The KNX AI Web Dashboard is now the official UI for live analysis of your KNX AI nodes: summary, anomalies, flow map, and chat.
This page keeps the historical name KNX-AI-Sidebar for backward compatibility.
What it is for
- Quickly browse the summary produced by a selected
knxUltimateAInode. - Inspect detected anomalies and spot suspicious behaviour (spam, flapping, bursts).
- Ask questions in the chat (answers are rendered in Markdown) to speed up troubleshooting.
How to use it
- Open it from the KNX AI node editor using Open Web Page.
- Alternatively open
/knxUltimateAI/sidebar/page(optionally with?nodeId=<id>). - Select the
knxUltimateAInode from the dropdown. - Use Auto or manual refresh and ask questions in chat.
Make the AI smarter (extra context)
In the selected knxUltimateAI node you can include additional context in the LLM prompt:
- Flow inventory: lets the AI “see” which KNX Ultimate nodes (and gateways) are present in your flows, so it can relate telegrams to your logic.
- Documentation snippets: adds relevant excerpts from built-in help/README/examples (and
docs/wikiwhen available) so the AI can explain node behaviour and suggest the right configuration.
Example usage (human scenarios)
- Loop / duplicated telegrams: ask what could generate repeated writes on the same GA and how to isolate the source.
- Noisy GA: ask why a GA is the top talker and which devices/nodes are involved.
- Unexpected behaviour after deploy: ask what changed in the last minutes and which patterns appeared.
- Routing bridge issues: ask how to filter/rewrite routed telegrams to avoid storms or feedback loops.
Example questions to paste in chat
- “Why is
2/4/2so active? What are the most likely causes?” - “Do you see any loop pattern between two group addresses?”
- “Which physical sources are writing to
x/y/zand how often?” - “What should I filter in Router Filter to stop the spam but keep normal traffic?”
Requirements
- At least one
knxUltimateAInode in your flows. - The selected
knxUltimateAInode must be connected to aknxUltimate-configgateway. - To use chat LLM answers: enable LLM in the
knxUltimateAInode and configure its credentials (API key) there.