AingDesk is an AI assistant that supports local models, knowledge bases, sharing, web search, and intelligent agents. It provides one-click deployment, server-side deployment options, and MCP client integration, with Docker and build steps available.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
AingDesk is an easy-to-use AI assistant that supports knowledge bases, model APIs, sharing, internet/search, and intelligent agents. It targets local AI model deployment, knowledge base setup, and server-side deployment.
How it works
The project enables one-click deployment of local AI models and mainstream model APIs, supports a local knowledge base, and includes an intelligent agent creation workflow. It also offers sharing capabilities and web search integration, with an MCP client and multi-model conversations in a session (coming soon).
Getting started
Quick start options
- Client Version (MacOS, Windows) install: Download from official website or mirrors (links in README).
- Server Version: Docker Run example provided, with volume mappings and port 7071.
- Server Version: Docker Compose example provided to run with docker compose up -d.
Build locally
git clone https://github.com/aingdesk/AingDesk.git
cd AingDesk
# For macOS users, please remove the `@rollup/rollup-win32-x64-msvc` dependency in [package.json](http://_vscodecontentref_/0)
cd frontend
yarn
cd ..
yarn
yarn dev
Recent releases
- v1.2.4 (2025-05-27): Add support for the Qwen3 model; optimize result parsing for certain search engines; add OCR support for pure image PDFs (macOS requires manual installation of Poppler); improve the effect of usi.
- v1.2.3 (2025-04-18): Fix compatibility with OpenAI GPT-O3; resolve issue with third-party embedding model creating a knowledge base with a missing model; optimize the model selector.
- v1.2.2 (2025-04-16): Add automatic update mechanism (Windows only); fix issues with combined invocation of knowledge base, agent, and tools; resolve other known issues.
- v1.2.1 (2025-04-14): Add MCP client functionality (configure MCP server in settings, then select the tool in the conversation); optimize model selector; miscellaneous adjustments.
- v1.2.0 (2025-04-08): Optimized default document chunking; improved full-text search accuracy in knowledge base; fixed incomplete chunking with multiple delimiters.
Traction
- Stars: 2529
Behind the repo
- Repository owner: aingdesk
- License: MIT
- Language: TypeScript
- Created: 2025-02-12
- Last push: 2026-06-04
- Topics: deepseek, electron, llm, localai, nodejs, ollama
Caveats
- License: MIT
- Age: Created 2025-02-12; last push 2026-06-04
- Open issues: 70
- Hosted via Docker for server deployment; client downloads available from project links