i have made simplevectorindex from 3 docs adding up to 1000 pages. now querying it is taking 1minute sometimes. what is the best way to reduce this time
I have almost 7000 articles indexed. Though I was using simplevectorindex as well at first but my queries took mostly 20-30 seconds. Now it takes 10-15 with pinecone index.
I noticed that the most relevant factors for reducing time are: top k similarities chosen, how you build the prompt, eg asking short answers i can get answers with 5 top similarities in 10 seconds. Less relevant but impact as well the size of chunks