Show HN: Reame – a CPU inference server that gets faster as it runs

Reame is a CPU-based LLM inference server designed for resource-constrained hardware like shared vCPUs and ARM boxes, optimized for repetitive tasks over private data where answers exist in provided context rather than model knowledge. Its key innovation is a persistent disk-based KV cache that avoids recomputing identical prompt prefixes, making subsequent requests dramatically cheaper while enabling high accuracy on narrow tasks like document extraction and RAG applications. The project positions itself not as a general-purpose ChatGPT replacement, but as a cost-effective solution for privacy-critical workloads and batch processing on already-owned hardware.