matthewdeanmartin / cheaper_openai

Recreate ChatGPT-Assistant, but with a IDE, CLI driven UI and the cheaper text-davinci-003 model

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

cheaper_openai

Recreate ChatGPT-Assistant, but with an IDE, CLI driven UI and the cheaper text-davinci-003 model

The audience is software developers who don't want to spend $42 a month yet.

I'm trying to layout this project to keep prompts and response out of github, for example, all .md files are excluded.

Some people report a month of API usage as low as $2.

installation

Go add a credit card and get an API key. Put the API key into .env, copy the .env.example. Do not commit your key!

This workflow doesn't work well for pypi packages.

git clone
# set an output folder
nano config.toml

Put a current.md file with your prompt in your output folder.

Open in pycharm and run create.

usage

Open your output folder and the chat module in your favorite IDE. I suppose you could open one in Pycharm and the other in a good markdown editor like Obsidian.

Right now, it looks like the API has no concept of conversation. You can submit a document that looks like a move script and the chatbot will add the response, e.g.

Please write the next response. You are davinci. You are a crazy wild comedian.

Me: Tell me a joke about chickens and roads.

davicni:

To keep the conversation going, tack on what Davinci said. I plan to try to automate this.

I ask Davinci to give a conversation a title to avoid response_01.md, response_02.md etc. I plan to make this controllable via config.toml

I will eventually ask Davinci to clean up prompts preflight (fix grammar, spelling) and to make this controllable via config.toml

About

Recreate ChatGPT-Assistant, but with a IDE, CLI driven UI and the cheaper text-davinci-003 model

License:MIT License


Languages

Language:Python 99.4%Language:Dockerfile 0.5%Language:Shell 0.1%