ultimate-pi 0.1.7 → 0.2.2

package/.pi/harness/specs/eval-verdict.schema.json

@@ -0,0 +1,59 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/eval-verdict.schema.json",
+	"title": "EvalVerdict",
+	"description": "Independent evaluator result consumed by policy gating.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"contract_version",
+		"run_id",
+		"status",
+		"failed_checks",
+		"regression_flags",
+		"confidence",
+		"recommended_action"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"contract_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"run_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"status": {
+			"type": "string",
+			"enum": ["pass", "conditional_pass", "fail"]
+		},
+		"failed_checks": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"regression_flags": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"confidence": {
+			"type": "number",
+			"minimum": 0,
+			"maximum": 1
+		},
+		"recommended_action": {
+			"type": "string",
+			"enum": ["proceed_to_adversary", "replan", "rollback"]
+		}
+	}
+}

package/.pi/harness/specs/incident-record.schema.json

@@ -0,0 +1,84 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/incident-record.schema.json",
+	"title": "IncidentRecord",
+	"description": "Incident, mitigation, rollback, and override audit record.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"contract_version",
+		"incident_id",
+		"trigger",
+		"blast_radius",
+		"mitigation",
+		"rollback_ref",
+		"postmortem_required"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"contract_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"incident_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"trigger": {
+			"type": "string",
+			"minLength": 1
+		},
+		"blast_radius": {
+			"type": "string",
+			"minLength": 1
+		},
+		"mitigation": {
+			"type": "string",
+			"minLength": 1
+		},
+		"rollback_ref": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["revert_command", "revert_branch", "patch_bundle"],
+			"properties": {
+				"revert_command": {
+					"type": "string",
+					"minLength": 1
+				},
+				"revert_branch": {
+					"type": "string",
+					"minLength": 1
+				},
+				"patch_bundle": {
+					"type": "string",
+					"minLength": 1
+				}
+			}
+		},
+		"postmortem_required": {
+			"type": "boolean"
+		},
+		"policy_override": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["human_override", "approver_count", "justification"],
+			"properties": {
+				"human_override": {
+					"type": "boolean"
+				},
+				"approver_count": {
+					"type": "integer",
+					"enum": [1]
+				},
+				"justification": {
+					"type": "string",
+					"minLength": 1
+				}
+			}
+		}
+	}
+}

package/.pi/harness/specs/plan-packet.schema.json

@@ -0,0 +1,90 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/plan-packet.schema.json",
+	"title": "PlanPacket",
+	"description": "Plan-gated mutation contract emitted before execution.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"contract_version",
+		"plan_id",
+		"task_id",
+		"scope",
+		"assumptions",
+		"risk_level",
+		"acceptance_checks",
+		"rollback_plan"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"contract_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"plan_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"task_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"scope": {
+			"type": "string",
+			"minLength": 1
+		},
+		"assumptions": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"risk_level": {
+			"type": "string",
+			"enum": ["low", "med", "high"]
+		},
+		"acceptance_checks": {
+			"type": "array",
+			"minItems": 1,
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"rollback_plan": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["rollback_artifacts", "revert_commit_ready"],
+			"properties": {
+				"revert_commit_ready": {
+					"type": "boolean",
+					"const": true
+				},
+				"rollback_artifacts": {
+					"type": "object",
+					"additionalProperties": false,
+					"required": ["revert_command", "revert_branch", "patch_bundle"],
+					"properties": {
+						"revert_command": {
+							"type": "string",
+							"minLength": 1
+						},
+						"revert_branch": {
+							"type": "string",
+							"minLength": 1
+						},
+						"patch_bundle": {
+							"type": "string",
+							"minLength": 1
+						}
+					}
+				}
+			}
+		}
+	}
+}

package/.pi/harness/specs/round-result.schema.json

@@ -0,0 +1,126 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/round-result.schema.json",
+	"title": "RoundResult",
+	"description": "One structured debate round with strict budget accounting.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"contract_version",
+		"run_id",
+		"debate_id",
+		"round_index",
+		"participants",
+		"claims",
+		"rebuttals",
+		"evidence_refs",
+		"token_usage",
+		"budget_profile",
+		"consensus_delta"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"contract_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"run_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"debate_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"round_index": {
+			"type": "integer",
+			"minimum": 1,
+			"maximum": 6
+		},
+		"participants": {
+			"type": "array",
+			"minItems": 2,
+			"items": {
+				"type": "string",
+				"enum": ["EvaluatorAgent", "AdversaryAgent", "TieBreakerAgent"]
+			}
+		},
+		"claims": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"rebuttals": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"evidence_refs": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"token_usage": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["per_agent", "round_total"],
+			"properties": {
+				"per_agent": {
+					"type": "object",
+					"additionalProperties": {
+						"type": "integer",
+						"minimum": 0,
+						"maximum": 2500
+					}
+				},
+				"round_total": {
+					"type": "integer",
+					"minimum": 0
+				}
+			}
+		},
+		"budget_profile": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": [
+				"name",
+				"max_rounds",
+				"round_token_cap",
+				"debate_global_cap"
+			],
+			"properties": {
+				"name": {
+					"type": "string",
+					"const": "aggressive"
+				},
+				"max_rounds": {
+					"type": "integer",
+					"const": 6
+				},
+				"round_token_cap": {
+					"type": "integer",
+					"const": 2500
+				},
+				"debate_global_cap": {
+					"type": "integer",
+					"const": 35000
+				}
+			}
+		},
+		"consensus_delta": {
+			"type": "number",
+			"minimum": -1,
+			"maximum": 1
+		}
+	}
+}

