AI features
AI where it helps, not where it becomes marketing fog
IoTSentinel uses AI to improve comprehension first: plain-English alerts, context-aware explanation, and guided reasoning. The point is not “AI for everything.” The point is making a security project more understandable than many closed competitors.
Translate security jargon
The primary AI job here is not to sound futuristic. It is to rewrite alerts and investigations into language a normal person can act on.
Stay explicit about privacy
AI is useful only if the data path is understandable. Privacy mode reorders the stack toward local execution so visitors can see there is a real alternative to always shipping data outward.
Prefer local where possible
A local Ollama-backed path keeps the project aligned with its local-first goal instead of making AI synonymous with mandatory cloud dependence.
Fallback stack
OpenAI
gpt-4o-mini
Anthropic Claude
claude-haiku-4-5
Groq
llama-3.1-8b-instant
Google Gemini
gemini-2.5-flash
Ollama
gemma2:2b
Smart templates
rule-based
The honest caveat
AI explanations can make security output far easier to follow, but they do not remove the need for good signals, good defaults, and careful product design. That is why the project keeps the fallback chain explicit and treats AI as an interface layer over detections rather than as magic authority.