Token Optimization and Context Efficiency
Load only the tools a step needs, keep the response fields you name, and compact the JSON when that saves tokens.
Sending the whole API catalog and every response field into the model raises cost and makes the wrong tool more likely. OL Gene loads the tools for this step, keeps the fields the task needs, and compacts the rest when that actually saves space.