DSPy 3.0 + Agent Bricks and SuperNetiX

Shashikant Jagtap

At Databricks Data + AI Summit , there are several big announcements in made for Databricks platforms related to AI, Agent and DSPy that are notable to mention as latest advancement in the AI field. Most of the announcement are tied to Databricks own tools and platforms but there are some that could be related to DSPy and upcoming framework SuperNetix from Superagentic AI. In this post we will cover snapshots of the announcement and how DSPy 3.0, Agent Bricks and SuperNetix are correlated.

Agent Bricks

Agent Bricks offers a straightforward, no-code solution for creating and refining specialised AI agent systems tailored to specific applications. By simplifying the development process, Agent Bricks enables users to concentrate on high-level priorities such as problem definition, data analysis, and performance metrics. Agent Bricks simplifies AI agent development by automating evaluation and optimization. Users define their task in natural language, and Agent Bricks handles the rest, creating benchmarks, fine-tuning models, and ensuring cost-effectiveness. This enables enterprises to quickly deploy high-quality, efficient agents in production, streamlining the process from lab to deployment.
It seems like Databricks took original DSPy concepts and made nice UI on top of the popular concepts like Signatures, Modules, evaluations  and optimisers for databricks customers.
You can fiddle with the documentation of the Agent Bricks here. Agent bricks is tightly coupled with Databricks ecosystem.

DSPy 3.0

On the same day, DSPy 3.0 released, as per the currently merged PRs it doesn’t seems like there is anything major changed but as part of the session on DSPy 3.0 , there are significant advancements in prompt optimisation and fine-tuning/RL capabilities, paired with seamless MLflow integration for observability. This needed to be validated in the recent release. There are also decommissioning of custom vector databases.

What this means for SuperNetiX

SuperNetiX is upcoming agent framework inspired by DSPy. Framework is still under development but with release of Agent Bricks some of the SuperNetiX ideas has been implemented but core of the SuperNetiX is still in place. SuperNetiX will be still different that Agent Bricks in multiple ways including

  •  Not tied to Databricks or any other commercial products
  • Not tied to any agent frameworks
  • Optimisation and evaluation done with differently than mentioned in the Agent Bricks documentation.
  • SuperNetiX won’t be no-code platform, it requires little bit of Python coding.

The release of DSPy 3.0 and Agent Bricks  might have minor impact on the current state of development which involve

  •  Change some features to avoid similarity with Agent Bricks
  • Make use of DSPy 3.0 features if there is significant improvements

Conclusion

Databricks team has done amazing job in launching Agent Bricks allowing Databricks users to build AI Agent on their platform using the full power of the DSPy features. DSPy team has released major version allowing open source community to build powerful features. Thanks to both Databricks and DSPy team.