feat: add agent runtime reliability report

README.ko.md

@@ -58,12 +58,17 @@ Recommended demo flow:
 ```bash
 poetry run inferedgelab demo-evidence-summary
 poetry run inferedgelab demo-evidence-summary --format json
+poetry run inferedgelab agent-runtime-report \
+  --orchestration-summary examples/agent_runtime/agent_3_orchestration_summary.json \
+  --guard-analysis examples/agent_runtime/aiguard_runtime_guard_analysis.json
 poetry run inferedgelab export-demo-evidence --output reports/studio_demo_evidence.md
 ```
 
 Load Demo Evidence는 bundled ONNX Runtime CPU / TensorRT Jetson result fixture를 불러오고, Run / Import / Jetson Helper는 기존 CLI/API workflow를 local UI로 확장하는 보조 기능입니다.
 Studio evidence와 jobs는 in-memory이며 local server process가 재시작되면 초기화됩니다.
 
+`agent-runtime-report`는 Orchestrator scheduling evidence와 AIGuard runtime reliability `guard_analysis`를 Lab-owned agent deployment decision context로 묶는 additive report path입니다. 기존 Runtime result나 compare contract는 변경하지 않습니다.
+
 ## 이 레포의 역할
 
 - Runtime benchmark/result JSON을 읽어 compare/report를 생성합니다.

