Asynchronous tasks
One of our customers creates vector indexes in an AWS Lambda environment, where operations must complete within a 15-minute window, and costs are based on computation time. This can be a problem because index creation on large datasets can take a long period of time, and Postgres generally requires maintaining a connection for the entire duration of a command.
To address this, we added support for asynchronous tasks inside Lantern. Asynchronous tasks allow you to run a long-running Postgres command without needing to maintain an active connection. You can monitor the progress of your tasks through the Lantern dashboard. Read the full documentation here.
Note: This feature is currently only available on Lantern Cloud’s production plan.
Weighted vector search
We added lantern.weighted_vector_search to enable searching over multiple vector columns with different weights, while still using vector indexes.
For example, if you have a products table with name and description columns, you can search for products that match a query in the name column with a higher weight than the description column. Under the hood, lantern.weighted_vector_search calculates a joint distance metric based on the weighted distance between the vectors in the table and the provided vectors.
Read the full documentation here.
Ubicloud control plane
Ubicloud is building an open, free, and portable cloud. They support multiple services including managed Postgres. We extended the managed Postgres service to support Google Cloud as an infrastructure provider.
Dashboard improvements
We did a light redesign, added a table viewer tab, and a database insights tab. At the moment, we support viewing the most frequently executed queries and slowest queries.
Note: Database insights is currently only available on Lantern Cloud’s production plan.
New embedding model
We added support for BAAI/bge-m3, a multi-lingual, multi-functional, and multi-granularity text embedding model.