llama_index
llama_index copied to clipboard
[wip] Example of implementing HyDE
https://github.com/jerryjliu/gpt_index/issues/385
Not sure exactly the best way to design this so that it fits the gpt_index abstractions, but this is the first step towards having a working HyDe configuration, and it functions quite well relative to the baseline embedding search
thanks for writing this! just a heads up i will be doing broader refactors to create some sort of "query processor" on index queries that can generically do operations such as hyde.
in the meantime i'm down to help land this as a beta feature if you'd like! (meaning params may change in the future). If you'r down, we could put in BaseGPTVectorStoreQuery or even just BaseGPTQuery (seems applicable to most indices)
Closed now that we have #437