README.md

@@ -113,6 +113,9 @@ poetry run inferedgelab demo-evidence-summary
 poetry run inferedgelab demo-evidence-summary --format json
 poetry run inferedgelab portfolio-demo-check
 poetry run inferedgelab core4-conformance-check
+poetry run inferedgelab agent-runtime-report \
+  --orchestration-summary examples/agent_runtime/agent_3_orchestration_summary.json \
+  --guard-analysis examples/agent_runtime/aiguard_runtime_guard_analysis.json
 poetry run inferedgelab export-demo-evidence --output reports/studio_demo_evidence.md
 ```
 
@@ -122,6 +125,9 @@ It validates the committed Studio fixtures, expected README/PPT metrics, portfol
 It validates the bundled Forge manifest/metadata fixture, Runtime result JSON, Lab compare/deployment decision surface, and AIGuard `guard_analysis` evidence without mutating existing schemas.
 The Lab decision surface now also exposes `policy_version`, `triggered_rules`, and `policy_summary` so reviewers can see which local policy rules produced deploy/review/block/unknown outcomes.
 
+`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.
+
 ![InferEdge Local Studio demo evidence](assets/images/local-studio-demo-evidence.png)
 
 Verified demo fixture values:

docs/portfolio/agent_runtime_reliability_report.md

@@ -0,0 +1,77 @@
+# Agent Runtime Reliability Report
+
+## Scope
+
+This report is the first Lab-side bundle view for the reliable edge agent
+runtime path.
+
+It connects:
+
+- Forge `agent_manifest.json` metadata
+- Runtime `result.agent` metadata
+- Orchestrator `inferedge-orchestration-summary-v1`
+- AIGuard `inferedge-aiguard-diagnosis-v1`
+- Lab-owned agent deployment decision context
+
+This is a local-first report path. It is not a production cloud orchestration
+dashboard and does not add DB/queue/auth/billing behavior.
+
+## Demo Bundle
+
+Committed lightweight fixtures:
+
+- `examples/agent_runtime/agent_3_orchestration_summary.json`
+- `examples/agent_runtime/aiguard_runtime_guard_analysis.json`
+
+Generate a Markdown report:
+
+```bash
+poetry run inferedgelab agent-runtime-report \
+  --orchestration-summary examples/agent_runtime/agent_3_orchestration_summary.json \
+  --guard-analysis examples/agent_runtime/aiguard_runtime_guard_analysis.json \
+  --format markdown \
+  --output reports/agent_runtime_reliability_report.md
+```
+
+## Evidence Summary
+
+| Evidence | Value |
+|---|---:|
+| executed_count | 10 |
+| dropped_count | 14 |
+| deadline_missed_count | 1 |
+| fallback_count | 14 |
+| drop_rate | 0.583333 |
+| fallback_rate | 0.583333 |
+| deadline_miss_rate | 0.1 |
+| queue_backlog_policy_decision_count | 1 |
+
+## Lab Decision Context
+
+Expected decision:
+
+```text
+blocked
+```
+
+Primary reason:
+
+```text
+Agent runtime reliability evidence indicates blocked deployment risk.
+```
+
+Triggered rules:
+
+- `guard_blocked_runtime_block`
+- `drop_rate_block`
+- `fallback_rate_block`
+- `deadline_miss_review`
+- `queue_backlog_review`
+
+## Boundary
+
+- Orchestrator records scheduling and policy 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
+  Runtime result, compare output, or classic deployment decision contracts.

examples/agent_runtime/agent_3_orchestration_summary.json

@@ -0,0 +1,69 @@
+{
+  "schema_version": "inferedge-orchestration-summary-v1",
+  "agent_runtime_summary": {
+    "schema_version": "inferedge-orchestration-summary-v1",
+    "source_contracts": {
+      "forge_agent_manifest": "inferedge-agent-manifest-v1",
+      "runtime_agent_result": "inferedge-runtime-agent-task-v1"
+    },
+    "agents": {
+      "safety_monitor_agent": {
+        "agent_id": "safety_monitor_agent",
+        "agent_type": "safety",
+        "priority": 100,
+        "latency_budget_ms": 20.0,
+        "fallback_policy": "protect",
+        "task_id": "task_safety_monitor_agent"
+      },
+      "vision_agent": {
+        "agent_id": "vision_agent",
+        "agent_type": "vision",
+        "priority": 90,
+        "latency_budget_ms": 33.0,
+        "fallback_policy": "drop_stale",
+        "task_id": "task_vision_agent"
+      },
+      "voice_command_agent": {
+        "agent_id": "voice_command_agent",
+        "agent_type": "voice",
+        "priority": 50,
+        "latency_budget_ms": 120.0,
+        "fallback_policy": "defer",
+        "task_id": "task_voice_command_agent"
+      }
+    },
+    "totals": {
+      "executed_count": 10,
+      "dropped_count": 14,
+      "deadline_missed_count": 1,
+      "fallback_count": 14,
+      "policy_decision_count": 14,
+      "overload_event_count": 14
+    }
+  },
+  "policy_decision_log": [
+    {
+      "agent_id": "vision_agent",
+      "task_id": "task_vision_agent",
+      "decision": "load_shedding",
+      "reason": "queue_backlog_threshold_exceeded",
+      "fallback_used": true,
+      "protected_agent_id": "safety_monitor_agent"
+    }
+  ],
+  "drop_events": [
+    {
+      "agent_id": "vision_agent",
+      "task_id": "task_vision_agent",
+      "reason": "load_shedding_backlog_threshold_exceeded"
+    }
+  ],
+  "overload_events": [
+    {
+      "agent_id": "vision_agent",
+      "task_id": "task_vision_agent",
+      "fallback_used": true,
+      "reason": "queue_backlog_threshold_exceeded"
+    }
+  ]
+}

examples/agent_runtime/aiguard_runtime_guard_analysis.json

@@ -0,0 +1,57 @@
+{
+  "schema_version": "inferedge-aiguard-diagnosis-v1",
+  "source": {
+    "orchestration_summary_schema_version": "inferedge-orchestration-summary-v1"
+  },
