Install MCP plugins from marketplace

Code & Artifacts

• Code execution: One-click JavaScript/TypeScript execution, live HTML preview
• Artifacts: Convert responses to editable Artifacts with version control
• Syntax highlighting: Multi-language code highlighting

Plugin System

• Built-in plugins: Calculator, document parsing, video parsing, image generation
• Plugin marketplace: Install more community plugins
• Gradio support: Configure Gradio apps as plugins
• LobeChat compatible: Compatible with some LobeChat plugins

Web Search

• SearXNG integration: Out-of-the-box web search
• Web scraping: Fetch web content via URL
• Concurrent processing: Support concurrent search and scraping

Data & Sync

• Local-first: Data stored locally, works offline, fast startup
• Cloud sync: Cross-device real-time sync after login
• Multi-window: Same browser multi-tab data synchronization

Other Features

• Message translation: Streaming translation, Markdown rendering
• Model capabilities display: Visual indicators for vision, reasoning, web search, and other model capabilities
• Dynamic prompts: Template syntax, variable reuse
• Dark mode: Dark/Light/Auto theme switching
• Custom theme: Customizable theme colors
• Keyboard shortcuts: Customizable shortcuts

Related Projects

• New API: AI model interface management and distribution system, supporting various large models with OpenAI-compatible format

Deployment

Docker Deployment

# Pull the image
docker pull ghcr.io/nsmao-com/ai-client:latest

# Run
docker run -d -p 80:80 ghcr.io/nsmao-com/ai-client:latest

MCP Proxy (for web version)

The web version requires MCP Proxy to bypass CORS restrictions when connecting to MCP servers:

# Pull and run MCP Proxy
docker pull ghcr.io/nsmao-com/ai-client/mcp-proxy:latest
docker run -d -p 8080:8080 ghcr.io/nsmao-com/ai-client/mcp-proxy:latest

See mcp-proxy/README.md for more details.

Development

Install the dependencies

pnpm i

Start the app in development mode (hot-code reloading, error reporting, etc.)

quasar dev

Lint the files

pnpm lint

Build the app for production

# SPA (Static Web App)
quasar build

# PWA (Progressive Web App)
quasar build -m pwa

# Electron (Desktop App for Windows/Linux/macOS)
quasar build -m electron