tgpt is a cross-platform command-line interface (CLI) tool that allows you to use AI chatbot in your Terminal without requiring API keys.
- KoboldAI (koboldcpp/HF_SPACE_Tiefighter-13B)
- Phind (Phind Model)
- Llama2 (Llama 2 70b)
- Blackbox AI (Blackbox model)
- OpenAI (All models, Requires API Key)
Image Generation Model: Craiyon V3
Usage: tgpt [Flags] [Prompt]
Flags:
-s, --shell Generate and Execute shell commands. (Experimental)
-c, --code Generate Code. (Experimental)
-q, --quiet Gives response back without loading animation
-w, --whole Gives response back as a whole text
-img, --image Generate images from text
--provider Set Provider. Detailed information has been provided below (Env: AI_PROVIDER)
Some additional options can be set. However not all options are supported by all providers. Not supported options will just be ignored.
--model Set Model
--key Set API Key
--url Set OpenAI API endpoint url
--temperature Set temperature
--top_p Set top_p
--max_length Set max response length
Options:
-v, --version Print version
-h, --help Print help message
-i, --interactive Start normal interactive mode
-m, --multiline Start multi-line interactive mode
-cl, --changelog See changelog of versions
-u, --update Update program
Providers:
The default provider is phind. The AI_PROVIDER environment variable can be used to specify a different provider.
Available providers to use: openai, opengpts, koboldai, phind, llama2, blackboxai, ollama and groq
Provider: openai
Needs API key to work and supports various models. Recognizes the OPENAI_API_KEY and OPENAI_MODEL environment variables.
Provider: opengpts
Uses gpt-3.5-turbo only. Do not use with sensitive data
Provider: koboldai
Uses koboldcpp/HF_SPACE_Tiefighter-13B only, answers from novels
Provider: phind
Uses Phind Model. Great for developers
Provider: llama2
Llama 2 is an open source large language model (LLM) developed by Meta AI. Uses llama2-70b by default. Supports other models.
Provider: blackboxai
Uses BlackBox model. Great for developers
Provider: ollama
Needs to be run locally. Supports many models
Provider: groq
Requires a free API Key. Supports LLaMA2-70b & Mixtral-8x7b
Examples:
tgpt "What is internet?"
tgpt -m
tgpt -s "How to update my system?"
tgpt --provider opengpts "What is 1+1"
tgpt --provider openai --key "sk-xxxx" --model "gpt-3.5-turbo" "What is 1+1"
cat install.sh | tgpt "Explain the code"
The default download location is /usr/local/bin
, but you can change it in the command to use a different location. However, make sure the location is added to your PATH environment variable for easy accessibility.
You can download it with the following command:
curl -sSL https://raw.githubusercontent.com/aandrew-me/tgpt/main/install | bash -s /usr/local/bin
If you are using Arch Linux, you can install the AUR package with paru
:
paru -S tgpt-bin
Or with yay
:
yay -S tgpt-bin
You need to add the Go install directory to your system's shell path.
go install github.com/aandrew-me/tgpt/v2@latest
-
Scoop: Package installation with Scoop can be done using the following command:
scoop install https://raw.githubusercontent.com/aandrew-me/tgpt/main/tgpt.json
If you installed the program with the installation script, you may update it with
tgpt -u
It may require admin privileges.
Support:
- Http Proxy [
http://ip:port
] - Http Auth [
http://user:pass@ip:port
] - Socks5 Proxy [
socks5://ip:port ]
- Socks5 Auth [
socks5://user:pass@ip:port
]
If you want to use a proxy, create proxy.txt
file in the same directory from where you are executing the file and write your proxy configuration there.
Example:
http://127.0.0.1:8080
You can download the executable for your operating system, rename it to tgpt
(or any other desired name), and then execute it by typing ./tgpt
while in that directory. Alternatively, you can add it to your PATH environmental variable and then execute it by simply typing tgpt
.
If you installed with the install script, you can execute the following command to remove the tgpt executable
sudo rm $(which tgpt)
Configuration file is usually located in ~/.config/tgpt
on GNU/Linux Systems and in "Library/Application Support/tgpt"
on MacOS