Signed URLs for HLS streams and thumbnails
The platform now encrypts HLS streams and signs the URLs for HLS streams and thumbnails.
A signed URL expires. How long it stays valid depends on what it points to:
- HLS streams: A signed URL is valid for 1 hour.
- Thumbnails: The URL is valid for at least 1 hour and up to 2 hours.
Higher rate limits for Jockey responses
You can now send more requests to create Jockey responses. The Free plan allows 10 requests per minute and 200 requests per day, up from 5 requests per minute and 15 requests per day. The Plus and Pro plans allow 30 requests per minute and 1,000 requests per day, up from 15 requests per minute and 200 requests per day.
For all limits, see the New billing rate limits page.
See the result for each item you add to or remove from an item collection
You can now see the result for each item you add to or remove from an item collection. The response lists, for each identifier of an item in the request, whether the platform added, removed, or skipped the item. For a skipped item, it also gives the reason.
Removing items returns this response with a 200 status. Before, removing items returned a 204 status with no response body.
Sort knowledge store items by supported metadata keys
When you list the items in a knowledge store, you can sort by these custom metadata keys: view_count, like_count, comment_count, share_count, video.view_count, video.like_count, video.comment_count, and video.share_count. Use the sort_by parameter with the metadata.<key> value. The platform sorts by numeric value. If the value is not a number, the platform treats the item as having no value for the key. For any other metadata key, the platform returns a 422 error with the parameter_invalid code. Sorting by created_at and updated_at does not change.
Video duration limits for knowledge store items
The platform now rejects a video that is shorter than 1.1 seconds or longer than 24 hours when you add the video to a knowledge store. The request returns a 422 error with the parameter_invalid code.
For details, see the Create a knowledge store item page in the API Reference section.
Responses API types change in the 1.3.8 SDKs
Version 1.3.8 of the Python and Node.js SDKs changes the types for the Responses API. Upgrading can require changes to your code, depending on the version that you have installed. The API is unchanged. Follow these steps to upgrade.
Update your code
Select the tab for your previous version of the SDK.
1.3.6 or earlier
1.3.7
SDK type changes (breaking change)
The usage field is removed. This change affects you if your code reads the usage field or uses the ResponseUsage type. To find the cost of a response, read the billing field. It contains the amount and the currency. It does not contain token counts.
Python
Node.js
Before:
After:
The keepalive stream event type is removed. This change affects you if your code uses the ResponseStreamKeepAliveEvent type. Remove the code that handles it.
The status field of an output item is an optional string. This change affects you if your code reads the status field and assumes that it has a value.
Other SDK type changes
- The
typefield ofResponseOutputItemis a string. The SDK still exportsResponseOutputItemTypewith the valuesmessage,function_call, andfunction_call_output. - The
typefield ofResponseAnnotationis a string. The SDK still exportsResponseAnnotationTypewith the valuesvideo_citation,image_citation, andcollection_citation. The platform also returns theurl_citationvalue. - An output item with the
jockey:web_searchtype contains the details of the search in theactionfield. The response stream includes thejockey:web_search.in_progress,jockey:web_search.searching, andjockey:web_search.completedevents. The response contains these items and events only when the request sets theincludeparameter to["intermediate_outputs"].
Filter knowledge stores by asset
When you list knowledge stores, you can now filter them by asset. Use the asset_id parameter to pass the identifier of the asset. A knowledge store matches when it contains an item that uses the asset.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Filter knowledge store items by asset
When you list the items in a knowledge store, you can now filter them by asset. Use the asset_id parameter to pass the identifiers of the assets. You can pass 1 to 50 identifiers. An item matches when it uses one of these assets.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Set embedding segment durations in fractions of a second
You can set the length of embedding segments in fractions of a second, for example 2.5 seconds. The API accepted these values before this change. The API reference and the SDK types now show them as valid. Marengo 3.0 ignores the fractional part of the value. For example, it treats 2.5 as 2.
