Nussknacker ads

A #streamhouse isn't a faster lakehouse. It's about keeping the current state of your business live and available to applications, analytics and AI agents. Two of its core capabilities are transform and serve. Here's how Nussknacker, #ApacheFlink and #ApacheFluss cover them: - Apache Flink runs the streaming transform. - Apache Fluss keeps the result as a live table: a stream of changes and a key-value lookup in one, open to everyone. - Nussknacker lets domain experts build both sides visually: the streaming transform on Flink and the serving API on top of Fluss, exposed through #OpenAPI or as an #MCP tool for #AIAgents. In Nussknacker it comes down to one component, the Fluss Key-Value Store. One scenario writes the latest state under a key, for example the last hour of activity per customer, device or SIM card. Any other scenario reads it back with a single lookup, with nothing recomputed. The logic is defined once, by the people who understand the domain. Every system and every agent gets the same up-to-date answer. #StreamProcessing

We’ve expanded AI support in #Nussknacker with #DecisionModels, starting with #Jev and #Clef, which share the System One protocol. Support for more models through Ollama is coming next. Instead of generating free-form text, a decision model evaluates the answers you define. It returns a typed assessment: a choice with probabilities, a score, or a yes/no probability. That makes it a natural complement to the decision tables already in Nu: - The model interprets unstructured input. - The decision table combines that assessment with business facts and explicit rules to choose the next action. The video shows a customer replying to a SIM replacement confirmation: “Yes, that’s exactly the problem: I did NOT request this SIM replacement.” A rule looking for “YES” misses the meaning. A decision model evaluates whether the reply expresses confirmation, denial, or insufficient information. The routing policy stays visible and editable in the scenario. With Nu’s AI Assistant, you describe the change in plain language and it configures the component within your existing flow. #DecisionAI #ApacheFlink

From a news article to a running #ApacheFlink outage detector - with one prompt. Our partner in Norway read about a telco outage and asked whether #Nussknacker could help detect something similar. I asked for the link and gave it to my #AIAgent. In this video, that prompt becomes a working demo: a Kafka input stream, synthetic data generator, detection logic, explanatory notes, and tests - ready to inspect in the visual designer and deploy. Watch the workflow in this one-minute video 👇 #StreamProcessing #ApacheKafka

One vessel goes dark. Another slows to 0.7 knots above a pipeline. A third stops in the middle of the sea. Thousands of AIS messages a minute, and every one of them could be nothing — or something. In our new 2-minute demo, live AIS data from the public aisstream.io WebSocket flows into Nussknacker, gets cleaned up, lands in Kafka, and feeds detection scenarios: going dark, loitering, anchor dragging near critical infrastructure. Then we add a brand new rule — sanctioned vessel crossing a line on the map — deploy it, and the first alert fires almost immediately. No Flink code involved. And we are not even maritime experts. What would you build with it? #ApacheFlink #StreamProcessing #Kafka #AIS #LowCode

Still think remote work is as productive as the office? 🤔

Build an #MCP server from your real-time #Kafka data - no integration code. An AI agent answering questions about your business is only as useful as the data behind it. And most of that data reaches the agent stale - pulled from a warehouse, minutes or hours old. What if the agent could read state computed from a live stream, seconds ago? In this walkthrough I build exactly that with #Nussknacker: a purchase stream turned into context an AI agent can query on demand through an MCP server. The request-response scenarios we've had for a while already speak #OpenAPI - and now any of them can join Nussknacker's built-in MCP server as a tool an agent discovers and calls on its own. #ApacheFlink #AIAgents #StreamProcessing