Langchain-Chatchat
chatchat-space/Langchain-ChatchatReviewed by AI
Checked by RepoPicks Radar on
What is Langchain-Chatchat?
Langchain-Chatchat builds a question-and-answer service over your own files. You drop documents into a knowledge base, it turns them into a searchable index, and the answers are written by a language model you choose. It can run entirely offline: the models are loaded by a separate program (Ollama, Xinference, LocalAI or FastChat), and online services such as OpenAI or Claude can be plugged in instead. It is installed with pip or Docker and used in a browser page on your own computer. The documentation is mainly in Chinese.
Langchain-Chatchat is free and open source (Apache-2.0) and can replace NotebookLM and ChatGPT.
How easy is it to install Langchain-Chatchat?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS D
- Linux D
There is no file to download in any of the 31 releases; the work is the same on Windows, macOS and Linux. You need Python, then `pip install langchain-chatchat`, then a second program to serve the models which the README asks you to install in a separate virtual environment, then `chatchat init`, hand edits in `model_settings.yaml`, `chatchat kb -r` to index the knowledge base and `chatchat start -a` to open it. Docker images exist as an alternative, which is still terminal work.
Coming soon: install in 1 click with the RepoPicks app.
Langchain-Chatchat on GitHub
- Stars
- 38,714on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +2stars in 7 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python
- To build from source
- Python
- Disk space
- 100 MB–300 MB · estimate
- Accounts or keys
- Online model services such as OpenAI, Azure OpenAI, Claude or Zhipu — only if you choose them instead of local models
- Cost
- Free · Apache-2.0 code, and free to run end to end with open models on your own machine. It can also be pointed at paid online services (OpenAI, Azure, Claude, Zhipu, Baichuan), which you then pay for.
- License
- Apache-2.0 (Permissive)
- Can replace
- NotebookLM, ChatGPT Plus
- Last change on GitHub
- Nov 10, 2025
Langchain-Chatchat risk signals (6)
- Attentionstale releases
Last release on 2024-07-12 (844 days ago)
automatic check
- Attentionweb page has no login if you open it up
The README warns that the default listening address `DEFAULT_BIND_HOST` is 127.0.0.1 "所以无法通过其他 ip 进行访问" and tells you to set it to 0.0.0.0 in `basic_settings.yaml` to reach it from another machine. Nothing in the content describes a password or user accounts, so anyone who can reach that address could read and change the knowledge base.
AI review
- Infoneeds an older python
pyproject.toml pins `python = ">=3.8.1,<3.12,!=3.9.7"`, so it does not install on Python 3.12 or newer, which is what current systems ship. The README names Python 3.8–3.11.
AI review
- Infono movement for about a year
The last commit is from 2025-11-10, 358 days before this check, and the README still says "🐳 Docker 镜像将会在近期更新" (the Docker image will be updated soon) next to an image tag dated 2024-08-29. The milestone list stops at June 2024.
AI review
- Infolicense stated twice and differently
GitHub and the LICENSE file give Apache-2.0, which the README repeats, but pyproject.toml declares `license = "MIT"`. Both are permissive, so the difference matters mostly to anyone reusing the code.
AI review
- Infodependencies fetched from a mirror by default
pyproject.toml sets a package source named "tsinghua" at `https://pypi.tuna.tsinghua.edu.cn/simple/` with `priority = "primary"`, so anyone installing from the source tree pulls the Python libraries from that mirror in China instead of PyPI.
AI review
Evidence is shown in its original language.
What Radar could not check about Langchain-Chatchat
- Whether the install still works today on a supported Python version, after a year without commits
- How well it answers in languages other than Chinese
- How much disk and memory the whole thing takes once the model framework and the models are added, since the figure here covers only the project's own files
- Whether the Docker images named in the README are still available
- What an online provider would cost if you use one instead of local models