cGPTerm

This is a refactoring project of the original one xiaoxx970/chatgpt-in-terminal in C/C++.

cGPTerm enables chatting with ChatGPT in the terminal.

It supports for rendering markdown contents from ChatGPT's API responses.

Supports history retrieval with the up arrow key, stream output, and tokens counting.

Slash (/) commands are available in the chat box to toggle raw response mode, undo the last question and answer, modify the system prompt and more, see the available commands below for details.

It also supports saving chat messages to a JSON file and loading them from the file.

Uses the gpt-3.5-turbo model as default model, which is the same model used by ChatGPT (Free Edition).

You can also change API Host to configure proxy (See Available Arguments).

Preparation

An OpenAI API key. You need to register an OpenAI account and obtain an API key.

OpenAI's API key can be generated on the page opened by clicking "View API keys" in the upper right corner of the homepage, direct link: https://platform.openai.com/account/api-keys

Installation

You can download all EXE file and DLLs within a zip file in Release, but you can also compile this project by yourself (Please see Section Compilation).

Before or after downloading the release package, you should install all dependencies which are:

Library	Library name with APT	Installation Command with APT
curl	libcurl4	sudo apt install libcurl4
argtable	libargtable2-0	sudo apt install libargtable2-0
jansson	libjansson4	sudo apt install libjansson4
readline	readline-common	sudo apt install readline-common
x11	x11-common	sudo apt install x11-common
pcre2	libpcre2-8-0	sudo apt install libpcre2-8-0

You need to check your Package manager to see how to install these libraries.

Then, unzip the package, run ./cgpterm in the outcome folder. If all thing goes well, you should see cGPTerm start to work.

You can also add this folder to environment path, in order to run cGPTerm anywhere.

Compilation

---

Attention: All my development is done under Debian using gcc 10.2.1. And I've already found out that using another gcc version may result in a segmentation fault after ChatGPT responses. And also when compiling with the release preset of cmake, the program cannot accept SSE Event and I don't know why. Therefore I highly suggest to use the compiled package in releases before this problem is solved. My compile settings are:

set(CMAKE_C_FLAGS "-Wall -g")
set(CMAKE_CXX_FLAGS "-Wall -g -O2")

You may edit the CMakeLists by yourself.

---

You need GCC and CMake to compile. Make sure that your GCC supports C++20.

1. clone this repo with:

   git clone --recursive https://github.com/Ace-Radom/cGPTerm.git
   

2. install all dependencies' dev packages

   Library	Installation Command with APT
   curl	sudo apt install libcurl4-openssl-dev
   argtable	sudo apt install libargtable2-dev
   jansson	sudo apt install libjansson-dev
   readline	sudo apt install libreadline-dev
   x11	sudo apt install libx11-dev
   pcre2	sudo apt install libpcre2-dev

   Same, you need to check your own Package manager to see how to install these libraries, when you are not using apt.

3. make build directory with:

   mkdir build
   cd build

4. configure CMake with:

   cmake -DCLIP_EXAMPLES=OFF -DCLIP_TESTS=OFF ..
   

5. build

   make
   

After that, you should see all EXE files, DLLs and static libraries under $/bin. You can run ./cgpterm there and it should work.

First launch

cGPTerm will automatic create log and config files under ~/.cgpterm by the first launch.

You should configure your API Key with:

./cgpterm --set-apikey <YOUR_API_KEY>

Same as setting API Key, you don't need to open the config file (which is ~/.cgpterm/config.ini) and edit it by yourself. cGPTerm provides a couple of commands to set all configs in config file. See Available Arguments to learn more.

How To Use

Run with the following command:

cgpterm

Use Enter to submit questions.

Here are some common shortcut keys (also shortcut keys for the shell):

• Ctrl+L: Clear screen, equivalent to clear command in shell
• Ctrl+C: Stop the current chat request
• Ctrl+D: Exit cGPTerm / Abort settings
• Tab: Autocomplete commands or parameters
• Ctrl+U: Delete all characters to the left of the cursor
• Ctrl+K: Delete all characters to the right of the cursor
• Ctrl+W: Delete the word to the left of the cursor

Original chat logs will be saved to ~/.cgpterm/chat.log

cGPTerm Updater

This project also provides an updater to fetch the lastest release package from GitHub and install it automatically. You can use it to update your cGPTerm to latest version since v1.1.0.

Run with the following command:

cgpterm-update

This updater will first check if the local version is out of date. If it is, it will then try to download the latest release package from GitHub and install it.

You should put the updater in the same folder as other cGPTerm installed files. Otherwise the updater cannot find cGPTerm and will ask you if you want to install it (yes, it's also an installer).

The updater needs tar to unpack release package. Use which tar to see if it is installed.

If you are using a version prior to v1.1.0, you can first remove all local installed cGPTerm files (excluding configuration files) and install the new version using the updater.

Available Arguments

cGPTerm provides a couple of arguments to control the software or set configs.

