Twelve LabsInstall

Summary

Search videos in a specified Marengo index using a natural language query and configurable search options to find relevant video clips or grouped video results.

Inputs

• Index ID (required) — ID of the Marengo index to search; must correspond to a Marengo model.
• Query Text (required) — The natural language search query describing what to find in the videos.
• Search Options (required) — The modalities to search within, including visual content, audio content, text visible in video (OCR), and logos; must match the index's model options.
• Additional Options — Optional parameters including JSON filters (e.g., duration constraints), grouping results by clip or video, inclusion of user metadata, logical operator to combine search modalities (AND/OR), pagination limit (max 50), sorting method (relevance score or clip count), and confidence threshold for filtering results.

Output shape

a JSON object containing search results grouped by clips or videos as specified, including matched metadata and optionally user metadata; each result reflects the confidence threshold and sorting options applied.

Results may be filtered and limited based on provided additional options; includes error handling that outputs error details per item if the operation encounters issues and 'Continue On Fail' is enabled.

Examples

Example 1: Search for videos featuring a man talking about technology with visual and audio modalities, grouping results by clips and including user metadata.

Set 'Index ID' to target Marengo index, 'Query Text' to 'A man talking about technology', select 'Visual' and 'Audio' for Search Options, leave Additional Options default with grouping by clip and default confidence threshold medium.