Author: Alice Cima, Kedro community
The Kedro community is excited to announce the release of Kedro 1.2.0 and Kedro-Viz 12.3.0, bringing new experimentation capabilities alongside foundational improvements that make building and managing modern AI workflows easier and more flexible. This release introduces early support for agentic and GenAI use cases and improves pipeline visibility to help teams move from experimentation to production more effectively.
What We’re Most Excited About
This release continues to strengthen Kedro’s core while unlocking more powerful experimentation capabilities. The new experimental LLM context nodes and agentic workflow starter make it easier to assemble, test, and iterate on GenAI workflows in a structured and production-aware way. On top of that, enhanced node previews in Kedro-Viz significantly improve pipeline visibility, making complex workflows easier to understand, debug, and communicate.
Key Highlights
Kedro 1.2.0
- Expanded multi-pipeline execution support
- Run multiple pipelines within a single session using the –pipelines CLI option and pipeline_names argument in KedroSession.run() for more flexible orchestration.
- Introduced the @experimental decorator
- Public APIs that are still evolving can now be clearly marked as experimental, helping users adopt new functionality with better transparency.
- Experimental LLM context nodes for GenAI workflows
- Added llm_context_node and LLMContextNode to assemble LLMs, prompts, and tools into runtime LLM contexts directly inside Kedro pipelines.
- Node preview extensibility
- Added experimental preview_fn support to the Node class, enabling user-defined preview functions for richer inspection and debugging.
- New agentic workflow starter
- Introduced support-agent-langgraph, an experimental starter showcasing agentic workflows built with LangGraph and integrated prompt management and tracing via Langfuse or Opik.
These experimental features are under active development and may continue to evolve in future releases, and the community welcomes feedback from users exploring these new capabilities.
Kedro-Viz 12.3.0
- Node previews for improved pipeline visibility
Added support for visual node previews including:- Mermaid diagrams
- Text previews
- Image previews
- These enhancements make it easier to inspect pipeline behavior and understand complex workflows visually.
Release Notes
What’s Next
Looking ahead, the Kedro community will continue investing in support for GenAI workflows, deployment guides, and broader developer experience improvements — while also strengthening the core foundations needed to support an expanding range of AI and GenAI use cases. This includes evolving Kedro’s session model to support repeated multi-run execution within a single session, alongside an API layer that enables pipelines to be invoked programmatically, including by external services and AI agents.
As data and AI systems increasingly combine deterministic pipelines with less-deterministic agentic components, Kedro will continue focusing on providing the engineering discipline, reproducibility, and system-level structure needed to scale these workflows reliably in production.
Now is a great time to get involved and help shape the future of the project. Whether you are interested in production AI systems, agentic workflows, experimentation tooling, or improving the developer experience, contributions, ideas, and feedback from the community are always welcome.
Visit Kedro Website, explore the project on Kedro GitHub, or join the community to get started.
