especially for a workflow of input_file >> preprocessing >> llm extraction it is very useful to be able to go from the file input to the result in your eval experiments, instead of only from the preprocessed content to the result.
for example when I extract information from a pdf I would like to experiment both with different ocr providers and settings, and with different models and prompts. For that I need access to the original pdf not just the ocr extracted text in chat input format.
Feature request
although it is possible to upload files and view them in the dashboard, when i download the run they do not show up:
Motivation
especially for a workflow of input_file >> preprocessing >> llm extraction it is very useful to be able to go from the file input to the result in your eval experiments, instead of only from the preprocessed content to the result.
for example when I extract information from a pdf I would like to experiment both with different ocr providers and settings, and with different models and prompts. For that I need access to the original pdf not just the ocr extracted text in chat input format.