FrameCite Video Knowledge Agent

FrameCite is a multimodal video knowledge agent that enables users to ask natural-language questions about video content and receive evidence-grounded answers with playable timestamp citations. It processes videos by transcribing speech, generating visual descriptions of sampled frames, and indexing both modalities for hybrid semantic and keyword retrieval. A LangGraph-based AI agent then searches relevant video segments, inspects original frames to verify visual evidence, and generates structured answers supported by precise video references.