[Feedback] Enabling low effort machine learning technology integrations (ml-commons plugin)

All, we’re planning to release the first phase of an extensibility framework within the ml-commons plugin to make it easier for users and partners to integrate machine learning (ML) services and APIs into OpenSearch.

We’ve release some information about our plans:

  1. Feature proposal: [FEATURE] Extensibility for OpenSearch Machine Learning, Phase 1 (ml-commons plugin) · Issue #881 · opensearch-project/ml-commons · GitHub
  2. Design and Integrator Experience (RFC): [RFC] OpenSearch ML Extensibility in ml-commons plugin (remote inference) · Issue #882 · opensearch-project/ml-commons · GitHub

At a high level, this framework will enable you to create and publish integrations simply by creating a blueprint (JSON document). These blueprints can then be used to provision a RESTful API based bridge between OpenSearch and an external ML service or technology, and allow users to interface with the external system through the ml-commons APIs.

This will allow the connectors will be plugged into search and analytics workflows to augment our VectorDB capabilities to support use cases like semantic search, visual search and retrieval augmented generation.

Phase 1 of the framework will be focused on supporting ML inference functionality, but we have plans to extend this functionality to other areas like deployment, data prep and training.

Would love to hear your feedback on the technologies, services and APIs you’re interested in integrating as well as your use cases.