Research: Experiment with Optimal Reranker and Hybrid Search Params (Top N, Alpha, Top K)

[davidgxue]

Context

• This is a followup action item from this research task #133 (comment)
• Cohere Reranker and Hybrid Search have been implemented with some parameter values that have shown to improve the RAG response quality greatly

Action Items

• Experiment with different parameter values for reranker and hybrid search to maximize the potential performance improvement gain
• If no changes to the current parameter values (top n = 10, alpha = 0.5, top k = 100) yields significant improvements, no implementation or changes to environment variables should be needed.