< 10MB, One-click self-hosted ChatGPT, allowing access to various data sources and non-OpenAI models.
WTFPL License
One-click local version of ChatGPT, allowing access to various data sources and non-OpenAI models.
Check the example directory, choose the right configuration and use.
# download the latest version
git clone https://github.com/soulteary/docker-chatgpt.git
# or use zipball
wget https://github.com/soulteary/docker-chatgpt/archive/refs/heads/main.zip
OPENAI_API_KEY: "sk-......"
# If your network is not directly accessible OpenAI API, enable proxy and set you own proxy server addr.
# OPENAI_API_PROXY_ENABLE: "on"
# OPENAI_API_PROXY_ADDR: "http://127.0.0.1:1234"
docker compose up
# or run in the daemon mode
docker compose up -d
Open your browser, visit http://localhost:8090
, and enjoy.
The Client will be updated along with the project to keep it consistent with the officially supported functions.
You can update the project by updating the mirror version used in this repository.
# x86_64
docker pull soulteary/chatgpt
# Mac M1/M2
docker pull soulteary/docker-chatgpt:arm64
Backend services will continue to complete and support new data source types.
You can download the latest version of the automatically built image by using the following command:
docker pull soulteary/sparrow
# or use the latest version
docker pull soulteary/sparrow:v0.9.2
Then, restart the project with docker compose down && docker compose up -d
to complete the upgrade.
For more advanced usage, and previous practices, such as searching various vertical websites through it, using MidJoruney to draw pictures, you can refer to the video in the Sparrow project documentation.
Welcome to submit your code in the project to support your data type.
The reason for the good experience of the project is because it is very fast and has a very good score.
In addition, the daily operation of the program as a service only requires about 10MB of hard disk space, about 10MB of content, and about 1% of CPU resources.
The project does not need to connect to any external network except for the backend service address that will be connected in the configuration.
You can prohibit the privacy leakage you are worried about by setting firewall rules or cloud server export access rules.
This does not affect the use of the program as it does not require an additional network connection.