If you use version 1.3.6 or earlier of the Python SDK, the SDK types the duration_sec and min_duration_sec fields as integers. It can reject or coerce fractional values depending on the installed Pydantic version. Upgrade to the latest version of the Python SDK to pass fractional values to these fields:
Show activity summaries while streaming a response
Set include to ["activity_summaries"] with stream: true to receive live progress summaries as jockey:activity events. Each snapshot contains an activity ID, revision, status, and summary. Activity summaries are delivered only in the stream and are not included in the final response’s output or saved message history.
intermediate_outputs is deprecated. Its existing streaming and non-streaming behavior is preserved, including intermediate messages, function calls, and function-call results in output. For progress displays, migrate to activity_summaries and handle jockey:activity events. The two include options cannot be combined.
Sort knowledge store items by custom metadata
You can now sort the items in a knowledge store by the value of a custom metadata key when you list them. Set the sort_by parameter to metadata.<key>. For example, metadata.view_count sorts the items by the view_count key.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Deleting an asset used in a knowledge store returns a 409 error
If knowledge store items use an asset, you cannot delete the asset. The request returns 409 asset_has_knowledge_store_items, even when you set the force parameter to true. To delete the asset, first delete every knowledge store item that uses it.
For details, see the Delete an asset API reference and the SDK reference for Python and Node.js.
Search a knowledge store by item collection
When you search a knowledge store, you can now limit the search to the items in specific item collections.
If a search matches more than 10,000 items, the platform returns a 422 error with the metadata_filter_too_broad code.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Introducing metadata, enrichment, and creation time filters for listing knowledge store items
You can now filter the items in a knowledge store when you list them. Use the metadata, enrichment, created_at_gte, and created_at_lte parameters to filter by custom metadata, enrichment values, and creation time.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Exclude items and filter custom metadata by range in knowledge store search
Knowledge store search accepts item_id.nin to exclude every clip or image belonging to 1 to 20,000 listed items. You can use it on its own or with either item_id.eq or item_id.in. An item listed in nin is excluded even when eq or in selects it.
The user_metadata filter also accepts inclusive gte and lte bounds on numbers and RFC 3339 timestamps with a timezone. You can use one or both bounds. Every metadata key you specify must match. These ranges apply to custom metadata you stored on the source asset or the knowledge store item.
The new billing starts on October 15
The new billing starts on October 15, 2026. It replaces the separate Free plan, Developer plan, and Jockey subscriptions with one set of plans: Free, Plus, Pro, and Enterprise. Each plan includes a Managed Intelligence allowance and a monthly Workload allowance. Managed Intelligence is the service that stores your assets and keeps them ready to use. It includes processing and updates to the latest models. A Workload is a task that you run on your assets. Examples are creating Jockey responses, analyzing, segmenting, and embedding.
Introducing item tags
Video items in a knowledge store now accept tags. A tag marks a moment or a range of the video and carries custom metadata as key-value pairs. A video item holds up to 100 tags.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Introducing tag filtering for knowledge store search
Knowledge store search now accepts a tags filter. The search returns only the clips that overlap the moments and ranges that matching tags mark on video items. A tag matches when its tag_metadata field matches the metadata you specify.
To use this feature, upgrade to the latest version of the Python or Node.js SDK:
Jockey is now generally available
Jockey is a unified agentic system that reasons across your videos and images.
What you can do
- Summarize all your videos and images: themes, subjects, patterns, and key statistics
- Search your videos and images by describing what you want, and receive an explanation of each match
- Organize and categorize your videos and images
- Find and sequence moments into highlight reels
Introducing custom metadata, enrichment, and creation time filters for knowledge store search
Knowledge store search now accepts the following filters:
user_metadata: Narrows results by custom metadata.enrichment: Narrows results by enrichment values.created_at: Narrows results by the creation time of the item.
If a search matches more than 10,000 items, the platform returns a 422 error with the metadata_filter_too_broad code.