Arguments	Description	Example
-h, --help	Show Help Messages and exit	cgpterm --help
-r, --raw	Enable raw mode	cgpterm --raw
--load FILE	Load chat history from file	cgpterm --load chat_history_code_check.json
--set-host HOST	Set API Host to use	cgpterm --set-host https://api.openai.com
--set-apikey KEY	Set API Key for OpenAI	cgpterm --set-apikey sk-....
--set-timeout TIMEOUT	Set maximum waiting time for API requests	cgpterm --set-timeout 40
--set-gentitle	Set whether to automatically generate a title for chat	cgpterm --set-gentitle
--set-saveperfix PERFIX	Set chat history file's save perfix	cgpterm --set-saveperfix chat_history_
--set-loglevel LEVEL	Set log level: DEBUG, INFO, ERROR, FATAL	cgpterm --set-loglevel DEBUG

Configuration File

The configuration file is located at ~/cgpterm/config.ini and is autogenerated. It can be modified using the program's --set option or edited manually.

The default configuration is as follows:

[DEFAULT]
# API Host, use for configuring proxy
# It must in format like "https://xxx.xxxxxx.xxx", no slash at the end is needed or allowed
OPENAI_HOST=https://api.openai.com

# API key for OpenAI
OPENAI_API_KEY=

# The maximum waiting time for API requests, the default is 30s
OPENAI_API_TIMEOUT=30

# Whether to automatically generate titles for conversations, enabled by default (generating titles will consume a small amount of tokens)
AUTO_GENERATE_TITLE=True

# Define the default file prefix when the /save command saves the chat history. The default value is "./chat_history_", which means that the chat history will be saved in the file starting with "chat_history_" in the current directory
# At the same time, the prefix can also be specified as a directory + / to allow the program to save the chat history in a folder (note that the corresponding folder needs to be created in advance), for example: CHAT_SAVE_PERFIX=chat_history/
CHAT_SAVE_PERFIX=./chat_history_

# Log level, default is INFO, available value: DEBUG, INFO, ERROR, FATAL
LOG_LEVEL=INFO

Available Slash Commands

• /raw: Switch response output style between raw and rendered Markdown format

  After switching, use the /last command to reprint the last reply

• /stream: Toggle stream output mode

  In stream mode, the answer will start outputting as soon as the first response arrives, which can reducing waiting time. Stream mode is on by default.

• /title: Toggle whether to enable automatic title generation

  When automatic title generation is on, cGPTerm will generate a new title after the first conversation of the current chat.

• /tokens: Display the total tokens spent and the tokens for the current conversation

  GPT-3.5 has a token limit of 4096; use this command to check if you're approaching the limit

• /usage: Show account credits summary

• /timeout [new_timeout]: Modify the API timeout

  The default timeout is 30 seconds, it can also be configured by setting OPENAI_API_TIMEOUT= in the config file, or using --set-timeout argument.

• /model [model_name]: Change AI model

  gpt-3.5-turbo, gpt-4 and gpt-4-32k are supported.

• /system [new_prompt]: Modify the system prompt

• /rand [randomness]: Set Model sampling randomness (temperature in OpenAI official docs)

  Default 1.

  Given randomness should be a real between 0 and 2.

• /save [filename_or_path]: Save the chat history to the specified JSON file

  If no filename or path is provided, the filename <SAVE PERFIX>YEAR-MONTH-DAY_HOUR,MINUTE,SECOND.json will be suggested on input.

• /undo: Delete the previous question and answer

• /delete or /delete first: Delete the first conversation in current chat

  When the token is about to reach the upper limit, the user will be warned.

• /delete all: Clear current chat, delete all questions and responses

• /last: Display last ChatGPT's reply

• /copy or /copy all: Copy the full ChatGPT's last reply (raw) to Clipboard

• /copy code: Copy the code in ChatGPT's last reply to Clipboard

• /version: Show cGPTerm local and remote version

• /list: List all settings in use

• /help: Show Slash Commands' help page

• /exit: Exit the application

You can press Ctrl_D to abort settings.

Exit Words

You can submit Exit Words to exit cGPTerm. Exit words will be sent as a question to ChatGPT, and cGPTerm will exit after GPT replies.

Exit words include:

['再见', 'bye', 'goodbye', '结束', 'end', '退出', 'exit', 'quit']

You can also use /exit command or Ctrl_D to exit directly.

Upon exit, the token count for the chat session will be showed.

Third Party Components

• dacap/clip: For accessing the clipboard.
• gh-markt/cpp-tiktoken: For tokens count when using stream mode.
• Ace-Radom/cpprich: For rich text output and Markdown render. (Rich Format library subproject under cGPTerm)
• ndevilla/iniparser: For parsering ini config file.
• jonhoo/pthread_pool: For thread-pool support.

You can find their licenses in THIRD_PARTY_LICENSE.

Sincere thanks to the developers of these projects.

Contributing

If you want, you can always dive in. Feel free to open an issue, submit PRs or report bugs!

License

This project is licensed under the GPLv3 License after v1.0.0 release.

From v0.1.0 to v0.3.1 is licensed under the Apache-2.0 License.