Quail adds prefix sharing to its open-source AI-SQL engine

A Quail team member announced prefix sharing for the open-source AI-SQL engine. They said an example tagging 1,772 long agent traces with issue types using qwen3-4b computed 3x fewer tokens and ran 2.6x faster. A separate commenter highlighted the optimizer's use of document overlap and KV-write cost to choose prefix sharing before LLM calls arrive.