Inside HotKV
Tags
Engineering.Measured and explained.
Browse HotKV blog articles by topic, including benchmarks, vector search, semantic caching, LLM costs, Redis compatibility and architecture.
Benchmarks
Measured HotKV results, including test configurations, competing systems, workload limitations and the evidence behind performance claims.
1 articleVector search
HNSW vector search in HotKV, covering recall, query throughput, ingestion speed and memory use on the datasets we have tested.
2 articlesRedis
HotKV and Redis comparisons, protocol compatibility and deployment choices, with the versions and conditions behind each benchmark.
2 articlesPerformance
How HotKV performs under measured workloads, from command throughput to vector queries, with hardware and configuration context.
4 articlesLLMs
Using HotKV in language model applications: model calls, token accounting, response reuse and the quality trade-offs of semantic caching.
2 articlesSemantic caching
Exact and semantic answer caching with HotKV, including similarity thresholds, cache hit rates and how we evaluate answer quality.
2 articlesCost optimization
Measured API costs and token savings with HotKV, including pricing assumptions, workload sensitivity and the limits of extrapolation.
1 articleReleases
HotKV product announcements and releases, with context on the engine, supported capabilities and where to find current documentation.
1 articleArchitecture
Design choices behind HotKV's C++ engine, concurrent command execution, memory management and compatibility with familiar Redis clients.
2 articlesAI agents
Memory, tools and connections for AI agents on HotKV, including conversation and event storage, the MCP server and how to keep agent state safe and bounded.
2 articlesPrompt caching
Provider prompt caching and response caching for LLM apps, what each one saves on tokens and cost, and where each one stops helping.
2 articlesAgent memory
Conversation, semantic, working and episodic memory for AI agents, with HotKV commands, retention rules and the failure modes to design against.
2 articlesMCP
Model Context Protocol servers for AI assistants, including the open-source HotKV MCP server, its fixed tool set, read-only mode and namespaces.