NodeSchool AI workshop

Getting started

1. Install Node.js (v18 recommended, check by running node -v in your terminal)
2. Install a code editor
   • We recommend Visual Studio Code
3. Clone or download this project
   • Download: https://github.com/pixelbar/nodeschool-ai/archive/refs/heads/main.zip
   • Clone: git clone https://github.com/pixelbar/nodeschool-ai.git
4. Open the cloned directory in your code editor
5. Install dependencies
   • Run in cloned directory: npm install
6. Download models
   • If you're at Pixelbar, go to: http://172.16.31.216/ and download t5-base.zip and vit-base-patch16-224.zip, unzip them, and put them (the whole folder, not just the contents) in the models/Xenova folder.
   • If you're not at Pixelbar, set allowRemoteModels in app.js to true to allow models to be downloaded at runtime.
7. If you want to use OpenAI, create an account at https://openai.com. You get $5 of free credits to be used in the first 3 months.
   • Create an API key: https://platform.openai.com/account/api-keys
   • Copy .env.example to .env and fill in the values
8. Run the app: node app.js

Models

Note that Transformers.js is limited in that it can only run models that have been converted to a special format (ONNX). You can find all models compatible with Transformers.js here: https://huggingface.co/models?library=transformers.js&sort=downloads.

Resources

• Express documentation - Express is the framework that the application is built on
• Transformers.js documentation - Transformers.js is a library to run machine learning models from Hugging Face on Node.js and the web
• Hugging Face Tasks - explore ML models for various specific tasks
• OpenAI API reference
• GPT best practices - best practices for prompting OpenAI GPT APIs
• OpenAI chat playground - playground to interact with Chat Completion API

Resources for after the workshop

• ChatGPT prompt engineering course