That can depend a lot on your use case, and often there is no clear answer as to what would work best other than testing different options. Playground may be helpful for you, another great thing is checking out the examples on github, and also chatGPT itself is super useful at helping with tuning recommendations. You can also feed documentation or even code from llamaindex into chatGPT to get its awesome assistance. Here is info on playground: https://github.com/jerryjliu/gpt_index/tree/6184d15456d0d387ba6336e79444bfdba721f6a5/examples/playground
thanks @afewell for the response! yeah cc @aioverdrivefriend the playground module should suited for generally testing the quality of responses across a bunch of indices