+  "guard_verdict": "blocked",
+  "severity": "high",
+  "confidence": 0.88,
+  "primary_reason": "drop_rate indicates runtime reliability risk under orchestrated multi-agent load.",
+  "evidence": [
+    {
+      "type": "excessive_drop_rate",
+      "metric_name": "drop_rate",
+      "observed_value": 0.5833333333333334,
+      "baseline_value": null,
+      "threshold": 0.2,
+      "delta": null,
+      "delta_pct": null,
+      "increase_factor": null,
+      "severity": "high",
+      "status": "failed",
+      "explanation": "Drop rate crossed the configured review threshold under synthetic 3-agent load.",
+      "why_it_matters": "High drop rate can make camera or command workloads stale even if selected high-priority tasks are protected.",
+      "suspected_causes": [
+        "queue_backlog",
+        "overload_load_shedding",
+        "producer_rate_exceeds_runtime_capacity"
+      ],
+      "recommendation": "Tune target FPS, queue size, drop policy, or fallback policy for affected agents.",
+      "raw_context": {
+        "executed_count": 10,
+        "dropped_count": 14
+      }
+    }
+  ],
+  "suspected_causes": [
+    "queue_backlog",
+    "overload_load_shedding",
+    "producer_rate_exceeds_runtime_capacity"
+  ],
+  "recommendations": [
+    "Tune target FPS, queue size, drop policy, or fallback policy for affected agents."
+  ],
+  "thresholds": {
+    "drop_rate_review": 0.2,
+    "drop_rate_blocked": 0.5
+  },
+  "baseline_summary": {},
+  "candidate_summary": {
+    "runtime_reliability": {
+      "drop_rate": 0.5833333333333334,
+      "fallback_rate": 0.5833333333333334,
+      "deadline_miss_rate": 0.1
+    }
+  },
+  "created_at": "2026-05-17T00:00:00Z"
+}

inferedgelab/cli.py

@@ -16,6 +16,7 @@
 from inferedgelab.commands.demo_evidence import export_demo_evidence_cmd
 from inferedgelab.commands.demo_evidence import portfolio_demo_check_cmd
 from inferedgelab.commands.core4_conformance import core4_conformance_check_cmd
+from inferedgelab.commands.agent_runtime_report import agent_runtime_report_cmd
 from inferedgelab.commands.list_results import list_results_cmd
 from inferedgelab.commands.history_report import history_report_cmd
 from inferedgelab.commands.serve import serve_cmd
@@ -52,6 +53,9 @@ def version_cmd() -> None:
 app.command("core4-conformance-check", help="Validate Forge/Runtime/Lab/AIGuard contract conformance")(
     core4_conformance_check_cmd
 )
+app.command("agent-runtime-report", help="Generate Agent Runtime Reliability report from Orchestrator/AIGuard evidence")(
+    agent_runtime_report_cmd
+)
 app.command("list-results", help="List recent structured benchmark results")(list_results_cmd)
 app.command("history-report", help="Generate HTML history report from structured benchmark results")(history_report_cmd)
 app.command("serve", help="Run InferEdgeLab FastAPI server")(serve_cmd)

inferedgelab/commands/agent_runtime_report.py

@@ -0,0 +1,70 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import typer
+from rich import print as rprint
+
+from inferedgelab.services.agent_runtime_report import (
+    agent_runtime_reliability_json,
+    build_agent_runtime_reliability_markdown,
+    load_agent_runtime_reliability_bundle,
+)
+
+
+def agent_runtime_report_cmd(
+    orchestration_summary: str = typer.Option(
+        ...,
+        "--orchestration-summary",
+        help="Path to InferEdgeOrchestrator orchestration_summary JSON",
+    ),
+    guard_analysis: str = typer.Option(
+        "",
+        "--guard-analysis",
+        help="Optional AIGuard runtime reliability guard_analysis JSON",
+    ),
+    format: str = typer.Option("text", "--format", "-f", help="text/json/markdown"),
+    output: str = typer.Option("", "--output", "-o", help="Optional output path"),
+) -> None:
+    report = load_agent_runtime_reliability_bundle(
+        orchestration_summary_path=orchestration_summary,
+        guard_analysis_path=guard_analysis or None,
+    )
+    normalized_format = format.strip().lower()
+    if normalized_format == "json":
+        text = agent_runtime_reliability_json(report)
+    elif normalized_format in {"markdown", "md"}:
+        text = build_agent_runtime_reliability_markdown(report)
+    elif normalized_format == "text":
+        text = _text_summary(report)
+    else:
+        raise typer.BadParameter("--format must be one of: text, json, markdown")
+
+    if output:
+        path = Path(output)
+        path.parent.mkdir(parents=True, exist_ok=True)
+        path.write_text(text, encoding="utf-8")
+        rprint(f"[green]Saved[/green]: {path}")
+    else:
+        print(text, end="")
+
+
+def _text_summary(report: dict) -> str:
+    metrics = report["agent_runtime_summary"]["metrics"]
+    decision = report["agent_deployment_decision"]
+    guard = report["guard_summary"]
+    lines = [
+        "InferEdge Agent Runtime Reliability Report",
+        f"schema_version: {report['schema_version']}",
+        f"decision: {decision['decision']}",
+        f"policy_version: {decision['policy_version']}",
+        f"reason: {decision['reason']}",
+        f"guard_verdict: {guard.get('guard_verdict')}",
+        f"drop_rate: {metrics['drop_rate']:.6g}",
+        f"fallback_rate: {metrics['fallback_rate']:.6g}",
+        f"deadline_miss_rate: {metrics['deadline_miss_rate']:.6g}",
+        "triggered_rules:",
+    ]
+    lines.extend(f"- {rule}" for rule in decision["triggered_rules"])
+    lines.append("")
+    return "\n".join(lines)