The Right Approach to Context Compression: Why APX Uses Two Controls, Not One

APX's design separates the timing of compression from the amount of recent context preserved, solving a critical problem: naive single-parameter systems either waste memory or lose important details. By splitting compact_threshold (when compression starts) from keep_recent (what stays fresh), APX gives developers explicit control over the memory-context trade-off. This demonstrates why fine-grained configuration is essential for systems managing long conversation histories.