Spanda

Predictive Recovery

The Recovery Orchestrator integrates telemetry to detect early failure indicators and trigger preventative recovery.

Supported indicators

Indicator	Severity	Recommended action
Memory leak	warning	`restart_component`
Increasing latency	warning	`graceful_degradation`
CPU spike	warning	`restart_component`
Battery degradation	warning	`graceful_degradation`
Temperature increase	critical	`safe_shutdown`
Connectivity instability	warning	`reconnect`
Packet loss	warning	`switch_network`
Sensor drift	warning	`reinitialize`
Repeated retries	warning	`restart_component`
Crash frequency	critical	`restart_robot`
Health degradation	warning	`graceful_degradation`

API

let (indicators, trigger) = orchestrator.check_predictive(&registry, telemetry.as_ref());
if trigger {
    // preventative recovery recommended
}

REST

# Global indicators (no program context)
curl -s http://127.0.0.1:8080/v1/recovery/predictive

# With optional telemetry in POST body
curl -s -X POST http://127.0.0.1:8080/v1/recovery/predictive \
  -H 'Content-Type: application/json' \
  -d '{"file":"rover.sd","entity_id":"Rover"}'

Response envelope: { "version": "v1", "indicators": [...], "should_trigger_preventative": bool }.

Knowledge-base recommendation

curl -s -X POST http://127.0.0.1:8080/v1/recovery/recommend \
  -H 'Content-Type: application/json' \
  -d '{"failure":"gps_loss"}'

SDK (0.5.6+)

client.get_recovery_predictive(None)?;
client.recommend_recovery(&json!({ "failure": "gps_loss" }))?;

client.get_recovery_predictive()
client.recommend_recovery({"failure": "gps_loss"})

await client.getRecoveryPredictive();
await client.recommendRecovery({ failure: "gps_loss" });

Prognostics integration

Program-level prognostics declarations in Spanda source complement telemetry scanning. See self-healing.md and assurance prognostics.

Learning

Historical recovery outcomes feed the rule-based knowledge base (success rates, strategy effectiveness). No machine learning in Phase 2 — statistics and rules only.