Composable Pipelines
14 registered step types assembled in any order per rule, including unified LLM calls, scene analysis, interactive prompts, conditional branching, and wait/resume workflows. Each rule defines its own pipeline.
An on-premise system that watches for moments where a gentle reminder might help, without automating away the daily routines that give seniors agency.
Cognitive decline doesn't have to mean loss of independence. Most monitoring systems either do too little (basic motion alerts) or too much (full automation that removes the daily routines giving seniors agency). Cognitive Companion occupies a different position: it understands context through vision and language models, then delivers gentle reminders only when they're actually warranted.
Built for real families. Designed for multigenerational households where a senior lives with family members who want to help but can't watch 24/7. The system tracks who is where, what activities have happened, and applies rules written in natural language to decide when a nudge is appropriate.
Built for privacy. Every model runs on your hardware. Camera frames are processed locally and never leave your network. The system is designed from the ground up to operate entirely within your home.
Built to be extended. Each pipeline step, notification channel, and context filter is a self-contained plugin. Drop a Python file in the right directory and the system discovers it at startup. No fork required.