Architect a rolling Count-Min Sketch filter to aggressively cache high-frequency fractional asset reads

[Damola-Sodiq]

Description

Integrate a Count-Min Sketch algorithm in the Node.js backend to probabilistically track hot fractional assets and aggressively cache them in memory.

Architecture & Context

Standard LRU caches consume excessive memory when dealing with thousands of long-tail assets. A Count-Min Sketch provides a mathematically bounded, low-memory footprint way to identify the top 1% of queried tokens and cache their metadata heavily.

Technical Requirements

• Implement the Count-Min Sketch data structure strictly in TypeScript.
• Hook the estimator into the Fastify request lifecycle to score asset IDs on read.
• Evict cold assets and promote hot assets into a V8 in-memory map dynamically.

Acceptance Criteria

• The filter accurately identifies trending assets with a known error bound.
• Memory consumption for the tracker stays under 15MB.
• Database read load is empirically reduced by > 50% during traffic spikes.