Investigate hybrid search for Qdrant

[demid-ns]

Description

Investigate and implement hybrid search functionality in Qdrant, combining vector similarity search with traditional filtering and keyword-based search capabilities. The goal is to encapsulate this functionality within a custom IMemoryDb type for seamless integration into the kernel-memory library

Technical Notes

• The hybrid search logic must be fully encapsulated within a custom IMemoryDb implementation, adhering to the architecture and design principles of the kernel-memory library
• Leverage the official Qdrant.Client library for interfacing with Qdrant. Ensure compatibility and efficient usage of its features
• Implement functionality to construct both dense and sparse vectors as required by the hybrid search process. Evaluate the need for and potentially develop a custom embedding generator for creating optimal vector representations