The Nu Blog
The latest news, releases, features and commentary
Decision models in Nussknacker: typed AI decisions for Kafka and API workflows
Decision models like TypeSafe Jev and Cloudflare Clef return typed answers with probabilities. See how to add one to a Nussknacker scenario and what you gain.
What is new in Nu – September 2026
Nussknacker in September 2026: Change Data Capture for tables, nodes with named outputs, a WebSocket Server source and live data for Request-Response.
What is new in Nu – August 2026
Nussknacker in August 2026: deployment targets, the World Map component, improved enum support, improved Compare versions and more.
Turn your Flink UDFs and PTFs into low-code components
Turn Apache Flink UDFs and PTFs into low-code components in Nussknacker — auto-discovered, typed, and usable without SQL.
Serving real-time Kafka data to AI agents through MCP
Serve real-time Kafka data to AI agents as an MCP tool: aggregate events in Flink, store the state, and expose it to agents — built visually in Nussknacker.
Typed structured output from LLMs in Nussknacker — no more parsing JSON by hand
Turn LLM output into typed fields in Nussknacker: define a JSON Schema once and get autocomplete, design-time validation, and null-safe parsing.
Real-Time Decision Logic Needs Its Spreadsheet Moment
Spreadsheets let business experts build their own models and see each change at once. Real-time decision logic still lacks that: why, and what it needs.
Turn REST API into a Kafka Event Stream — No Code Required
Bridge any REST API to Apache Kafka without code. Visually turn the OpenSky flight tracking pull API into a live Kafka event stream with Nussknacker.
Complex Event Processing with Apache Flink — Visual CEP with Nussknacker
Learn how to build visual Complex Event Processing scenarios with Apache Flink MATCH_RECOGNIZE and Nussknacker using real-time ADS-B flight data.
Building Iceberg Lakehouse with AWS Glue
Build a medallion lakehouse on S3 with Apache Iceberg and the AWS Glue catalog in Nussknacker Cloud, moving sales data from the Bronze to the Gold layer.
Understanding Event Time in Nussknacker
Learn why Event time is crucial in stream processing and how Nussknacker leverages it for reliable stateful computations.
Generation-Augmented Retrieval (GAR) for Product Recommendation
The LLM answers first and a vector store matches the answer to your catalogue. This inverts RAG, avoids prompt size limits and recommends only real products.
Why Streaming SQL Is Not the Right Tool for Authoring Event-Driven Stream Based Algorithms
Streaming SQL suits ETL and dashboards but breaks down on state, time, branching and external calls in event-driven apps. Flink SQL is the reference point.
Integrate Azure Databricks MLflow for machine learning model management and inference
Connect Nussknacker Cloud to Azure Databricks managed MLflow. Registered models then appear as enricher components for inference in a fraud detection scenario.
How to train and register ML models in Azure Databricks
See how to train and register ML models in Azure Databricks and deploy them in Nussknacker Cloud for credit card fraud detection.
Next-Generation Real-Time Rating Systems
Synchronous architectures struggle when rating must keep up with live usage. This post shows how stateful stream processing in Nussknacker handles it.
A real-time rating system prototype built in one day
I built a real-time rating prototype in Nussknacker in one day. Flink provided the session windows, the customer balance state and recovery from failure.
Real-time IoT data from MQTT in Nussknacker Cloud
Nussknacker Cloud has MQTT source and sink nodes. This guide creates an MQTT broker, configures the integration and runs a scenario on machine data.
Real-time transaction fraud detection
Transaction fraud has to be caught while the payment happens. This demo scenario in Nussknacker flags suspicious transactions and sends alerts to Kafka.
Telecom’s credit scoring system with ML inference
Mobile carriers score credit risk for handset installments. Data scientists deploy MLflow models in Nussknacker, and business analysts change the credit rules.
Streaming SQL alternative
Streaming SQL makes you treat each event as a row in a table. This post takes a streaming SQL example from RisingWave and builds the same logic in Nussknacker.
Real-Time ML-Driven Recommendations Use Case
Snowplow tracks clicks in a Shopify store, Nussknacker scores them with an ML model in real time and sends the recommendations back to the shop.
Using Nussknacker with Apache Iceberg: Periodical report example
Use Nussknacker with Apache Iceberg via Flink catalogs for data ingestion, transformation, enrichment, and low-code business logic.
ML models inference in fraud detection
Train a fraud detection model, track it in MLflow, call it from a Nussknacker scenario, then improve the results with A/B tests and model ensembles.
Next Best Action recommendation management simplified
A Polish telecom runs next best action recommendations in Nussknacker's request-response mode. A Dunder Mifflin-style example shows how to read and edit them.
Exactly-once event processing
Nussknacker delivers messages exactly once end to end through the Kafka sink, using Flink. Side effects such as enrichments may still repeat after a failure.
Batch processing on Apache Flink
Nussknacker is adding a batch mode on Apache Flink. This early preview ingests transactions, deploys with parameters and keeps business logic in one place.
Data processing modes: Streaming, Batch, Request-Response
Understand the difference between streaming, batch, and request-response data processing — and when to use each mode in real-time systems.
