todo.md

1. Index is not persisted. Because of Hugging face limitation.

Easy fix > Persist on Azure Storage (similar to s3) It's not that bad, it's using text-embedding-3-small which is $0.020 / 1M tokens. Fix in step_1_index_documents.py

2. Improve Python type in function

3. Add errors message and try catch

4. Add automated test

5. Steps are displayed live

6. Being able to test from any step (from step 3, avoiding to index and comparing, ...)

7. Add time it took between each step

8. Add price for each step

9. Improve which file is with who

Put the file names here, where in it

 text += f'### Discrepancy {i}:\n\n'