Level 2 - API 101? What’s an embedding? How do you use APIs?

Code Sample

https://github.com/garethpaul/gpt-docs-api

Local Install

To get started you can run the following commands:

make build run

You should then be able to access streamlit in your browser with http://localhost:8502

For more information see the Makefile

Docker

If you'd prefer to use Docker to build and run.

docker build -t streamlit-app . docker run -p 8502:8502 streamlit-app

App Information

The app.py file provides a Streamlit application for generating responses to user queries using OpenAI's GPT-3 models. The application uses embeddings and metadata to create augmented queries for generating detailed and context-aware answers.

The application offers two modes of operation:

Use Embeddings: The application retrieves embeddings for the query, finds the top-k closest metadata entries, creates an augmented query, and uses OpenAI's ChatCompletion API to generate a response. The URLs for the top-k metadata are also displayed as sources.
Generic Response: If "Use Embeddings" is not selected, the application uses OpenAI's Completion API to generate a generic response to the query.

Functions in this module include:

load_embeddings_and_train_model(pickle_file_path): Load embeddings from a pickle file and train a nearest neighbors model.
get_embeddings(query): Retrieve embeddings for a query using OpenAI's Embedding API.
get_top_k_metadata(embedding, nn_model, metadata): Retrieve the top-k metadata entries corresponding to the nearest neighbors.
create_augmented_query(top_k_metadata, query): Construct an augmented query by combining top-k metadata contexts with the query.
get_model_response(augmented_query): Generate a response to the augmented query using OpenAI's ChatCompletion API.
get_generic_response(query): Generate a generic response to the query using OpenAI's Completion API.
main(): The main function for the Streamlit app, responsible for handling user input and generating responses.

Example usage: