Nit: in the context of this notebook, we're re-declaring an llm we already declared up above in ln 233. I'd probably omit this line, and omit the related import from langchain_openai import ChatOpenAI in ln 343 above.

I also don't love re-declaring the retriever with one additional param. It would be great if we could set enable_limit when we initially declare the retriever in ln 234, and then remove the re-initializing here.

It makes sense to have these things on a docs page if we want this to be a stand-alone code example, but here in the context of the notebook, it's not needed.

Good catch!

I got some query results that seem unrelated to the filter. i.e. for "toys", I got this document:

Document(id='685eaec1edc703d86a4c7201', metadata={'_id': '685eaec1edc703d86a4c7201', 'year': 1979, 'rating': 9.9, 'genre': 'science fiction'}, page_content='Three men walk into the Zone, three men walk out of the Zone')

For thriller and action, I got this document:

Document(id='685eaec1edc703d86a4c7203', metadata={'_id': '685eaec1edc703d86a4c7203', 'year': 1995, 'genre': 'animated', 'rating': 9.3}, page_content='Toys come alive and have a blast doing so')

I'm sure this is related to the limited amount of sample data we're providing, but it doesn't show the feature great to have these seemingly unrelated results being returned. I wonder if we want to add more sample data to show only obviously related results being retrieved?

Done! Added more data and improved the queries to produce so the outputs are more descriptive