A simple chat client in single Python script

stream

1. clone this repo
2. pip3 install -U -r requirements.txt
3. copy demo_config.json to conf/config.json
4. get your OPENAI_API_KEY and put it in conf/config.json

$ ./gptcli.py -h
usage: gptcli.py [-h] [-c CONFIG]
options:
 -h, --help show this help message and exit
 -c CONFIG path to your config.json (default: config.json)

Sample config.json:

{
 "api_key": "sk-xxx",
 "base_url": "https://chat.pppan.net/v1",
 "model": "gemini-2.5-pro-exp-03-25",
 "context": 2,
 "stream": true,
 "stream_render": true,
 "showtokens": false,
 "proxy": "socks5://localhost:1080",
 "prompt": [
 { "role": "system", "content": "You are a helpful assistant" }
 ],
 "model_choices": [
 "gemini-2.5-pro-exp-03-25",
 "gemini-2.0-flash",
 "gemini-2.0-flash-lite",
 "gemini-1.5-flash",
 "gemini-1.5-flash-8b",
 "gemini-1.5-pro"
 ]
}

• (required) api_key: OpenAI's api key. will read from OPENAI_API_KEY envronment variable if not set
• (optional) base_url: OpenAI's api base url. Can set to a server reverse proxy, for example Azure OpenAI Service or chatgptProxyAPI. By default it's from OPENAI_API_BASE or just https://api.openai.com/v1;
• (optional) model: LLM chat model, by default it's gpt-3.5-turbo;
• (optional) context: Chat session context, choices are:
  • 0: no context provided for every chat request, cost least tokens, but AI don't kown what you said before;
  • 1: only use previous user questions as context;
  • 2: use both previous questions and answers as context, would cost more tokens;
• (optional) stream: Output in stream mode;
• (optional) stream_render: Render markdown in stream mode, you can disable it to avoid some UI bugs;
• (optional) showtokens: Print used tokens after every chat;
• (optional) proxy: Use http/https/socks4a/socks5 proxy for requests to api_base;
• (optional) prompt: Customize your prompt. This will appear in every chat request;
• (optional) model_choices: List of available models;

Console help (with tab-complete):

gptcli> .help -v
gptcli commands (use '.help -v' for verbose/'.help <topic>' for details):
======================================================================================================
.edit Run a text editor and optionally open a file with it
.help List available commands or provide detailed help for a specific command
.load Load conversation from Markdown/JSON file
.multiline input multiple lines, end with ctrl-d(Linux/macOS) or ctrl-z(Windows). Cancel
 with ctrl-c
.prompt Load different prompts
.quit Exit this application
.reset Reset session, i.e. clear chat history
.save Save current conversation to Markdown/JSON file
.set Set a settable parameter or show current settings of parameters
.usage Tokens usage of current session / last N days, or print detail billing info

Run in Docker:

# build
$ docker build -t gptcli:latest .
# run
$ docker run -it --rm -v $PWD/.key:/gptcli/.key gptcli:latest -h
# for host proxy access:
$ docker run --rm -it -v $PWD/config.json:/gptcli/config.json --network host gptcli:latest -c /gptcli/config.json

• Single Python script
• Support any OpenAI-Compatible API and models
• Session based
• Markdown support with code syntax highlight
• Stream output support
• Proxy support (HTTP/HTTPS/SOCKS4A/SOCKS5)
• Multiline input support (via .multiline command)
• Save and load session from file (Markdown/JSON) (via .save and .load command)
• Print tokens usage in realtime, and tokens usage for last N days, and billing details (only works for OpenAI)

This script only support text models. If you want a more feature-rich client, for example, with functions like RAG, image generation, Function Calling, etc., please consult other projects, for instance, aichat.

• https://platform.openai.com/docs/introduction
• https://platform.openai.com/docs/api-reference/completions
• https://platform.openai.com/docs/models/overview
• https://platform.openai.com/account/api-keys