kuavi-index

name: kuavi-index description: Index a video file for KUAVi analysis argument-hint: disable-model-invocation: true

KUAVi Index Video

Index a video file to enable semantic search, scene detection, and analysis.

Instructions

1. The user provides a video file path via $ARGUMENTS or in the conversation.
2. Call the kuavi_index_video MCP tool with the video path.
3. Report the indexing results: number of segments, scenes, transcript entries, duration.
4. Suggest next steps: /kuavi-search for finding specific content, or /kuavi-analyze for full analysis.

Example

User: /kuavi-index /path/to/video.mp4

Call kuavi_index_video with video_path="/path/to/video.mp4", then report:

• Video duration
• Number of segments detected
• Number of scene boundaries
• Whether transcript was generated
• Whether embeddings are available

Advanced Options

• Overlapping V-JEPA 2 windows: Set scene_stride for per-frame averaged embeddings (smoother scene detection)
• Feature map storage: Enable store_feature_maps=True for spatial analysis
• Auto FPS: Set auto_fps=True to automatically compute FPS from video duration (targets 120 frames)

Corpus Indexing

For multi-video workflows, use kuavi_index_corpus instead:

kuavi_index_corpus(video_paths=["/path/to/video1.mp4", "/path/to/video2.mp4"])

This builds a cross-video index enabling kuavi_search_corpus for cross-video semantic search.