014 GitHub stars
02Intelligent prompt softening to prevent tool overtriggering and over-engineering
03Automated removal of unsupported beta headers and 1M context parameters
04One-shot model string updates for Anthropic, AWS Bedrock, Vertex AI, and Azure
05Integrated support for setting the 'high' effort parameter in API calls
06Heuristic-based replacements for 'thinking' sensitivity in non-extended thinking requests