System Overview

Aura is a polyglot microservices architecture with a rich React 19 + Vite 6 frontend. Seven AI agents communicate via the SymphonyBus (a custom EventTarget-based event bus) with commission chaining for multi-step orchestration.

Services

Service	Stack	Port	Role
Frontend	React 19 + Vite 6 + TypeScript	3000	SPA, agent orchestration, frame extraction
API Proxy	Express	3005	Gemini key isolation, rate limiting, usage metering
Vector Search	FastAPI + ChromaDB	3001	Semantic search with adaptive chunking
Graph Knowledge	Express + SQLite	3004	Concept graph traversal, learning paths
Media Pipeline	Express + FFmpeg + WebSocket	3002/3003	Cloud-side frame extraction, transcription
CLIP Embeddings	FastAPI + CLIP ViT-B/32	3006	Multimodal visual search

All backend services are optional. The frontend degrades gracefully to browser-local alternatives — in-browser frame extraction, hash-based embeddings, IndexedDB persistence. The full stack is orchestrated via docker-compose.yml.

Defense-in-depth AI safety

Every LLM interaction passes through layered safety machinery:

Zod schema validation on all LLM function calls (conductor-schemas.ts)
Critic agent — adversarial quality gate scoring relevance, factual consistency, quality
Valhalla 3-layer sandbox — static analysis, import blocking, execution timeout
Plugin sandboxed execution — isolated scope + SHA-256 integrity verification

Frontend pillars

Lens Laboratory — the analysis workspace where Insights from the Virtuosos land
Creator Studio — structured, AI-assisted content generation
Agent Studio — build custom Virtuosos; install third-party ones from the marketplace
Valhalla Gateway — sandboxed automation of external creative software
Course View — adaptive learning modules generated from any video

Quality figures

438 tests across 20 files (Vitest + backend pytest/node test suites)
Production build in ~29s
Streaming responses reduce perceived latency by 60–80%