Datasets:

anon-ops
/

ops-lite

+{
+  "@context": {
+    "@language": "en",
+    "@vocab": "https://schema.org/",
+    "citeAs": "cr:citeAs",
+    "column": "cr:column",
+    "conformsTo": "dct:conformsTo",
+    "cr": "http://mlcommons.org/croissant/",
+    "rai": "http://mlcommons.org/croissant/RAI/",
+    "data": {
+      "@id": "cr:data",
+      "@type": "@json"
+    },
+    "dataType": {
+      "@id": "cr:dataType",
+      "@type": "@vocab"
+    },
+    "dct": "http://purl.org/dc/terms/",
+    "examples": {
+      "@id": "cr:examples",
+      "@type": "@json"
+    },
+    "extract": "cr:extract",
+    "field": "cr:field",
+    "fileProperty": "cr:fileProperty",
+    "fileObject": "cr:fileObject",
+    "fileSet": "cr:fileSet",
+    "format": "cr:format",
+    "includes": "cr:includes",
+    "isLiveDataset": "cr:isLiveDataset",
+    "jsonPath": "cr:jsonPath",
+    "key": "cr:key",
+    "md5": "cr:md5",
+    "parentField": "cr:parentField",
+    "path": "cr:path",
+    "recordSet": "cr:recordSet",
+    "references": "cr:references",
+    "regex": "cr:regex",
+    "repeated": "cr:repeated",
+    "replace": "cr:replace",
+    "sc": "https://schema.org/",
+    "separator": "cr:separator",
+    "source": "cr:source",
+    "subField": "cr:subField",
+    "transform": "cr:transform",
+    "samplingRate": "cr:samplingRate",
+    "equivalentProperty": "cr:equivalentProperty"
+  },
+  "@type": "sc:Dataset",
+  "conformsTo": "http://mlcommons.org/croissant/1.0",
+  "name": "ops-lite",
+  "description": "OpenRCA 2.0: a root-cause-analysis (RCA) benchmark for microservice systems with process-level ground truth. 457 instances, each a recorded fault injection on one of three open-source microservice testbeds (TrainTicket, Hotel Reservation from DeathStarBench, OpenTelemetry Demo). Each instance carries ten parquet telemetry tables (traces, metrics, logs; normal and abnormal windows), the raw injection record, a label giving each injected fault's role (required, optional, not effective), and the verified propagation graph along which the effects of the injected faults travelled between services.",
+  "url": "https://huggingface.co/datasets/anon-ops/ops-lite",
+  "version": "2.0.0",
+  "datePublished": "2026-09-26",
+  "license": "https://www.apache.org/licenses/LICENSE-2.0",
+  "citeAs": "@misc{openrca2_2026, title={OpenRCA 2.0: Process-Level Evaluation of LLM Agents for Microservice Root Cause Analysis}, year={2026}, howpublished={\\url{https://huggingface.co/datasets/anon-ops/ops-lite}}}",
+  "creator": {
+    "@type": "Organization",
+    "name": "anon-ops"
+  },
+  "keywords": [
+    "root-cause-analysis",
+    "microservices",
+    "observability",
+    "AIOps",
+    "chaos-engineering",
+    "causal-graph",
+    "llm-agents",
+    "benchmark"
+  ],
+  "isLiveDataset": false,
+  "distribution": [
+    {
+      "@type": "cr:FileObject",
+      "@id": "repo",
+      "name": "repo",
+      "description": "The Hugging Face git repository.",
+      "contentUrl": "https://huggingface.co/datasets/anon-ops/ops-lite/tree/refs%2Fconvert%2Fparquet",
+      "encodingFormat": "git+https",
+      "sha256": "https://github.com/mlcommons/croissant/issues/80"
+    },
+    {
+      "@type": "cr:FileSet",
+      "@id": "parquet-files-for-config-default",
+      "containedIn": {
+        "@id": "repo"
+      },
+      "encodingFormat": "application/x-parquet",
+      "includes": "default/*/*.parquet"
+    },
+    {
+      "@type": "cr:FileObject",
+      "@id": "manifest.jsonl",
+      "name": "manifest.jsonl",
+      "description": "Instance-level manifest: one JSON record per instance.",
+      "contentUrl": "https://huggingface.co/datasets/anon-ops/ops-lite/resolve/main/manifest.jsonl",
+      "encodingFormat": "application/jsonlines",
+      "sha256": "3356646fa120c6b0c16f81c6218ae90b9c8497e8f7ddafb2bfebf30828187be8"
+    }
+  ],
+  "recordSet": [
+    {
+      "@type": "cr:RecordSet",
+      "@id": "cases_index",
+      "name": "cases_index",
+      "description": "One record per instance.",
+      "field": [
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/name",
+          "name": "name",
+          "description": "Stable instance identifier; also the directory name under cases/.",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.name"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/system",
+          "name": "system",
+          "description": "Microservice testbed: ts (TrainTicket), hs (Hotel Reservation / DeathStarBench), or otel-demo (OpenTelemetry Demo).",
+          "dataType": "sc:Text",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.system"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/root_services",
+          "name": "root_services",
+          "description": "Services of the required faults: the root-cause set an agent must recover.",
+          "dataType": "sc:Text",
+          "repeated": true,
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.root_services"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/optional_services",
+          "name": "optional_services",
+          "description": "Services of optional faults (effect confined to their own resources, or masked by a co-injected fault); naming them is neither required nor penalised.",
+          "dataType": "sc:Text",
+          "repeated": true,
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.optional_services"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/chaos_types",
+          "name": "chaos_types",
+          "description": "Chaos Mesh fault types of the required faults.",
+          "dataType": "sc:Text",
+          "repeated": true,
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.chaos_types"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/n_injected_faults",
+          "name": "n_injected_faults",
+          "description": "Number of faults injected in the instance.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.n_injected_faults"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/n_required_faults",
+          "name": "n_required_faults",
