feat: replace openai with llmhub

[vatsalaggarwal]

Summary

This PR integrates LLMHub. This integration allows users to get started right away (1-click!), and automatically integrates all relevant LLM providers without requiring users to set up API keys with each. Additionally we provide various tools to make it easier to improve its accuracy/latency/costs (via tools for finetuning & prompt engineering).

A user would only need to do the following to get started:

pip3 install llmhub
# llmhub was globally installed as part of previous command.
llmhub auth

llmhub auth points users to a webpage where they need to sign in via GitHub. That's it!

Features

💆 Allow users to use your app in 1-click.

💻 Deploy in 1-click.

💇 Integrate ANY LLM in 3-lines with a standardised API.

🔧 Monitor and improve your prompts.

🌎 Discover new prompts easily, and showcase your work.

💪 Contribute to popular prompts, and improve your skills.

💵 Free credits while we're in beta

Some things to note

• The current prompts are deployed to be an exact replica of existing code (models & parameters like temperature/frequency_penalty). Easy to edit.
• We'd love to keep this LLM work public/open-source. Is that OK?

About Us

We're @vatsalaggarwal and @sidroopdaska. @sidroopdaska was the second engineering hire at the largest autonomous driving company in Europe, and I was an ML Scientist at Amazon for 4 years and researched deep generative models!

[tom-doerr]

Hi, thank you for adding the changes. However I thought it would be optional to use LLMHub, the current version requires it. Could you edit this pull request and the one for tom-doerr/fix#3 to make it optional? Otherwise it would break existing installations and make it incompatible with my other Codex projects that read the API keys from the same location.