vss-search-archive

Archived video search accepts natural-language queries to find relevant clips across registered sources using embedding-based retrieval and attribute-based filtering. It resolves sources through the VSS search pipeline and presents results in professional reports.

Yes, you can ingest new video clips or live RTSP streams directly for indexing. Once ingested, the streams become registered sources available for natural-language search and embedding-based retrieval across the archive.

Yes, prerequisites include a deployed VSS search profile, access to the agent backend, and valid credentials for models and embeddings. Without these environment configurations, the search pipeline cannot resolve sources or perform retrieval.

Embedding-based retrieval indexes video clips by generating vector embeddings from ingested streams. When you submit a natural-language query, the system matches query embeddings against indexed video embeddings to locate relevant clips across registered sources.

The best way to filter archived video search results is combining natural-language queries with attribute-based filtering. This approach narrows embedding-based retrieval results using metadata from registered sources to find specific occurrences efficiently.

Video search requires credentials for models and embeddings because the pipeline must authenticate with backend services to generate vector embeddings from video clips and process natural-language queries for accurate retrieval results.