ai_tokenopt 0.5.10

Adaptive token optimization engine for LLM inference pipelines — compresses prompts, conversation history, tool schemas, and output streams to minimize token usage while preserving response quality.