ChuanhuChatGPT
GaiZhenbiao/ChuanhuChatGPTReviewed by AI
Checked by RepoPicks Radar on
What is ChuanhuChatGPT?
Chuanhu Chat (ChuanhuChatGPT) gives you one chat window for many AI models: paid services such as ChatGPT, Azure OpenAI, Gemini, Claude and DeepSeek through your own key, or models running on your machine through Ollama, ChatGLM, LLaMA and Qwen. It adds saved history with search, answers based on files you upload, web search, an automatic assistant and GPT-3.5 fine-tuning. You install Python, copy the example settings file, paste a key and start it; the page then opens in your browser. The README and the install guide are written mainly in Chinese.
How easy is it to install ChuanhuChatGPT?
C · install 1 or 2 tools · Windows
- Windows C
- macOS C
- Linux C
Every system is grade C and the steps are the same: install Python, clone the project, run "pip install -r requirements.txt", rename config_example.json to config.json, paste a key and run "python ChuanhuChatbot.py". There are ready starter scripts for the three systems (run_Windows.bat, run_macOS.command, run_Linux.sh), but no installer and no ready download in the 27 releases, and the detailed install guide is the project wiki, in Chinese.
ChuanhuChatGPT on GitHub
- Stars
- 15,244on Nov 10, 2026
- Downloads
- 4,0875 this week
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python
- To build from source
- Python, Docker, Graphics card
- Disk space
- 950 MB–4.3 GB · estimate
- Accounts or keys
- OpenAI (ChatGPT, DALL-E 3, fine-tuning), Google Gemini, Anthropic Claude, DeepSeek
- Cost
- Free, API costs extra · The program is free and GPL-3.0. Chatting through ChatGPT, Gemini, Claude or DeepSeek is billed by each of those services on your own key; using local models through Ollama costs nothing but needs a strong computer.
- License
- GPL-3.0 (Copyleft)
- Can replace
- ChatGPT Plus, TypingMind, Perplexity, NotebookLM
- Last change on GitHub
- Sep 16, 2026
ChuanhuChatGPT risk signals (8)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): gradio: GHSA-26jh-r8g2-6fpr, GHSA-279j-x4gx-hfrh, GHSA-37qc-qgx6-9xjv; langchain: GHSA-3644-q5cj-c5c7, GHSA-3hjh-jh2h-vrg6, GHSA-45pg-36p6-83v9; protobuf: GHSA-7gcm-g887-7qv7, GHSA-8qvm-5x2c-j2w7, PYSEC-2026-1805; python-multipart: GHSA-59g5-xgcq-4qw3, GHSA-5rvq-cxj2-64vf, GHSA-6jv3-5f52-599m
automatic check
- Attentionpaid api key needed for the main use
README quick start: after installing you must copy config_example.json to config.json and "fill in the API-Key and other settings". The model table lists ChatGPT, Azure OpenAI, Google Gemini, Claude, DeepSeek, iFlytek Spark and MiniMax, all services you pay per use. The program includes no model and no free quota; only the local-model route (Ollama, ChatGLM, LLaMA, Qwen) avoids that bill.
AI review
- Attentionpublic online demo is run by another account
The repository homepage points to "https://huggingface.co/spaces/JohnSmith9982/ChuanhuChatGPT" and the README links it as "try online". That page is hosted under the HuggingFace account JohnSmith9982, not under the repository owner GaiZhenbiao, so pasting your own API key there means handing it to a server you do not control. The README's own advice is to duplicate the Space into your account first.
AI review
- Attentionsettings can expose the chat to the internet
README, deployment section: set "\"server_name\": \"0.0.0.0\", \"server_port\": <your port>" to deploy to a server, and "\"share\": true" to "get a public link". Both make your chat window, and the keys saved in it, reachable from outside your computer; the README does not describe a password for that link.
AI review
- Attentiondependencies pinned to 2024 versions
requirements.txt pins gradio==4.29.0, gradio_client==0.16.1, openai==1.16.2, langchain==0.1.14, anthropic==0.18.1, pydantic==2.5.2, protobuf==3.20.3 and faiss-cpu==1.7.4. The automatic check already flags known vulnerabilities in gradio, langchain, protobuf and python-multipart; the pins also mean the installed libraries are from 2024 even though the README announces GPT-5 support.
AI review
- Infolocal models need a second install and a graphics card
The local-model libraries live in a separate requirements_advanced.txt (transformers, torch, llama-cpp-python, auto-gptq, cpm-kernels) and the Dockerfile leaves that line commented out: "# RUN pip install --user --no-cache-dir -r requirements_advanced.txt". Running ChatGLM, LLaMA or Qwen on your own machine means installing that second list and having a graphics card with enough memory.
AI review
- Infomidjourney listed though it has no public api
The model table lists "Midjourney" among the "API models" (marked as not supporting streaming). Midjourney sells access through its own app and does not publish a public API, so this connection depends on a route outside that service's terms; the repository does not say how it is made.
AI review
- Inforeadme and install guide mainly in chinese
The main README is in Simplified Chinese with translations in readme/ (English, Japanese, Korean, Russian), and the install and troubleshooting pages it sends you to are Chinese wiki pages ("使用教程", "常见问题"). The Telegram group is the support channel.
AI review
Evidence is shown in its original language.
What Radar could not check about ChuanhuChatGPT
- Whether each listed model really answers today, since the pinned libraries are from 2024 while the README announces GPT-5 and DeepSeek R1
- Real disk use: the checks estimate 950 to 4450 MB, and the local-model route downloads models on top of that
- What the online demo Space does with the API key you type into it
- How much each chat costs, which depends on the service and model you choose
- Whether the uploaded files used for file-based answers stay on your computer or are sent to the model service
- Whether the automatic assistant and web-search features still work, as they depend on outside search services