Miyamoto AI
An operator that lives inside the platform. Ask it a question about your estate and it queries the real data; ask it to do something and it calls the real service. Not a chatbot bolted onto a help page.
How it works
Ask in plain language
Type or speak a request — a status question, a mission order, a report. No query syntax, no menu path to memorise, in English or Arabic.
It calls real services
Miyamoto is wired to the platform tools: telemetry queries, asset lookups, mission dispatch, report generation. Answers come from your live data, not from a canned script.
Results render in place
Responses come back as working UI — a dashboard panel, an alert feed, a mission timeline, or a downloadable PDF — rendered directly in the conversation.
Capabilities
Natural language command
Skip the menu tree. Ask Miyamoto to navigate to a specific turbine or dispatch a perimeter sweep, and it executes against the live platform rather than describing how you could.
Telemetry queries
Miyamoto reads the InfluxDB time-series streams behind your sensors, surfacing thermal anomalies and degradation trends that would otherwise need a manual dashboard trawl.
Generative UI
Answers arrive as interactive components — KPI cards, alert feeds, mission timelines, dashboard panels — assembled for the question you actually asked.
PDF reporting
Ask for an incident summary or a monthly inspection report and Miyamoto generates it server-side as a formatted PDF, charts included, ready to send onward.
Guided walkthroughs
New operators can ask to be shown around. Miyamoto runs a step-by-step spotlight tour of the interface, highlighting each control as it explains it.
Organisation-aware
Every response is scoped to your organisation and the asking user role. Miyamoto cannot surface data the person in front of it is not entitled to see.
Under the hood
- Model
- Gemini 2.5 Flash via the Vercel AI SDK
- Tools
- Telemetry, assets, missions, reporting, and navigation services
- Output
- Streamed text, generative React UI, and pdfmake PDF documents
- Data sources
- InfluxDB time series and Firestore asset metadata
- Access control
- Scoped by organisation and user role on every call
- Languages
- English and Arabic
Frequently asked questions
What is Miyamoto AI?
Miyamoto AI is the conversational operator built into Katana IoTwin. It answers questions about your assets and telemetry, dispatches drone missions, generates dashboards, and produces PDF reports — all from natural language, inside the platform where the data already lives.
How is it different from a support chatbot?
A support chatbot retrieves documentation. Miyamoto calls real platform services: it queries your live InfluxDB telemetry, reads your Firestore asset records, and triggers actual missions. The answers reflect the current state of your estate, not a help article.
Which model powers Miyamoto?
Miyamoto runs on Gemini 2.5 Flash through the Vercel AI SDK, with a tool layer that connects it to the platform services. The model handles language; the tools handle the data, which is what keeps responses grounded in your actual records.
Can Miyamoto see data it should not?
No. Every tool call is scoped to the asking user organisation and role, enforced server-side. Miyamoto operates within exactly the same access boundaries as the person talking to it.
What kind of reports can it produce?
Incident summaries, inspection reports, asset status rollups, and drone flight logs. Reports are generated server-side as formatted PDFs with charts included, so they can be filed or forwarded without reformatting.
Explore the rest of the platform
See it on your own site
Start a 30-day trial, or have KatanaTech operators run the capture and hand you a finished twin.