Developed advanced video metadata handling and model-version aware media resolution for Gemini models in the BerriAI/litellm repository. Focused on backend development and API integration, the work introduced fine-grained controls for video processing, such as configurable frame extraction rates and time ranges, applicable across all Gemini model versions. Conditional logic was implemented in Python to ensure that media resolution checks apply only to Gemini 3+ models, maintaining compatibility with earlier versions. Comprehensive updates to documentation and test coverage in Markdown and Python ensured consistent behavior across model versions, emphasizing robust cross-version support and maintainability without addressing major bug fixes during this period.