The project is OpenAI-like API service of SkywardAI ecosystem.
setup-usage.mov
Introducing our new CLI tool!
Make sure you can normally run
make,docker compose,gcc,shin your host machine.
Simply run make setup in the root folder to compile & run the CLI tool.
Don't want to set? Directly go to Save & Build menu in it and use Build and start the server option to load the app in default settings.
No gcc compiler? You can choose to compile the file /setup/setup.c yourself.
Explore it yourself to find more settings!
- Please make sure you installed
Node.json your local machine. - This project developed on Node Version
v20.15.0.
# Manage package by pnpm
# Install pnpm globally, or change it to your local machine location
npm install -g pnpm
# Install dependencies
pnpm install
# OR
npm install
# RUN
npm run- Please make sure you have
dockerandmakeinstalled in your server. - Docker version for testing is
27.0.3, build 7d4bcd8.
# to simply start with all needed containers started, please run
make up
# if you just want to build this project to docker container, please run
make build
# if you want to start only this project in docker, please run
make start
# PLEASE NOTE: make start will automatically run make build first
# to run a container bind to your local machine volume, run
make dev
# this will do the same thing as `make up` but allows you to make changes and sync with containerNOTE: make dev Requires Node.js environment installed, or at least have node_modules specified in package.json installed on your server. Please see Local Machine section.
Go to the url of your project, default http://localhost:8000 if you didn't disabled the Docs route, then you can see docs and try it on.
See demo video.
This project got monitor build with swagger-stats, when you got this project running, just go to <Your Server>:<Your Port>/stats.
For example, http://localhost:8000/stats
When you set up the project and didn't disabled the
chatboxAPI, you can get a quick-setup chatbot with some basic styles on your own website, which calls the/v1/chat/completionsAPI for inference.
To set it up, simply add
<script src='http://localhost:8000/chatbox' defer></script>into the bottom of your html body element. So easy!
If you want to hide the real link, in your javascript code you can do
const chatbox_script = await (await fetch("http://localhost:8000/chatbox")).blob();
const chatbox_url = URL.createObjectURL(chatbox_script);
const script_elem = document.createElement('script');
script_elem.src = chatbox_url;
document.body.append(script_elem);And remember to use URL.revokeObjectURL(chatbox_url) if you don't need it anymore.
Extra parameters (request query) you can add to it are:
base_url:StringAdd this when in production, otherwise the requests won't send to correct route.
Defaulthttp://localhost:8000.max_tokens:IntegerAdd this when you want to limit tokens can be generated, is useful in production. Default
128