Feature Request: Support video metadata and media resolution Parameters for Gemini Video Understanding
Feature Request: Support video metadata and media resolution Parameters for Gemini Video Understanding: a task in LegoFlow-SWE (Harbor dataset). When using Google Gemini models (Vertex AI / Google AI Studio) for video understanding, users currently cannot control how the model samples video…
The task
When using Google Gemini models (Vertex AI / Google AI Studio) for video understanding, users currently cannot control how the model samples video inputs. The Gemini API supports advanced video processing parameters that allow trading off between token usage/latency and model accuracy, but LiteLLM does not expose…
Part of Lego-X/LegoFlow-SWE.