AI Enhancements in 7.1.3

[ppekrol]

With RavenDB version 7.1.3, we’re continuing to expand what you can do with AI natively in your database. This release is extensive (full changelog can be found under this link), but it brings interesting improvements for AI agents, GenAI, and embeddings generation:

AI Agents

Temperature for OpenAI / Azure OpenAI

During the creation of a GenAI task when using OpenAI or Azure OpenAI, you can now select temperature. This controls how deterministic or creative the AI’s responses are. Lower values yield more focused and consistent answers, while higher values encourage more varied and imaginative output.

Azure OpenAI support

AI Agents can now call Azure OpenAI, making it ideal for Azure-first deployments. Configure your Azure base URL and model, and run with the same agent workflow.

OpenRouter (OpenAI-compatible API)

For developers who want to centralize their AI services, RavenDB AI Agents can now be used through the OpenAI-compatible API. This unlocks access to a vast catalog of models and endpoints while preserving your existing agent workflows - a seamless, drop-in change with no code rewrites required.

Server-side streaming

AI Agent responses can now be streamed directly from the server as they are generated. This means text arrives progressively, reducing latency and making long outputs feel much faster in UIs and CLIs.

Embeddings Generation

Google Vertex AI support

You can now use Google Vertex AI as a provider for embedding generation. This extends RavenDB’s support beyond OpenAI, MistralAI, and Hugging Face, giving you more flexibility in how you power vector search.

OverlapTokens for paragraph chunking

We have exposed the OverlapTokens parameter to PlainTextSplitParagraphs and MarkdownSplitParagraphs chunking methods to better preserve the semantic meaning of your text. When provided, it defines how many tokens from the end of one chunk are also included at the beginning of the next chunk, ensuring context isn't lost when splitting text.

You can read more about text chunking in our documentation.

---

This is only part of the changes. Explore the full changelog under this link.

Stay up to date with RavenDB news, connect with our community, and share your feedback! Join our Discord community server - https://discord.gg/ravendb

Happy coding!