Get Started

Migration guide

Note

The timeline and actions on this page apply to the first-party API. If you use Marengo through Amazon Bedrock, see the Amazon Bedrock migration guide instead.

Marengo 2.7 will be deprecated after TwelveLabs automatically reindexes your videos to Marengo 3.0. Starting mid-March 2026, reindexing will roll out progressively.

Update your application to ensure continuous service when the cutover happens.

The following fields are preserved during migration:

  • Index ID
  • Video ID
  • Video created date
  • User-provided metadata

The updated_at field will reflect the migration date.

Migration steps

1

Update your SDK

Install the latest version of the Python or Node.js SDK.

2

Update the version strings in your code

Replace the old version strings with the new ones:

FeatureMarengo 2.7 valueMarengo 3.0 value
Indexesmarengo2.7marengo3.0
Embeddingsmarengo-retrieval-2.7marengo3.0
3

Update your search requests

Deprecated parameters are now ignored, and audio processing has changed. Your code will not return errors, but results may differ. See the Update your search requests section for details.

4

Regenerate all embeddings

Marengo 2.7 embeddings are not compatible with Marengo 3.0. This is a breaking change. See the Regenerate all embeddings section for details.

Update your search requests

MethodDescriptionAction
POST /searchThe adjust_confidence_level, sort_option, and threshold parameters and are deprecated in Marengo 3.0Update your logic use the rank field instead.
POST /searchThe score and confidence fields in the response and are deprecated in Marengo 3.0Update your logic use the rank field instead.
POST /searchThe audio search option now processes only non-speech audio (music, sound effects, environmental sounds).If you search for spoken content, set search_options to transcription instead of audio. Using audio for spoken content returns incomplete results.
POST /searchYou can now search for spoken words using the transcription search option.Set search_options to transcription in your search requests.
POST /searchYou can now control how the platform matches your query against spoken words.Set transcription_options to lexical for exact word matching, semantic for meaning-based matching, or both for comprehensive results.

For details about how search options work, see the Search options section.

For a complete guide on making search requests, see the Search page.

Regenerate all embeddings

MethodDescriptionRequired action
- POST /embed/tasks
- POST /embed		The embeddings types and dimensions are not compatible with Marengo 2.7.You must regenerate all your text, audio, image, and video embeddings with Marengo 3.0.
GET /embed/tasks/{task_id}The visual-text value has been renamed.Set embedding_option to visual instead of visual-text.
GET /embed/tasks/{task_id}The audio embedding option now processes only non-verbal audio.If you need embeddings for spoken content, set the embedding_option parameter to transcription.

For a complete example of creating embeddings with Marengo 3.0, see the Create video embeddings page.

For details about embedding options, see the Embedding options section.

Additional resources

Python SDK Reference
Node.js SDK Reference
API Reference
Discord