Create a new deep search over one or more collections. Deep search uses agentic retrieval and LLM summarization to find specific moments across your video data.
The search can be processed synchronously (default), streamed via SSE, or run in the background.
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Deep search creation parameters
Knowledge base configuration. Determines which content to search.
Three source modes are supported:
collections: Search within specific collectionsfiles: Search specific files by URL or file ID (uses default index)default: Search all default-indexed files for the accountThe search query.
1The scope of results to return. 'segment' returns individual segments, 'file' returns file-level results.
segment, file Maximum number of results to return. Actual count may be lower when exclude_weak_results is enabled.
1 <= x <= 500When true, removes results tagged as weak matches by the synthesis LLM.
Additional fields to include in the response. 'search_queries' includes the intermediate search query plan.
search_queries Stream the response via SSE. Mutually exclusive with background.
Process in the background. Returns immediately with status 'in_progress'. Mutually exclusive with stream.
Deep search created successfully
Deep search ID
Object type identifier
deep_search Current status of the deep search
in_progress, completed, failed, cancelled Unix timestamp of when the deep search was created
The original search query
The scope of the search results
segment, file LLM-generated synthesis text summarizing the results
Array of search results
Total number of results
Maximum number of results requested
Intermediate search query plans (included when requested via include=['search_queries'])
Token and search call usage
Error details if the deep search failed