This postprocessor optimizes token usage by removing sentences that are not relevant to the query (this is done using embeddings).The percentile cutoff is a measure for using the top percentage of relevant sentences. The threshold cutoff can be specified instead, which uses a raw similarity cutoff for picking which sentences to keep.
Would want to:
test out to make sure it does as advertized
?is it worth adding if it adds more complexity to the manuscripot?
https://docs.llamaindex.ai/en/stable/examples/node_postprocessor/OptimizerDemo/
This postprocessor optimizes token usage by removing sentences that are not relevant to the query (this is done using embeddings).The percentile cutoff is a measure for using the top percentage of relevant sentences. The threshold cutoff can be specified instead, which uses a raw similarity cutoff for picking which sentences to keep.
Would want to: