Paged Kv Cache Management With Prefix Sharing

via “prefix caching with semantic token matching”

High-throughput LLM serving engine — PagedAttention, continuous batching, OpenAI-compatible API.

Unique: Implements semantic-aware prefix caching using a trie-based prefix tree with hash-based matching and zero-copy KV page sharing, enabling cross-request cache reuse without explicit user configuration

vs others: Reduces KV cache computation by 30-50% for RAG/few-shot workloads vs no caching, with minimal overhead due to hash-based matching vs tree traversal

OpenAI and Anthropic compatible server for Apple Silicon. Run LLMs and vision-language models (Llama, Qwen-VL, LLaVA) with continuous batching, MCP tool calling, and multimodal support. Native MLX backend, 400+ tok/s. Works with Claude Code.

Unique: Adapts vLLM's paged KV cache design to MLX's unified memory architecture, enabling efficient cache sharing across requests while respecting Apple Silicon's memory constraints; tracks page allocation state to prevent fragmentation

vs others: More memory-efficient than contiguous caching for multi-request scenarios; enables longer context windows than naive caching; better cache utilization than request-level caching

via “multi-level kv cache management with prefix caching”

A high-throughput and memory-efficient inference and serving engine for LLMs

Unique: Implements block-level KV cache with prefix caching that tracks cache blocks as first-class objects with ownership and eviction policies, enabling cache reuse across requests without recomputation. Supports disaggregated serving via KV cache transfer protocol, allowing cache to be stored on dedicated cache servers separate from compute workers.

vs others: Reduces memory usage by 20-40% on multi-turn conversations vs. standard KV cache by reusing cached prefixes; disaggregated serving enables 10x larger batch sizes by decoupling cache capacity from compute capacity.