0.1a8

• Now uses GPT-4o mini, which is around 30 times cheaper than GPT-4o for text tasks, though the same price for image tasks. #30
• New feature: on the "Extract data into this table" page there is now a link to "Duplicate these columns to a new table" which pre-fills the create table form with the same columns and hints. #29

0.1a7

• Now uses GPT-4o instead of GPT-4 Turbo - the new model is cheaper, faster and likely a little bit better too. #28

0.1a6

• Create table with AI extracted data page now has a "Add another column" button for if you need more than ten columns. #25

0.1a5

• Switch to using datasette-secrets for configuration. If you were previously using the environment variable OPENAI_API_KEY you should change that to DATASETTE_SECRETS_OPENAI_API_KEY. #24

0.1a4

• This plugin now uses the general availability gpt-4-turbo model, which adds structured function support to GPT-4 Vision. #19
• If an error occurs during an extraction operation it is now displayed on the progress page. #10
• Table and database action menu items now have better copy. #20

0.1a3

• Extraction jobs can now provide additional instructions to pass to the model, which are persisted and suggested for reuse when data is imported into the table in the future. #17

0.1a2

• Single drop zone for all types of upload - text files, images and HEIC images. #16

0.1a1

• Fixed bug where content without images failed to process. #14

0.1a0

• Initial plugin alpha release. #1