Nussknacker 1.15 release highlights
Nussknacker 1.15 adds a batch processing mode built on Flink's Table API and Catalogs, support for Flink 1.18, node search in the editor and a new theme.
Webinar video: Real-time data processing for the people
In this webinar recording, Zbigniew Małachowski lists the features real-time data tools need beyond drag and drop, for fraud detection, streaming ML or IoT.
How to use Nussknacker together with Kafka & Schema Registry
This IoT example reads Kafka topics with Avro schemas from Schema Registry, flags anomalies in Nussknacker and shows what happens when a schema evolves.
SMS Pumping Fraud: Detecting Artificially Inflated Traffic
Fraudsters register an MVNO and pump SMS traffic through other operators. A Nussknacker scenario flags senders of 10,000+ messages to a few numbers in 8 hours.
A guide to OpenAPI components
Nussknacker reads OpenAPI definitions and turns REST services into enricher components for scenarios. This guide shows it on premises and in Nussknacker Cloud.
Introduction to Telecommunications Fraud Detection
Learn how real-time tools help detect telecom fraud, from spoofing to false capacity and how Nussknacker fits into modern fraud management.
Introduction to integrating Google Firebase with Kafka Connect and Nussknacker
Discover how to connect Firebase with Kafka Connect and Nussknacker to enable real-time analytics, event-driven workflows and data processing
How to build a bridge between IT and the business teams?
Fast-changing data sources, formats, schemas and processing models keep business teams waiting on IT. This post describes a bridge application between the two.
Azure Active Directory (Entra ID) Integration in Nussknacker
Set up single sign-on to Nussknacker with Azure Active Directory, now Microsoft Entra ID: app registration, a secret, API permissions, roles and OIDC settings.
Nu chapter for TouK
Respect for the inimitable software engineers of TouK who have joined Snowflake, or who will continue working with us on Nussknacker.
Combine streaming and historical data for real-time decisioning
React to customer behaviour in Confluent Cloud using customer profiles from Snowflake. Nussknacker lets non-developers combine both in one scenario.
What is wrong with low-code tools in streaming?
Real-time actions-on-data apps such as ML on streams are rarely simple enough for drag and drop. This post names the features low-code streaming tools miss.
Translating business case into streaming concepts
Streaming data is not enough in most business cases. You need to enrich the events with additional information to perform meaningful business actions.
Adapt Nussknacker to your domain with configuration only
Nussknacker can be adapted to your domain without code. This post shows environment labels, comment validation, component settings and OpenAPI filters.
How to handle quickly evolving Elasticsearch documents in Nussknacker
Learn how Nussknacker handles evolving Elasticsearch schemas and enables writing rules on flexible, dynamic data structures.
Speed up your OpenBanking decision scenarios
Check how Nussknacker uses open banking standards to help domain experts turn data into fast, actionable decisions with low-code tools.
Stream Designer and Nussknacker Designer comparison
We compare Confluent Stream Designer with Nussknacker Designer on target users, deployment, connectors, data enrichment, testability and observability.
Real-time sentiment analysis using Nussknacker and SpEL
Explore how to use Nussknacker, Kafka, and SpEL to build real-time sentiment analysis with schema-based topics and dynamic decision logic.
Low-code for business rules – are we there yet?
Low-code tools now cover analytics, web apps and data pipelines. Defining business rules is still hard for domain experts, and this post looks at why.
Building a credit rating verification service with Nussknacker
Discover how visual low-code tools like Nussknacker make complex credit rating algorithms easier to build, understand, and maintain.
Is SQL enough to democratise stream processing?
Running streaming SQL isn't enough for real-time apps. Learn why tools like Nussknacker are key beyond simple ETL in stream processing.
Nussknacker 1.5
Nussknacker 1.5 brings single Kafka source and sink components for Avro and JSON Schema topics, improved Filter and Choice nodes and a full outer join.
When ML serving needs enrichments and decision logic
See how Nussknacker lets you deploy ML models and build advanced decision logic with transformations, enrichments, and real-time data flows.
Stretching old solutions to fit new problems doesn’t work – a genealogy of Nussknacker
Stream processing grew from Hadoop batch jobs through Kafka and Flink, yet it still left business teams behind. This is the story of why we built Nussknacker.
Nussknacker has many flavours: which engine is for you?
One Nussknacker Designer can deploy scenarios to different engines: Flink for stateful streaming, or Streaming-Lite when a use case does not need Flink.
Running low-code scenarios with GraalVM native image
We compiled Nussknacker scenarios to a GraalVM native image. The container shrank to 111 MB and used under 10 MB of memory, against 433 MiB on an untuned JVM.
Nussknacker compared with similar open-source tools
We compare Nussknacker with NiFi, Node-RED, BPMN engines and Drools by logic style, business versus integration logic, scalability and fit for the domain.
Apache Ignite in Nussknacker environment
See how Apache Ignite with Nussknacker enables fast, distributed caching and data enrichment for real-time streaming use cases.
Event streams always need Flink, or …not really?
Flink is powerful but complex. Discover how Nussknacker delivers stream processing value without the operational burden of heavy engines like Flink.
Nussknacker 1.1
Released in December 2021, Nussknacker 1.1 brings support for Flink 1.14, an integration with Apache Ignite and simpler configuration.
No posts match.