What is it? pyWhat but with LLMs
Project description
LangWhat
Answer "What is it?" on the command line with the power of large language models (LLMs).
pyWhat LLM version, leveraging OpenAI API and Sydney (Bing AI Chat).
Screenshots
OpenAI
langwhat 'f7316ffccd4d2d555a7522328cf792dd73bfbcd9'
langwhat 'f7316ffccd4d2d555a7522328cf792dd73bfbcd9' --zh
Sydney
Sydney fixed my type "marry" automatically.
langwhat 'marry ball washington' -s
langwhat 'marry ball washington' -s -z
Caching
Responses are much faster when cache hits, and token usage becomes 0.
Note that Sydney doesn't support counting token usage, and always shows 0.
lw teddy --show-token-usage
lw teddy --show-token-usage
Features
- Uses few-shot prompting to reduce model mis-behavior
- English by default for superior response speed and accuracy
- Supports supplying API Key via either environment variable or config file at
~/.config/langwhat/api_key.txt
- Supports using Sydney with
-s
flag - Caching responses in a local sqlite database
Installation
pipx
This is the recommended installation method.
$ pipx install langwhat
# python 3.11 or higher is required, if your pipx uses a lower version of python by default,
# you could run the following command to install langwhat with python 3.11
# pipx install --python "$(which python3.11)"
pip
$ pip install langwhat
Usage
$ langwhat --help
usage: lw [-h] [-z] [-s] [-C] [--show-token-usage] [-V] what
positional arguments:
what what is it
options:
-h, --help show this help message and exit
-z, --zh Use Mandarin to prompt and answer
-s, --sydney Use Sydney (Bing AI) instead of OpenAI
-C, --no-cache Disable cache
--show-token-usage Show token usage
-V, --version show program's version number and exit
Develop
$ git clone https://github.com/tddschn/langwhat.git
$ cd langwhat
$ poetry install
Credits
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
langwhat-2.0.3.tar.gz
(8.9 kB
view hashes)