Operations
Telemetry & observability
src/lib/telemetry.ts provides structured logging with pluggable sinks:
export type LogLevel = 'debug' | 'info' | 'warn' | 'error'
export function addLogSink(sink: LogSink): void
// Domain-specific helpers:
logVirtuosoCommission(...) // a specialist was commissioned
logVirtuosoResult(...) // a commission completed
logConductorQuery(...) // user intent received
logValidationFailure(...) // a Zod schema rejected a function call
logSearchExecuted(...) // semantic search ran
logValhallaAnalysis(...) // sandbox safety analysis
startTimer() // { elapsed(): number } latency helperValhalla executions additionally log through valhalla-telemetry.ts, and the worker pool
reports queue depths and steal counts.
Task persistence
src/lib/task-persistence.ts checkpoints Symphony Bus task state into IndexedDB, so in-flight
work survives reloads and integrates with the
offline mutation queue.
Testing
438 tests across 20 files, spanning every advanced system:
| Layer | Tooling | Examples |
|---|---|---|
| Frontend units | Vitest | semantic-chunker, react-planner, valhalla-sandbox, knowledge-tracing, crdt-collaboration, plugin-marketplace, worker-pool, offline-sync, federated-learning, clip-embeddings |
| Conductor safety | Vitest | conductor-schemas, conductor-validator |
| Backend (Node) | node test runners | api-proxy, graph-knowledge, media-pipeline (+ WebSocket tests) |
| Backend (Python) | pytest | vector-search |
| End-to-end | Playwright | e2e/aura.spec.ts |
npm test # Vitest suite
npx playwright test # E2ECI pipeline
.github/workflows/ci.yml runs on pushes to main/develop and on PRs:
- Frontend tests — Node 22,
npm ci, Vitest, then a TypeScript type check - Vector Search tests — Python service tested with pytest
- Additional jobs cover the Node backends
Production build completes in ~29s (npm run build).