Dataiku Version 13.3

Dataiku 13.3 delivers significant updates to Deploy Anywhere capabilities, testing frameworks, Generative AI features, visualization enhancements, and user experience improvements. Learn more in our release notes and find instructions to update your instance.

Highlighted Updates

Deploy Anywhere to Databricks completes the Dataiku MLOps cloud integration support
Comprehensive testing framework for Dataiku projects
Official support for image generation in LLM Mesh
AI Generate recipe with natural language
Updates to the enterprise-ready, GenAI application Dataiku Answers(Check out the new featurette!)

Explore many more feature updates organized by Dataiku capability below.

Generative AI
XOps
Data Prep
Visualization & Data Storytelling
Homepage and UI Improvements
Dataiku Answers

Generative AI

Explore all Generative AI capabilities in Dataiku

Image Generation

The LLM Mesh API now officially supports image generation capabilities with both text-to-image and image-to-image generation. Supported models include:

DALL-E 3
Imagen 3
Stability AI models
FLUX.1

Available through LLM Mesh connections and accessible in Dataiku Answers, these capabilities enable:

Text-to-image generation
Image-to-image transformations (edit, inpainting, outpainting)
Integration with existing Dataiku workflows while maintaining security and governance

READ THE BLOG BREAKING DOWN VISUAL AI WITH DATAIKU >

AI Generate Recipe

Access AI-powered recipe generation directly in the Flow or Explore view when selecting datasets. Key features include:

Enhanced productivity through natural language descriptions of desired data transformations
Contextual suggestions based on dataset schema and sample data
Single recipe suggestions with continuous improvement through user feedback
Automatic naming of output datasets based on prompt context for better project understanding

AI GENERATE RECIPE CONCEPT ON THE DATAIKU KNOWLEDGE BASE >

Updates to LLM Mesh

The latest LLM Mesh improvements introduce support for new models, connections, and features, including:

JSON mode support for structured outputs with Vertex AI Gemini Pro and other providers
LLM-as-a-judge for enhanced prompt injection detection
Support for AWS Bedrock Llama 3.2
Integration with Snowflake Cortex
Added AWS OpenSearch as vector store option
Function calling support for Vertex AI Gemini and HuggingFace local models

READ THE VECTOR STORE TYPES DOCUMENTATION >

Knowledge Bank Updates

New update modes provide greater flexibility:

Overwrite (default mode): Completely rebuilds the Knowledge Bank with new content
Append: Adds new input rows to the existing Knowledge Bank, allowing for potential duplicates
Smart Sync: Inserts new rows, updates modified ones, and deletes rows that were in the last run but not in the current input
Upsert: Inserts new rows and updates modified ones, without performing any deletions

HOW TO USE KNOWLEDGE BANKS INCLUDED IN THE RAG TUTORIAL >

READ THE UPDATE METHODS DOCUMENTATION >