text-to-regex

Uses GPT-3's latest text-davanci-003 model via LLMHub to convert text to Python re expressions.

There are two main features:

• explicit "assumptions". This lets you find errors, and iterate towards getting a final regex that "works".
• quickly test the regex on the same page and iterate towards

Running your own server

# Setup Python environment.
conda create -n t2r python
conda activate t2r
pip install -r requirements.txt

# Setup LLMHub dependency.
npm install -g llmhub
llmhub auth

# Start Streamlit app.
streamlit run app.py

TODO

1. Add ability to gather feedback from users (upvote/downvote).
2. People usually think of a string they want to match and then come up with a regex. So, we should change prompt so that it includes the string people are trying to again match against whilst coming up with the regex.