@kici-dev/agent 0.0.0 → 0.1.2

package/dist/metrics/prometheus.d.ts

@@ -0,0 +1,95 @@
+/**
+ * Total completed jobs.
+ * Labels:
+ * - status: success | failed | cancelled
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const jobsTotal: import("@opentelemetry/api").Counter<import("@opentelemetry/api").Attributes>;
+/**
+ * Currently running jobs.
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const jobsActive: import("@opentelemetry/api").UpDownCounter<import("@opentelemetry/api").Attributes>;
+/**
+ * Total completed steps.
+ * Labels:
+ * - status: success | failed | skipped
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const stepsTotal: import("@opentelemetry/api").Counter<import("@opentelemetry/api").Attributes>;
+/**
+ * Step execution duration in seconds.
+ * Advisory boundaries cover sub-second steps through 30-minute long-running steps.
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const stepDurationSeconds: import("@opentelemetry/api").Histogram<import("@opentelemetry/api").Attributes>;
+/**
+ * Git clone duration in seconds.
+ * Advisory boundaries cover fast shallow clones through large repo clones.
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const cloneDurationSeconds: import("@opentelemetry/api").Histogram<import("@opentelemetry/api").Attributes>;
+/**
+ * Total log bytes streamed back to orchestrator.
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const logBytesTotal: import("@opentelemetry/api").Counter<import("@opentelemetry/api").Attributes>;
+/**
+ * Rising-edge counter for log-streaming backpressure events.
+ *
+ * Incremented once each time a LogStreamer transitions from "normal" into
+ * a backpressured state. Not incremented again until the streamer
+ * recovers and trips again, so the series measures how often operators
+ * saw pressure — not how long it lasted. Pair with
+ * {@link logBackpressureActive} (current state) and
+ * {@link logLinesDroppedTotal} (cumulative loss) for a complete picture.
+ *
+ * Labels:
+ * - mode: `pause` (producer paused, no data loss) | `drop` (lines
+ *   discarded with a `[N lines dropped due to backpressure]` marker)
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const logBackpressureEventsTotal: import("@opentelemetry/api").Counter<import("@opentelemetry/api").Attributes>;
+/**
+ * Cumulative log lines dropped by the backpressure "drop" mode.
+ *
+ * Incremented in batches by the LogStreamer whenever it emits the
+ * `[N lines dropped due to backpressure]` marker. Never decremented.
+ * A non-zero rate on this series means operators are losing real log
+ * content and should either expand the streamer's buffer or shrink the
+ * producer.
+ *
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const logLinesDroppedTotal: import("@opentelemetry/api").Counter<import("@opentelemetry/api").Attributes>;
+/**
+ * Current backpressure state as a 0/1 up/down counter.
+ *
+ * Toggles to 1 when a LogStreamer enters backpressure and back to 0 on
+ * recovery. Exposed as an up/down counter (not an observable gauge) so
+ * the existing LogStreamer callbacks can update it imperatively from
+ * both the single-job and step-streamer wiring paths.
+ *
+ * Grafana treats this as a step function: the area under the curve is
+ * the total time spent backpressured. Labelled with `mode` so dashboards
+ * can distinguish pause-mode from drop-mode pressure.
+ *
+ * Labels:
+ * - mode: `pause` | `drop`
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const logBackpressureActive: import("@opentelemetry/api").UpDownCounter<import("@opentelemetry/api").Attributes>;
+/**
+ * Orchestrator WebSocket connection status.
+ * Use add(1) for connected, add(-1) for disconnected.
+ *
+ * Labels:
+ * - scaler: injected by orch-side AgentMetricsAggregator (Phase 5b); `stateful` for static agents, backend name (`container` / `firecracker` / `bare-metal`) for scaler-managed
+ */
+export declare const connectionStatus: import("@opentelemetry/api").UpDownCounter<import("@opentelemetry/api").Attributes>;
+//# sourceMappingURL=prometheus.d.ts.map

package/dist/routes/health.d.ts

@@ -0,0 +1,27 @@
+import { Hono } from 'hono';
+import { type MetricsRoutesDeps } from '@kici-dev/shared';
+/**
+ * Health status provided by the agent runtime.
+ */
+interface HealthStatus {
+    agentId: string;
+    connected: boolean;
+    activeJobs: number;
+}
+export interface HealthRoutesDeps extends MetricsRoutesDeps {
+    /** Function returning current agent health status */
+    getStatus: () => HealthStatus;
+}
+/**
+ * Create health, readiness, and metrics routes.
+ *
+ * Delegates liveness and readiness to the shared health route helper,
+ * adding agent-specific status info and a /metrics endpoint.
+ *
+ * - GET /health  - Liveness probe (always 200, includes agent status)
+ * - GET /ready   - Readiness probe (200 if connected, 503 if not)
+ * - GET /metrics - Prometheus metrics
+ */
+export declare function createHealthRoutes(deps: HealthRoutesDeps): Hono;
+export {};
+//# sourceMappingURL=health.d.ts.map

package/dist/server.d.ts

@@ -0,0 +1,20 @@
+/**
+ * Agent entry point.
+ *
+ * Startup sequence:
+ * 1. Load config
+ * 2. Create logger
+ * 3. Create JobRunner with send/sendDirect callbacks
+ * 4. Create OrchestratorClient with dispatch and cancel handlers
+ * 5. Add WS log transport (if not scaler-managed)
+ * 6. Connect OrchestratorClient
+ * 7. Create Hono app with health routes
+ * 8. Start HTTP server
+ * 9. Register signal handlers
+ *
+ * Graceful shutdown:
+ * - SIGTERM/SIGINT: 10s grace for running jobs, then force-kill
+ * - SIGUSR1: Drain mode (stop accepting, finish current, exit)
+ */
+export {};
+//# sourceMappingURL=server.d.ts.map