Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -128,7 +128,7 @@ The Lab decision surface now also exposes `policy_version`, `triggered_rules`, a
`agent-runtime-report` is an additive reliable edge agent runtime report path.
It bundles Orchestrator scheduling evidence and AIGuard runtime reliability `guard_analysis` into a Lab-owned agent deployment decision context without changing existing Runtime result or compare contracts.
The current bundled evidence is a synthetic/dummy sustained high-load 3-agent scenario.
The report preserves sustained queue-depth, policy decision reason, and `sustained_overload_risk` evidence as local-first deployment review context.
The report preserves sustained queue-depth, worker health, runtime event summary/timeline, policy decision reason, and `sustained_overload_risk` evidence as local-first deployment review context.

![InferEdge Local Studio demo evidence](assets/images/local-studio-demo-evidence.png)

Expand Down
18 changes: 18 additions & 0 deletions docs/portfolio/agent_runtime_reliability_report.md
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,23 @@ AIGuard `guard_analysis` also includes `sustained_overload_risk`, which Lab
preserves as report evidence and reflects in the agent deployment decision
context.

The report also preserves the Orchestrator operation-health fields added for
runtime operation review:

- `queue_state_summary` for queue pressure, max backlog, final queue depth, and
overload threshold.
- `worker_health_snapshot` for healthy/constrained/degraded worker state,
executed/drop/deadline/fallback counts, and latency context.
- `runtime_event_summary` for event type counts.
- `runtime_event_timeline` sample rows for queue snapshots, policy decisions,
drops, and execution outcomes.

These fields make the report path explicit:

```text
Orchestrator operation evidence -> AIGuard reliability explanation -> Lab-owned deployment risk context
```

## Lab Decision Context

Expected decision:
Expand All @@ -83,6 +100,7 @@ Triggered rules:
## Boundary

- Orchestrator records scheduling and policy evidence.
- Orchestrator operation-health fields are displayed as local runtime evidence.
- AIGuard explains runtime reliability risk.
- Lab remains the final deployment decision owner.
- This report is an additive agent-runtime path and does not change existing
Expand Down
118 changes: 118 additions & 0 deletions examples/agent_runtime/agent_3_orchestration_summary.json
Original file line number Diff line number Diff line change
Expand Up @@ -97,6 +97,124 @@
"protected_agent_id": "safety_monitor_agent"
}
],
"queue_state_summary": {
"schema_version": "inferedge-orchestrator-queue-state-v1",
"sample_count": 1,
"overload_backlog_threshold": 3,
"max_total_queue_depth": 6,
"average_total_queue_depth": 6.0,
"final_queue_depth": {
"vision_agent": 4,
"voice_command_agent": 2,
"safety_monitor_agent": 0
},
"max_queue_depth_by_task": {
"vision_agent": 4,
"voice_command_agent": 2,
"safety_monitor_agent": 0
},
"queue_pressure_state": "overloaded"
},
"worker_health_snapshot": {
"schema_version": "inferedge-orchestrator-worker-health-v1",
"workers": {
"safety_monitor_agent": {
"task": "safety_monitor_agent",
"agent_id": "safety_monitor_agent",
"task_id": "task_safety_monitor_agent",
"agent_type": "safety",
"scheduled_priority": 100,
"latency_budget_ms": 20.0,
"worker": "dummy",
"health_state": "healthy",
"executed_count": 2,
"dropped_count": 0,
"deadline_missed_count": 0,
"fallback_count": 0,
"mean_latency_ms": 6.0,
"p95_latency_ms": 8.0,
"max_queue_backlog": 0,
"queue_size": 1,
"queue_pressure_ratio": 0.0
},
"vision_agent": {
"task": "vision_agent",
"agent_id": "vision_agent",
"task_id": "task_vision_agent",
"agent_type": "vision",
"scheduled_priority": 90,
"latency_budget_ms": 33.0,
"worker": "dummy",
"health_state": "degraded",
"executed_count": 8,
"dropped_count": 14,
"deadline_missed_count": 1,
"fallback_count": 14,
"mean_latency_ms": 41.0,
"p95_latency_ms": 41.0,
"max_queue_backlog": 4,
"queue_size": 4,
"queue_pressure_ratio": 1.0
},
"voice_command_agent": {
"task": "voice_command_agent",
"agent_id": "voice_command_agent",
"task_id": "task_voice_command_agent",
"agent_type": "voice",
"scheduled_priority": 50,
"latency_budget_ms": 120.0,
"worker": "dummy",
"health_state": "constrained",
"executed_count": 0,
"dropped_count": 0,
"deadline_missed_count": 0,
"fallback_count": 0,
"mean_latency_ms": null,
"p95_latency_ms": null,
"max_queue_backlog": 2,
"queue_size": 2,
"queue_pressure_ratio": 1.0
}
}
},
"runtime_event_summary": {
"schema_version": "inferedge-orchestrator-runtime-event-summary-v1",
"event_count": 4,
"event_type_counts": {
"queue_snapshot": 1,
"policy_decision": 1,
"drop": 1,
"execution": 1
}
},
"runtime_event_timeline": [
{
"event_index": 0,
"event_type": "queue_snapshot",
"reason": "queue_depth_sampled"
},
{
"event_index": 1,
"event_type": "policy_decision",
"agent_id": "vision_agent",
"task_id": "task_vision_agent",
"reason": "queue_backlog_threshold_exceeded"
},
{
"event_index": 2,
"event_type": "drop",
"agent_id": "vision_agent",
"task_id": "task_vision_agent",
"reason": "load_shedding_backlog_threshold_exceeded"
},
{
"event_index": 3,
"event_type": "execution",
"agent_id": "vision_agent",
"task_id": "task_vision_agent",
"reason": "deadline_missed"
}
],
"drop_events": [
{
"agent_id": "vision_agent",
Expand Down
Loading
Loading