CoffeeShop demonstrates how to use JsonTranslator to create a natural language interface to take orders for a coffee shop. This requires a complex schema. CoffeeShop shows how GPT 3.5 and 4.0 can effectively be used to translate a broad range of user intent into strongly typed objects that can then be processed with conventional software.
CoffeeShop emulates natural language ordering at a Coffee Shop serving:
- Coffee
- Espresso
- Lattes
- Baked Good
A range of syrups and other familiar options are also offered.
CoffeeShop also demonstrates how to use the JsonVocab attribute to:
- Specify string vocabularies like:
[JsonVocab(whole milk | two percent milk | nonfat milk | soy milk | almond milk | oat milk)]- Vocabularies differ from enums because they can also be loaded on demand and/or be customized for each request (based on the request's context)
- Automatic validation and repair of strings using these vocabularies.
Works with gpt-35-turbo and gpt-4.
Example prompts can be found in input.txt.
Input:
we'd like a cappuccino with a pack of sugar, 1 tall latte with extra foam, 1 latte with vanilla syrup and 3 blueberry muffins.