Skip to content

overview.md

Latest commit

 

History

History
81 lines (64 loc) · 3.2 KB

overview.md

File metadata and controls

81 lines (64 loc) · 3.2 KB
Raw

Requirements — Vision, Status Labels, Non-Functional & Out of Scope

Part of Trinity's requirements set. Index & write-path rule: requirements.md.

Vision

Trinity is autonomous agent orchestration and infrastructure — sovereign infrastructure for deploying, orchestrating, and governing fleets of autonomous AI agents on your own hardware.

Capability Description Implementation Status
Hierarchical Delegation Orchestrator-Worker with context quarantine ✅ Agent-to-Agent via MCP
Persistent Memory Virtual filesystems, memory folding ✅ Chat persistence, file browser
Agent Configuration Template system with CLAUDE.md and credential injection ✅ Templates with CLAUDE.md
Autonomous Operations Scheduling, monitoring, and fleet management ✅ Cron scheduling, health monitoring

Status Labels

  • ⏳ Not Started
  • 🚧 In Progress
  • ✅ Implemented
  • ❌ Removed

Non-Functional Requirements

Performance

  • Agent creation: < 30 seconds
  • Chat response: < 5 seconds for simple queries
  • UI responsiveness: < 100ms for interactions
  • WebSocket latency: < 500ms for status updates
  • Task list loading: < 1 second (PERF-001)

PERF-001: Task List Performance Optimization

  • Status: ✅ Implemented (2026-02-21)
  • Requirement ID: PERF-001
  • Priority: MEDIUM
  • Description: Optimize task loading on Agent Detail Tasks tab. Current implementation transfers full execution_log (100KB+ per execution) for list views that don't display it.
  • Key Features:
    • Lightweight ExecutionSummary response model excluding heavy fields
    • Optimized SQL query selecting only needed columns
    • Composite index on (agent_name, started_at DESC)
    • On-demand detail loading when user expands a task
  • Achieved Impact: 50-100x reduction in data transfer (10MB → 100KB)
  • Spec: docs/requirements/TASK_LIST_PERFORMANCE.md

Security

  • All credentials encrypted at rest (Redis)
  • HTTPS in production (Let's Encrypt)
  • Container isolation (network, filesystem)
  • Comprehensive audit logging via audit_log table (SEC-001)
  • Encryption key endpoint admin-only (C-001, 2026-03-09)
  • WebSocket authentication required before accept (C-002, 2026-03-09; hardened #178, 2026-03-27 — reject before websocket.accept(), removed first-message auth)
  • Internal API shared-secret auth (C-003, 2026-03-09)
  • Agent access control on chat/credential endpoints (M-004/M-006, 2026-03-09)
  • DOMPurify XSS protection on all v-html (H-005, 2026-03-09)

Scalability

  • Support 50+ concurrent agents
  • Multiple backend workers (uvicorn)
  • Stateless backend (Docker as truth)

Reliability

  • Agent state survives backend restart
  • SQLite persists across container recreation
  • Redis AOF for secret durability

Testing

  • Feature flows include Testing sections
  • Manual testing before marking complete
  • Automated tests for critical paths only

Out of Scope

  • Multi-tenant deployment (single org only)
  • Mobile application → See Section 27 (MOB-001) for PWA-based mobile admin
  • Fiat/Stripe payment integration (Nevermined handles crypto payments)
  • Agent marketplace