We've rolled out three significant updates to Moonshine, designed to enhance reliability and efficiency for our users. These improvements focus on subindexing, UI enhancements, and expanded LLM support.

Subindexes

The first major update introduces subindexing. This feature allows users to create targeted subsections within an existing index. Imagine an index containing videos of cats, dogs, and elephants. Instead of querying the entire index, which can be resource-intensive, subindexing allows you to isolate and analyze only the relevant videos, such as those featuring cats.

This targeted approach offers several key benefits: reduced token consumption, faster runtimes, and elimination of redundant re-indexing. The videos within a subindex remain readily available for all Moonshine functionalities without requiring further processing. This is particularly useful for comparing two specific videos, a common use case for many of our users. highlights this scenario, explaining how subindexing simplifies comparative analysis.

Chat Improvements

The second update focuses on UI enhancements, bringing advanced features for "inquire" and "search" directly to the Moonshine platform. Users can now create and manage subindexes directly within the chat interface. We've also integrated JSON schema enforcement. This addresses concerns about LLM hallucinations and ensures consistent, structured output. Users can now define the desired JSON format, and Moonshine will enforce it, providing predictable and reliable results.

Multi VLMs

Finally, we've introduced VLM switcher support. This gives users greater control over their queries by allowing them to select their preferred VLM. We're launching with Gemini, but plan to expand our offerings in the future. Pricing and other VLM-specific details will be available in our documentation. This update provides increased flexibility and customization, allowing users to tailor their experience to their specific needs and budget.

These three updates represent a significant step forward in making Moonshine a more powerful and user-friendly platform. We are confident that these enhancements will streamline workflows, improve accuracy, and provide greater control over video analysis. We encourage our users to explore these new features and discover how they can leverage them to unlock deeper insights from their video data.