Helm

At its core, Helm is personal cognition management. Sonar AI chat is the single entry point that connects tasks, notes, the knowledge base and family affairs — one message can check tasks, search notes or query the knowledge base, with the same data flowing across mobile and web.

Retrieval runs on a self-built RAG pipeline: documents are vectorized and indexed, and semantic search feeds a knowledge Q&A layer powered by mixed LLM inference. n8n pipelines take care of data collection and distribution, while a multi-model gateway balances quotas across providers.

Mobile is a first-class citizen rather than an appendix: check-ins, quick notes and family matters — baby growth logs, shopping lists — are captured in an Android app (packaged with Capacitor) and synced with the web in real time. Core screens get dedicated phone layouts, so most operations can be done one-handed on a commute.

Everything from design to operations is built by one person: Vue 3 + Capacitor + FastAPI + PostgreSQL, self-hosted on a cloud server with CI deployment, monitoring and alerting.

Vue3 · FastAPI · PostgreSQL · RAG · LLM · n8n · Capacitor