""" Gitea workflow / runner health contract snapshot. Loads the latest committed, read-only Gitea workflow and runner health contract. This module never calls Gitea, mutates workflows, restarts runners, stops containers, reads Secret payloads, triggers deploys, or sends notifications. """ from __future__ import annotations import json from pathlib import Path from typing import Any from src.services.snapshot_paths import default_evaluations_dir _DEFAULT_EVALUATIONS_DIR = default_evaluations_dir(Path(__file__)) _SNAPSHOT_PATTERN = "gitea_workflow_runner_health_*.json" _SCHEMA_VERSION = "gitea_workflow_runner_health_v1" def load_latest_gitea_workflow_runner_health( evaluations_dir: Path | None = None, ) -> dict[str, Any]: """Load the newest committed Gitea workflow / runner health snapshot.""" directory = evaluations_dir or _DEFAULT_EVALUATIONS_DIR candidates = sorted(directory.glob(_SNAPSHOT_PATTERN)) if not candidates: raise FileNotFoundError(f"no Gitea workflow / runner health snapshots found in {directory}") latest = candidates[-1] with latest.open(encoding="utf-8") as handle: payload = json.load(handle) if not isinstance(payload, dict): raise ValueError(f"{latest}: expected JSON object") _require_schema(payload, _SCHEMA_VERSION, str(latest)) _require_read_only_boundaries(payload, str(latest)) _require_operation_boundaries(payload, str(latest)) _require_rollup_consistency(payload, str(latest)) _require_workflow_evidence(payload, str(latest)) _require_runner_contracts(payload, str(latest)) _require_notification_contracts(payload, str(latest)) _require_operator_denials(payload, str(latest)) _require_no_plaintext_secret_payload_keys(payload, str(latest)) return payload def _require_schema(payload: dict[str, Any], expected: str, label: str) -> None: actual = payload.get("schema_version") if actual != expected: raise ValueError(f"{label}: expected schema_version={expected}, got {actual!r}") def _require_read_only_boundaries(payload: dict[str, Any], label: str) -> None: program_status = payload.get("program_status") or {} if program_status.get("read_only_mode") is not True: raise ValueError(f"{label}: program_status.read_only_mode must be true") approval_boundaries = payload.get("approval_boundaries") or {} allowed = sorted(key for key, value in approval_boundaries.items() if value is not False) if allowed: raise ValueError(f"{label}: approval boundaries must remain false: {allowed}") def _require_operation_boundaries(payload: dict[str, Any], label: str) -> None: boundaries = payload.get("operation_boundaries") or {} if boundaries.get("read_only_api_allowed") is not True: raise ValueError(f"{label}: read_only_api_allowed must be true") blocked_flags = { "workflow_modification_allowed", "runner_restart_allowed", "runner_container_stop_allowed", "runner_label_change_allowed", "runner_registration_allowed", "secret_read_allowed", "secret_plaintext_allowed", "notification_send_allowed", "schedule_enable_allowed", "gitea_api_write_allowed", "deploy_trigger_allowed", "migration_trigger_allowed", } allowed = sorted(flag for flag in blocked_flags if boundaries.get(flag) is not False) if allowed: raise ValueError(f"{label}: operation boundaries must remain false: {allowed}") def _require_rollup_consistency(payload: dict[str, Any], label: str) -> None: workflows = payload.get("workflow_records") or [] runners = payload.get("runner_contracts") or [] notifications = payload.get("notification_contracts") or [] rollups = payload.get("rollups") or {} if rollups.get("total_workflows") != len(workflows): raise ValueError(f"{label}: rollups.total_workflows must match workflow_records") if rollups.get("by_workflow_status") != _count_by(workflows, "status"): raise ValueError(f"{label}: rollups.by_workflow_status must match workflow_records") if rollups.get("by_runner_evidence_status") != _count_by(workflows, "runner_evidence_status"): raise ValueError(f"{label}: rollups.by_runner_evidence_status must match workflow_records") if rollups.get("workflows_with_schedule") != _count_where(workflows, lambda row: "schedule" in row.get("triggers", [])): raise ValueError(f"{label}: rollups.workflows_with_schedule must match workflow_records") if rollups.get("workflows_with_workflow_dispatch") != _count_where( workflows, lambda row: "workflow_dispatch" in row.get("triggers", []), ): raise ValueError(f"{label}: rollups.workflows_with_workflow_dispatch must match workflow_records") if rollups.get("workflows_with_notify_bridge") != _count_where( workflows, lambda row: (row.get("notify_bridge_calls") or 0) > 0, ): raise ValueError(f"{label}: rollups.workflows_with_notify_bridge must match workflow_records") if rollups.get("workflows_with_actionable_or_failure_quiet_policy") != _count_where( workflows, lambda row: row.get("notification_policy") in {"actionable_only_no_success_noise", "failure_only"}, ): raise ValueError( f"{label}: rollups.workflows_with_actionable_or_failure_quiet_policy must match workflow_records" ) attestation_required = sorted( row.get("workflow_id") for row in workflows if row.get("runner_evidence_status") in {"owner_attestation_required", "comment_ambiguous"} ) if sorted(rollups.get("workflow_ids_requiring_runner_attestation") or []) != attestation_required: raise ValueError( f"{label}: rollups.workflow_ids_requiring_runner_attestation must match workflow_records" ) if rollups.get("total_runner_contracts") != len(runners): raise ValueError(f"{label}: rollups.total_runner_contracts must match runner_contracts") action_required_runners = sorted( runner.get("contract_id") for runner in runners if runner.get("status") == "action_required" ) if sorted(rollups.get("runner_contracts_requiring_action") or []) != action_required_runners: raise ValueError(f"{label}: rollups.runner_contracts_requiring_action must match runner_contracts") if rollups.get("notification_contracts_total") != len(notifications): raise ValueError(f"{label}: rollups.notification_contracts_total must match notification_contracts") quiet_contract_ids = sorted( contract.get("contract_id") for contract in notifications if contract.get("policy_kind") in {"failure_only", "actionable_only"} ) if rollups.get("notification_contracts_quiet_success_count") != len(quiet_contract_ids): raise ValueError( f"{label}: rollups.notification_contracts_quiet_success_count must match notification_contracts" ) if sorted(rollups.get("notification_contracts_quiet_success_ids") or []) != quiet_contract_ids: raise ValueError(f"{label}: rollups.notification_contracts_quiet_success_ids must match contracts") def _require_workflow_evidence(payload: dict[str, Any], label: str) -> None: workflows = payload.get("workflow_records") or [] missing = sorted( workflow.get("workflow_id") for workflow in workflows if not workflow.get("file_ref") or not workflow.get("evidence_refs") or not workflow.get("triggers") or not workflow.get("runner_labels") ) if missing: raise ValueError(f"{label}: workflow_records must include file, evidence, trigger, runner labels: {missing}") def _require_runner_contracts(payload: dict[str, Any], label: str) -> None: runners = payload.get("runner_contracts") or [] missing = sorted( runner.get("contract_id") for runner in runners if not runner.get("evidence_refs") or not runner.get("health_contract") or not runner.get("next_action") ) if missing: raise ValueError(f"{label}: runner_contracts must include evidence, health contract, next_action: {missing}") def _require_notification_contracts(payload: dict[str, Any], label: str) -> None: contracts = payload.get("notification_contracts") or [] missing = sorted( contract.get("contract_id") for contract in contracts if not contract.get("success_noise_policy") or not contract.get("failure_policy") or not contract.get("evidence_refs") ) if missing: raise ValueError(f"{label}: notification_contracts must include policy and evidence: {missing}") quiet_contracts = [ contract for contract in contracts if contract.get("policy_kind") in {"failure_only", "actionable_only"} ] if len(quiet_contracts) < 2: raise ValueError(f"{label}: must preserve failure-only and actionable-only notification contracts") noisy_quiet_contracts = sorted( contract.get("contract_id") for contract in quiet_contracts if "不" not in str(contract.get("success_noise_policy")) and "quiet" not in str(contract.get("success_noise_policy")).lower() ) if noisy_quiet_contracts: raise ValueError(f"{label}: quiet notification contracts must suppress success noise: {noisy_quiet_contracts}") def _require_operator_denials(payload: dict[str, Any], label: str) -> None: contract = payload.get("operator_contract") or {} must_not_interpret_as = set(contract.get("must_not_interpret_as") or []) required_denials = { "workflow 修改批准", "runner restart / stop 批准", "Secret 已讀取或可輸出", "Telegram 測試通知批准", "Gitea write token 授權", "deploy / migration workflow 觸發批准", } if not required_denials.issubset(must_not_interpret_as): raise ValueError(f"{label}: operator_contract.must_not_interpret_as is missing required denials") def _require_no_plaintext_secret_payload_keys(value: Any, label: str, path: str = "$") -> None: if isinstance(value, dict): forbidden_key_fragments = { "secret_value", "token_value", "authorization_header", "private_key", "webhook_secret", "runner_token", } for key, nested in value.items(): lowered = str(key).lower() if any(fragment in lowered for fragment in forbidden_key_fragments): raise ValueError(f"{label}: forbidden secret payload key at {path}.{key}") _require_no_plaintext_secret_payload_keys(nested, label, f"{path}.{key}") elif isinstance(value, list): for index, nested in enumerate(value): _require_no_plaintext_secret_payload_keys(nested, label, f"{path}[{index}]") def _count_by(items: list[dict[str, Any]], key: str) -> dict[str, int]: counts: dict[str, int] = {} for item in items: value = item.get(key) counts[value] = counts.get(value, 0) + 1 return counts def _count_where(items: list[dict[str, Any]], predicate) -> int: return sum(1 for item in items if predicate(item))