Module load 

Graceful degradation under load — rate limiting, concurrency control, and load shedding.

This module provides the infrastructure for RavenClaws to degrade gracefully under load rather than failing hard. It includes:

• Rate limiting — Token bucket algorithm for per-endpoint and global rate limits
• Concurrency control — Semaphore-based limit on in-flight requests
• Load shedding — Metrics-based overload detection and 503 responses
• Backpressure — Queue depth tracking and admission control

Architecture

┌──────────────┐     ┌──────────────────┐     ┌────────────────┐
│  Incoming    │ ──→ │  LoadManager     │ ──→ │  Agent/Server  │
│  Requests    │     │  ┌────────────┐  │     │  Handler       │
│              │     │  │ RateLimiter│  │     │                │
│              │     │  ├────────────┤  │     │                │
│              │     │  │ Concurrency│  │     │                │
│              │     │  │  Limiter   │  │     │                │
│              │     │  ├────────────┤  │     │                │
│              │     │  │ LoadShedder│  │     │                │
└──────────────┘     └──┴────────────┴──┘     └────────────────┘

Stability

All public types are #[non_exhaustive] — new fields and variants may be added in minor releases.

Structs

ErrorTracker

Tracks error rates within a sliding time window for overload detection.

LoadConfig

Load management configuration (v1.1.0)

LoadManager

Central load management coordinator.

LoadMetrics

Snapshot of load manager metrics.

TokenBucket

Token bucket rate limiter.

Enums

Admission

Result of an admission check — whether a request is allowed through.

RequestOutcome

Outcome of a request processed through the load manager.