package/.pi/harness/specs/router-tuning-proposal.schema.json

@@ -0,0 +1,114 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/router-tuning-proposal.schema.json",
+	"title": "RouterTuningProposal",
+	"description": "Evidence-gated, approval-required proposal for model-router updates.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"proposal_id",
+		"created_at",
+		"router_path",
+		"base_router_sha256",
+		"candidate_router_sha256",
+		"evidence",
+		"status",
+		"approval",
+		"candidate_router"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"proposal_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"created_at": {
+			"type": "string",
+			"minLength": 1
+		},
+		"router_path": {
+			"type": "string",
+			"const": ".pi/model-router.json"
+		},
+		"base_router_sha256": {
+			"type": "string",
+			"pattern": "^[a-f0-9]{64}$"
+		},
+		"candidate_router_sha256": {
+			"type": "string",
+			"pattern": "^[a-f0-9]{64}$"
+		},
+		"status": {
+			"type": "string",
+			"enum": ["proposed", "approved_applied", "rejected"]
+		},
+		"evidence": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": [
+				"sample_count",
+				"min_sample_count",
+				"success_rate_delta",
+				"cost_per_task_delta",
+				"regression_guard_passed",
+				"trace_refs"
+			],
+			"properties": {
+				"sample_count": {
+					"type": "integer",
+					"minimum": 1
+				},
+				"min_sample_count": {
+					"type": "integer",
+					"minimum": 1
+				},
+				"success_rate_delta": {
+					"type": "number"
+				},
+				"cost_per_task_delta": {
+					"type": "number"
+				},
+				"regression_guard_passed": {
+					"type": "boolean"
+				},
+				"trace_refs": {
+					"type": "array",
+					"items": {
+						"type": "string",
+						"minLength": 1
+					}
+				},
+				"notes": {
+					"type": "string"
+				}
+			}
+		},
+		"approval": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["required", "approved_by", "approved_at", "justification"],
+			"properties": {
+				"required": {
+					"type": "boolean",
+					"const": true
+				},
+				"approved_by": {
+					"type": ["string", "null"]
+				},
+				"approved_at": {
+					"type": ["string", "null"]
+				},
+				"justification": {
+					"type": ["string", "null"]
+				}
+			}
+		},
+		"candidate_router": {
+			"type": "object"
+		}
+	}
+}

package/.pi/harness/specs/run-trace.schema.json

@@ -0,0 +1,107 @@
+{
+	"$schema": "https://json-schema.org/draft/2020-12/schema",
+	"$id": "https://ultimate-pi.local/.pi/harness/specs/run-trace.schema.json",
+	"title": "RunTrace",
+	"description": "Append-only execution trace row for harness phases.",
+	"type": "object",
+	"additionalProperties": false,
+	"required": [
+		"schema_version",
+		"contract_version",
+		"run_id",
+		"plan_id",
+		"agent_id",
+		"phase",
+		"model",
+		"thinking_level",
+		"tool_spans",
+		"artifact_refs",
+		"cost"
+	],
+	"properties": {
+		"schema_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"contract_version": {
+			"type": "string",
+			"const": "1.0.0"
+		},
+		"run_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"plan_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"agent_id": {
+			"type": "string",
+			"minLength": 1
+		},
+		"phase": {
+			"type": "string",
+			"enum": ["plan", "execute", "evaluate", "adversary", "merge"]
+		},
+		"model": {
+			"type": "string",
+			"minLength": 1
+		},
+		"thinking_level": {
+			"type": "string",
+			"enum": ["off", "low", "medium", "high", "xhigh"]
+		},
+		"tool_spans": {
+			"type": "array",
+			"items": {
+				"type": "object",
+				"additionalProperties": false,
+				"required": ["tool_call_id", "tool_name", "started_at", "ended_at"],
+				"properties": {
+					"tool_call_id": {
+						"type": "string",
+						"minLength": 1
+					},
+					"tool_name": {
+						"type": "string",
+						"minLength": 1
+					},
+					"started_at": {
+						"type": "string",
+						"format": "date-time"
+					},
+					"ended_at": {
+						"type": "string",
+						"format": "date-time"
+					}
+				}
+			}
+		},
+		"artifact_refs": {
+			"type": "array",
+			"items": {
+				"type": "string",
+				"minLength": 1
+			}
+		},
+		"cost": {
+			"type": "object",
+			"additionalProperties": false,
+			"required": ["input_tokens", "output_tokens", "total_tokens"],
+			"properties": {
+				"input_tokens": {
+					"type": "integer",
+					"minimum": 0
+				},
+				"output_tokens": {
+					"type": "integer",
+					"minimum": 0
+				},
+				"total_tokens": {
+					"type": "integer",
+					"minimum": 0
+				}
+			}
+		}
+	}
+}