Skip to main content

S.I.R.I.U.S. (Smart Intelligent Responsive Interface for Universal Support)

Privacy-first, local-first AI assistant that automates tasks, email, and workflows across tools. Stable Node.js + Ollama today; Next‑Gen TypeScript + Mistral + n8n (EU‑compliant) in progress.
Source Code

Summary

S.I.R.I.U.S. is a privacy-first, offline-capable AI assistant for automating tasks, email, and calendars—designed for self-hosting, data control, and extensibility. It runs fully local via Ollama today, with a TypeScript “Next” rewrite adding Mistral (EU) and n8n’s 200+ integrations.

Challenges Addressed

  • Privacy/data sovereignty: Keep AI inference and user data local by default
  • Tool fragmentation: Unify Notion, Asana, Google Workspace, Jira, Trello, and more
  • Resilience: Operate offline; no hard cloud dependencies

Core Capabilities

  • Automation & integrations: Notion, Asana, Google Calendar/Gmail, Jira, Trello, n8n
  • Natural language commands with context and memory
  • Local-first AI via Ollama; optional vector memory with Pinecone
  • Learning: Circadian rhythm analysis, semi‑supervised pattern recognition, anomaly signals
  • Multi-interface: Web, React Native mobile app, browser extension

Architecture (high level)

  • Current (stable)
    • Runtime: Node.js services, WebSocket real‑time events, workers for heavy tasks
    • AI: Ollama local inference; Matryoshka quantization optimization
    • Data: Local JSON memory; optional Pinecone for vector recall
    • Clients: Web UI, React Native app, browser extension
    • Integrations: Modular services (Notion/Asana/Google/Jira/Trello/n8n)
  • Next (TypeScript rewrite)
    • TypeScript + stricter types, maintainable modules
    • AI: Mistral (EU‑centric) for chat/embeddings
    • Automations: n8n workflows + webhook processing (200+ integrations)
    • Focus: EU compliance, observability, testability

Role & Contributions

  • Led system architecture and integration strategy (functional/OOP where appropriate; SOLID, DRY, SRP)
  • Built service adapters, real‑time messaging, memory layer, and workers
  • Implemented local‑first AI (Ollama) and EU‑aligned model path (Mistral) in Next
  • Authored developer docs, onboarding scripts, and comprehensive tests

Outcomes

  • Private by default: No external data sharing; offline operation
  • Reduced context switching via unified command and integration layer
  • Extensible: Clear adapter contracts; easy to add services/workflows
  • Next‑gen: Type-safe TS codebase + n8n expands integrations to 200+ with minimal glue

Accessibility & UX

  • Accessibility-first (ARIA), responsive/adaptive design, minimalistic UI, classic modular structure, cross‑browser support

Status

  • Shipping: Local-first Node.js + Ollama, core integrations, tests
  • In progress: sirius-next TypeScript rewrite with Mistral + n8n, circadian/productivity analytics APIs