can you use the huggingface datasets library with this?

Is there any way to filter out by retreived file name? For instance, I have a file Cuc113_Rolfson709_e74ac32b-102c-8a1d-cd6c-3b337cf8614d.json and a query that is looking for someone named Cuc113 Rolfson709, however the retriever occasionally pulls from the files of other patients. Can I somehow add the file metadata information and only pull from the files that matche the users name?

When using HuggingFaceEmbeddings to generate embeddings, how do I specify that I want the GPU to be used? Is there something like langchains model_kwargs?

I keep getting a ModuleNotFoundError: No module named 'llama_index.readers.base' even on code that used to run without issues before, was there an update to the llama-index repo that depreciated that???

as in, below x score results should be ignored

so for things like faithfulness and relevancy, the scores are generated by some outside LLM right? There's no formula that's used to generate that value?