Agent Lifecycle

Agent lifecycle management defines what happens when an agent starts, runs, and shuts down. LODE encodes lifecycle configurations as verified binaries so that bootstrap hooks, teardown procedures, and health checks are deterministic and tamper-proof.

Schema: LifecycleConfig

lifecycle-config-v1.lode


{
  "schema": "lode",
  "version": 1,
  "name": "LifecycleConfig",
  "fields": [
    { "id": 1, "name": "agentName",      "type": "string",             "required": true },
    { "id": 2, "name": "bootstrapHooks", "type": "string[]",           "required": true },
    { "id": 3, "name": "teardownHooks",  "type": "string[]",           "required": false },
    { "id": 4, "name": "healthCheck",    "type": "bool",               "required": true },
    { "id": 5, "name": "triggers",       "type": "map<string,string>", "required": false },
    { "id": 6, "name": "timeout",        "type": "string",             "required": false },
    { "id": 7, "name": "retryPolicy",    "type": "string",             "required": false }
  ]
}

TypeScript Interface


interface LifecycleConfig {
  agentName: string;                       // Target agent
  bootstrapHooks: string[];                // Run in order at startup
  teardownHooks?: string[];                // Run in order at shutdown
  healthCheck: boolean;                    // Enable periodic health probes
  triggers?: Record<string, string>;       // Event → action mappings
  timeout?: string;                        // Max execution time ("300s")
  retryPolicy?: string;                    // Retry strategy ("exponential:3")
}

Example: Data Pipeline Agent

data-pipeline-lifecycle.json


{
  "agentName": "data-pipeline-agent",
  "bootstrapHooks": ["load-credentials", "init-connections", "verify-schemas"],
  "teardownHooks": ["flush-buffers", "close-connections", "archive-logs"],
  "healthCheck": true,
  "triggers": {
    "onStart": "notify:slack",
    "onFailure": "alert:pagerduty",
    "onComplete": "report:dashboard"
  },
  "timeout": "300s",
  "retryPolicy": "exponential:3"
}

Compilation


$ lodec compile data-pipeline-lifecycle.json -s lifecycle-config-v1.lode -o lifecycle/pipeline.vein
Compiled data-pipeline-lifecycle.json → lifecycle/pipeline.vein (224 bytes)
Fingerprint: c3d4e5f6...

Bootstrap Execution

The runtime loads and verifies the lifecycle config before executing hooks:


async function bootstrap(agentName: string): Promise<void> {
  const config = await loadVerifiedConfig(agentName);
 
  for (const hook of config.bootstrapHooks) {
    console.log(`[bootstrap] Running: ${hook}`);
    await executeHook(hook, { agentName, phase: "bootstrap" });
  }
 
  if (config.healthCheck) {
    startHealthProbe(agentName, config.timeout);
  }
 
  if (config.triggers?.onStart) {
    await fire(config.triggers.onStart, { agentName, status: "started" });
  }
}

Teardown

Teardown hooks run in order when the agent shuts down, ensuring clean resource release:


async function teardown(agentName: string): Promise<void> {
  const config = await loadVerifiedConfig(agentName);
 
  for (const hook of config.teardownHooks ?? []) {
    try {
      await executeHook(hook, { agentName, phase: "teardown" });
    } catch (err) {
      console.error(`[teardown] Hook "${hook}" failed: ${err}`);
      // Continue with remaining hooks
    }
  }
 
  if (config.triggers?.onComplete) {
    await fire(config.triggers.onComplete, { agentName, status: "stopped" });
  }
}

Retry Policy

The retryPolicy field encodes the retry strategy as a compact string:


function parseRetryPolicy(policy: string): { strategy: string; maxAttempts: number } {
  const [strategy, max] = policy.split(":");
  return { strategy, maxAttempts: parseInt(max, 10) };
}
 
parseRetryPolicy("exponential:3"); // → { strategy: "exponential", maxAttempts: 3 }
parseRetryPolicy("linear:5");      // → { strategy: "linear", maxAttempts: 5 }