Simple semantic search for all your tweets

1. request your personal data archive. Follow the directions here
2. There will be a file called data/tweets.js. Mine had a single variable assigned to an array of tweet objects. Edit it, leave only the array and rename it to tweets.json.
3. pip install -r requirements.txt
4. run python embed-tweets-instructor.py to create embeddings for all your tweets using the instructorXL model. This model is large, and it will take a while to download. I tested this on a GPU with 8GB of VRAM, not sure if it works with something smaller. You can easily change this code to use smaller models.
5. run python search-instructor.py to run semantic queries.