Performance Standards

In an autonomous agent system, latency is the enemy of reasoning. We enforce strict timing targets across all surfaces.

⚡ Latency Targets

Layer	Max Latency	Monitoring Tool
A2A Handshake	< 150ms	NATS Monitor
Tool Execution	< 500ms	HookSystem Logs
Model Inference	< 2.0s	Antigravity Metrics
Web UI Load	< 800ms	Lighthouse / Pages

Optimization Mandates

1. Neural-Bus Local Cache: Never fetch metadata twice; use the Redis-backed cache for A2A discovery.
2. Context Compaction: Always compress prompts to the minimum necessary tokens to reduce inference time.
3. Rust Engine Preference: Use the Rust-based execution engine for heavy file operations or search.

---

⚖️ Performance Audits

Environments that consistently exceed 1.5s total loop latency are automatically flagged for hardware or network upgrade.