Skip to content

Caching and resetting output

Extractor caches every LLM response so repeated calls with the same configuration reuse previous results.

Export the cache before you shut down and import it in a future session to avoid running the model on reports that have already been extracted:

extractor.export_cache("my_cache.pkl")
...
extractor.import_cache("my_cache.pkl")

If you want to start fresh, call reset() to clear cached feature values and token estimates. This is useful when you wish to re-run extract_features on the same DataFrame with a different feature model. reset returns the instance so you can immediately chain another call:

clean_df = extractor.reset().extract_features(feature_model=NewModel)

The returned DataFrame contains your newly extracted features and an empty cache ready for further runs.