OpenAI Api Exploration with Embeddings

The goal of this is to create a program that acts as an NYC bar/happy hour recommender. It uses OpenAI's embeddings API to base the recommendations off a list of bars found in the csv files in the data directory.

Note: This README needs some TLC that's simply not available from me at 2am.

Running

1. Create a .env file in the repository (see .env.example) and add your OpenAI credentials.
2. Run yarn start "The prompt goes here".
3. Usually the first run will request and create the embeddings, but I'm commiting the embeddings to this repo for drinks-happy-hour.csv.

If you want to try out another data file you can create a new embedding context by basically copying the existing code in src/index.ts. Expect a 30-60 second first run of the program when doing this. Subsequent executions will be significantly faster at around 5 seconds.

Examples

I'm pretty damn impressed with how good it is off the bat. Pretty much no tweaking of anything, no error handling (lol), no logging, no prompt engineering.