vss-ask-video

You can analyze video frames by submitting a named sensor video clip to a visual question answering agent, which performs frame-level analysis to identify objects, actions, timing, and safety details. This returns a natural-language answer fresh from the video content.

Yes, you can ask fresh visual questions about recorded video clips without relying on prior tool output, search hits, or metadata. The video understanding agent processes the clip directly to extract visual facts such as whether a worker wears PPE or when a person climbs a ladder.

To run video understanding queries you need a VSS profile with the video understanding tool enabled, access to a named sensor video, and the ability to call the VSS /generate endpoint. Provide the sensor id and the video clip to request visual analysis.

Video understanding supports safety-related visual analysis by examining frames to verify PPE usage, determine when a person climbs a ladder, or extract other safety-relevant visual facts from a recorded clip.

To get a natural-language answer from a visual question, call the VSS /generate endpoint with a named sensor video clip. The agent runs the video understanding tool to perform frame-level analysis and returns the answer directly from the visual content.