emmetsurmeli/track-sprinting-ai

Track Sprint AI is a local Streamlit sprint-video review application. An athlete sets up a profile, uploads and trims a recording, then reviews locally computed pose and movement measurements. The app supports human-reviewed contact timing and landing posture, and sends eligible measurements, athlete context, and selected research evidence to OpenAI for constrained coaching—not the video. Results can be inspected, revisited in a dated training calendar, compared across eligible sessions, and exported. The graph follows documented capabilities and sampled orchestration; some UI-to-module wiring and unsampled implementation details are intentionally left out.