One desktop workspace for comparing AI conversations
Deutsch | English | Espanol | Francais | Italiano | Japanese | Korean | Russian | Tieng Viet | Chinese
AI Chat Hub is an Electron desktop workspace for sending a prompt to several AI models and comparing their results in one place. It keeps model selection, conversations, and follow-up work in a single local workspace.
- Compare responses in one to six model panels.
- Sign in to supported official web chats, including DeepSeek, Qianwen, Kimi, ChatGLM, and Doubao.
- Configure API models from OpenAI, Azure OpenAI, Anthropic, Google, Cohere, Groq, xAI, Baidu, and OpenAI-compatible endpoints.
- Improve prompts, attach text or code files as context, and generate images with a configured OpenAI API model.
- Produce a quick summary or analyze consensus and differences across selected responses.
- Save chats, prompts, and settings locally; manage multiple conversations, layouts, and proxy settings.
The following providers are available through the embedded official web workspace. The planned entries mirror the former project's web-access scope, but are not currently available.
| Provider | URL | Official web login | Notes |
|---|---|---|---|
| DeepSeek | https://chat.deepseek.com/ | Supported | Embedded official web workspace |
| Qianwen | https://chat.qwen.ai/ | Supported | Embedded official web workspace |
| Kimi | https://kimi.moonshot.cn/ | Supported | Embedded official web workspace |
| ChatGLM | https://chatglm.cn/ | Supported | Embedded official web workspace |
| Doubao | https://www.doubao.com/chat/ | Supported | Embedded official web workspace |
| 360 AI Brain | https://ai.360.cn/ | Planned | Former web-access scope |
| Character.AI | https://character.ai/ | Planned | Former web-access scope |
| ChatGPT | https://chatgpt.com/ | Planned | Former web-access scope |
| Claude | https://www.anthropic.com/claude | Planned | Former web-access scope |
| Code Llama | https://ai.meta.com/blog/code-llama-large-language-model-coding/ | Planned | Former web-access scope |
| Copilot | https://copilot.microsoft.com/ | Planned | Former web-access scope |
| Dedao Learning Assistant | https://ai.dedao.cn/ | Planned | Former planned web-access scope |
| Falcon 180B | https://huggingface.co/tiiuae/falcon-180b-chat | Planned | Former web-access scope |
| Gemini | https://gemini.google.com/ | Planned | Former web-access scope |
| Gemma 2B & 7B | https://blog.google/technology/developers/gemma-open-models/ | Planned | Former web-access scope |
| Gradio | https://gradio.app/ | Planned | Former web-access scope |
| HuggingChat | https://huggingface.co/chat/ | Planned | Former web-access scope |
| iFLYTEK SPARK | http://xinghuo.xfyun.cn/ | Planned | Former web-access scope |
| Llama 2 | https://ai.meta.com/llama/ | Planned | Former web-access scope |
| MOSS | https://moss.fastnlp.top/ | Planned | Former web-access scope |
| Perplexity | https://www.perplexity.ai/ | Planned | Former web-access scope |
| Phind | https://www.phind.com/ | Planned | Former web-access scope |
| Pi | https://pi.ai/ | Planned | Former web-access scope |
| Poe | https://poe.com/ | Planned | Former web-access scope |
| SkyWork | https://neice.tiangong.cn/ | Planned | Former web-access scope |
| Vicuna | https://lmsys.org/blog/2023-03-30-vicuna/ | Planned | Former web-access scope |
| WizardLM | https://github.com/nlpxucan/WizardLM | Planned | Former web-access scope |
| YouChat | https://you.com/ | Planned | Former web-access scope |
| You | https://you.com/ | Planned | Former web-access scope |
| Zephyr | https://huggingface.co/spaces/HuggingFaceH4/zephyr-chat | Planned | Former web-access scope |
- Launch
AI Chat Huband choose a model for each comparison panel. - For an official web model, sign in in its embedded login panel. For an API model, add its credentials in Settings.
- Enter a prompt and send it to the selected panels.
- Review the responses side by side, then summarize or analyze the results when needed.
You need a valid account or API key for the models you use, plus network access to the relevant service.
Chats, settings, and credentials are stored locally on your computer. API keys are not included when exporting chat data.
Install Node.js 20 and npm, then run these commands from the repository root:
npm install
npm run electron:serveKeep the terminal open while using the app. Use the Electron window that opens automatically; opening http://localhost:8080 in a browser does not support the embedded official chat pages.
- Download the Windows installer
.exe, or build it from source with the command below. Local builds are written todist_electron/. - Double-click the
.exeand complete the installation wizard. - Start the installed app from its desktop shortcut, Start menu entry, or installation directory.
Local builds are unsigned. Windows may show a SmartScreen warning; only choose More info > Run anyway when you trust the file's source.
Build a desktop package for the current platform:
npm run electron:buildReport bugs or feature requests in GitHub Issues.