vss-search-archive

You can search recorded video archives using natural-language queries to find events, objects, actions, and people without manually reviewing footage. The Skill matches your text queries against visual content to retrieve relevant historical footage.

Yes, natural-language search allows you to locate visual content like people in red jackets or forklifts in video footage. You simply describe the objects, actions, or people you are looking for to retrieve matching archive results.

Yes, searching video archives requires Cosmos Embed1 vectors indexed in Elasticsearch. You also need active embedding ingestion through rtvi-embed and a reachable VSS Agent endpoint to process and retrieve the semantic visual content.

To retrieve historical footage recorded between specific times, use a natural-language query with the VSS Agent. Semantic video retrieval searches time- or location-oriented events across ingested video streams indexed with Cosmos Embed1 embeddings.

Video archive search requires a VSS search profile and active embedding ingestion through rtvi-embed to generate and store Cosmos Embed1 vectors in Elasticsearch. This setup enables the semantic matching of natural-language queries to visual content.

Searching for visual entities in ingested video streams is limited to footage that has already been processed and indexed. You must have active embedding ingestion and an operational VSS Agent endpoint to retrieve historical footage successfully.