v0.0.5

Fixes the navigation button in the python UI for changes to Mesop.

v0.0.4

Automatically truncate text before sending to embedding model.

The embedding models have a maximum context length of 2048 tokens (see here). To avoid breaching this limit, we truncate the input to 2000 characters before calling the embedding models. This is conservative, but avoids an additional API call to check the tokens of the input, and in most cases the limit should not be breached.

v0.0.3

• Previously the Google Sheet being read by the UI required certain columns to be frozen. This is now no longer a requirement.
• The demo data used to have non numeric campaign ids. To make it clear what data is expected, these were changed to numeric.
• The requirements were updated to require pydantic 2.0 or greater

v0.0.2

Adds two new columns when generating ads in the UI:

1. The count of headlines
2. The count of descriptions.

This helps when evaluating if any generated ads have empty slots.

v0.0.1

First version, containing the python module and the user interface.