+          "description": "Number of required faults.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.n_required_faults"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/process_applicable",
+          "name": "process_applicable",
+          "description": "Whether the verified propagation graph has at least one service-level edge; process-level metrics apply only then.",
+          "dataType": "sc:Boolean",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.process_applicable"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/n_service_edges",
+          "name": "n_service_edges",
+          "description": "Number of service-level edges in the verified propagation graph.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.n_service_edges"
+            }
+          }
+        },
+        {
+          "@type": "cr:Field",
+          "@id": "cases_index/longest_service_path",
+          "name": "longest_service_path",
+          "description": "Longest path, in hops, of the service-level verified propagation graph.",
+          "dataType": "sc:Integer",
+          "source": {
+            "fileObject": {
+              "@id": "manifest.jsonl"
+            },
+            "extract": {
+              "jsonPath": "$.longest_service_path"
+            }
+          }
+        }
+      ]
+    }
+  ],
+  "rai:dataCollection": "Each instance originates from a controlled fault-injection experiment on one of three open-source microservice testbeds (TrainTicket, Hotel Reservation from DeathStarBench, OpenTelemetry Demo) running under synthetic load. Faults, single or co-injected, are applied with Chaos Mesh, and traces, metrics, and logs are exported through OpenTelemetry for a normal window before the injection and an abnormal window during it. Log lines printed by the fault-injection tooling itself are removed because they would reveal the injection target; no other telemetry is edited.",
+  "rai:dataCollectionType": [
+    "Synthetic / Simulation",
+    "Automated machine instrumentation"
+  ],
+  "rai:dataCollectionTimeframe": {
+    "startDate": "2026-04-20",
+    "endDate": "2026-05-03"
+  },
+  "rai:dataAnnotationProtocol": "Labels are produced by PAVE (Path Annotation via Verified Effects), which verifies the recorded injection forward against the telemetry. Each injected fault is first tested on the injected service's own telemetry with a test matched to its fault family; a fault with no observable effect is marked not_effective, one whose effect stays on its own resources optional, and one with a service-level effect required. A propagation edge enters the verified graph only if it follows a known propagation mechanism, the downstream service deviates beyond both its pre-injection level and the variation seen in fault-free windows of the same system, and the downstream anomaly starts no earlier than the upstream one. Instances without a required fault are removed; a small number of labels are set by manual adjudication of the raw telemetry.",
+  "rai:dataAnnotationPlatform": "PAVE, implemented in rcabench-platform v3 (internal/reasoning), with manual adjudication overrides.",
+  "rai:dataAnnotationAnalysis": "A blind per-edge audit on 28 held-out instances, by independent auditor agents with query access to the raw telemetry and no access to any annotation, found 97.5% of the annotated service edges correct and 85.2% of the true edges recovered. Mean service edges per instance: 3.9 (ts 5.3, otel-demo 2.9, hs 1.5).",
+  "rai:annotationsPerItem": "1 (algorithmic, with manual adjudication where the audit disagreed).",
+  "rai:annotatorDemographics": "Not applicable: annotations are produced by software; audits were run by LLM agents.",
+  "rai:machineAnnotationTools": [
+    "PAVE (rcabench-platform v3 internal/reasoning)"
+  ],
+  "rai:personalSensitiveInformation": "None. The dataset contains only synthetic load and fault telemetry from open-source microservice testbeds running in isolated lab clusters. No real user data, PII, payment data, or production traffic is present. Service / endpoint names come from the public Train-Ticket, Hotel Reservation / DeathStarBench, and OpenTelemetry Demo projects.",
+  "rai:dataBiases": "System coverage is uneven: ts (TrainTicket) 275 instances, hs (Hotel Reservation) 148, otel-demo 34. Hotel Reservation instances have shallow propagation (1.5 service edges on average) and are much easier for current agents than TrainTicket instances; report per-system results alongside pooled ones. Only faults whose effect is observable in telemetry are labelled as required, so silent faults are absent by construction.",
+  "rai:dataUseCases": "Recommended: evaluating LLM agents and other RCA methods on microservice root-cause identification at the outcome level (root-cause set) and the process level (propagation graph). Not recommended: training or evaluating production incident triage on real customer traffic, since the load is synthetic and the clusters are isolated.",
+  "rai:dataLimitations": "Each instance covers one fault window in an isolated lab cluster; 178 of 457 instances carry co-injected faults. Propagation edges are limited to the mechanisms in PAVE's rule set, so effects travelling through mechanisms it does not model can be missing from the graph (edge recall 85.2% in the audit). For HTTP faults the altered end of a call is not always identifiable from telemetry; three such faults accept either end.",
+  "rai:dataSocialImpact": "The dataset is intended to advance reproducible RCA research for microservice systems. Risks are low: there is no human or operational data. Possible indirect risk: an RCA algorithm tuned only on this benchmark may transfer poorly to production fault distributions, potentially giving operators false confidence. Users are encouraged to report any observed gaps between benchmark performance and production performance via the dataset's HF discussion page.",
+  "rai:dataReleaseMaintenancePlan": "Versioned releases on Hugging Face under anon-ops/ops-lite; version 2.0.0 replaces the 500-instance 1.0.0 release, which remains accessible through the repository history."
+}