feat(api): 預填修復候選 PlayBook 草案

apps/api/src/services/repair_candidate_service.py

@@ -350,7 +350,13 @@ class RepairCandidateService:
             namespace=namespace,
         )
         metadata["playbook_draft_required"] = True
-        metadata["repair_candidate_next_step"] = draft_package["next_step"]
+        metadata["repair_candidate_prefilled_draft_summary"] = self._draft_summary_for_operator(
+            draft_package.get("playbook_draft_template") or {}
+        )
+        metadata["repair_candidate_next_step"] = self._next_step_with_draft_summary(
+            str(draft_package["next_step"]),
+            str(metadata["repair_candidate_prefilled_draft_summary"]),
+        )
         metadata["repair_candidate_draft_package"] = draft_package
         metadata["fallback_action"] = fallback_action
         return RepairCandidateResult(
@@ -487,6 +493,16 @@ class RepairCandidateService:
             evidence=evidence,
             playbook=playbook,
         )
+        playbook_draft_template = self._build_playbook_draft_template(
+            coverage_gap=coverage_gap,
+            blockers=blockers,
+            lane=lane,
+            alertname=alertname,
+            target_resource=target_resource,
+            namespace=namespace,
+            evidence=evidence,
+            playbook=playbook,
+        )
         incident_id = getattr(incident, "incident_id", "") if incident else ""
         project_id = str(getattr(incident, "project_id", "awoooi") or "awoooi")
         work_item: dict[str, Any] | None = None
@@ -517,6 +533,7 @@ class RepairCandidateService:
                 "next_step": next_step,
                 "required_fields": required_fields,
                 "coverage_gap": coverage_gap,
+                "playbook_draft_template": playbook_draft_template,
                 "blocked_operations": blocked_operations,
                 "target_href": f"/awooop/runs?{run_query}",
                 "work_item_href": f"/awooop/work-items?{work_item_query}",
@@ -542,12 +559,128 @@ class RepairCandidateService:
             "evidence_snapshot_id": evidence_ref,
             "required_fields": required_fields,
             "coverage_gap": coverage_gap,
+            "playbook_draft_template": playbook_draft_template,
             "blocked_operations": blocked_operations,
         }
         if work_item:
             package["awooop_work_item"] = work_item
         return package
 
+    def _draft_summary_for_operator(self, template: dict[str, Any]) -> str:
+        route = str(template.get("suggested_route") or "--")
+        repair = str(template.get("repair_command_template") or "--")
+        verifier = template.get("verifier_plan_template") or []
+        verifier_head = "--"
+        if isinstance(verifier, list) and verifier:
+            verifier_head = str(verifier[0])
+        return f"route={route}; repair_template={repair}; verifier={verifier_head}"
+
+    def _next_step_with_draft_summary(self, next_step: str, draft_summary: str) -> str:
+        if not draft_summary or draft_summary == "route=--; repair_template=--; verifier=--":
+            return next_step
+        return f"{next_step} 草案已預填：{draft_summary}"
+
+    def _build_playbook_draft_template(
+        self,
+        *,
+        coverage_gap: dict[str, Any],
+        blockers: list[str],
+        lane: str,
+        alertname: str,
+        target_resource: str,
+        namespace: str,
+        evidence: EvidenceSnapshot | None,
+        playbook: Playbook | None,
+    ) -> dict[str, Any]:
+        """Prefill a service-specific PlayBook draft without authorizing execution."""
+
+        target_kind = str(coverage_gap.get("target_kind") or "unknown")
+        target = (target_resource or "").strip()
+        ns = (namespace or "").strip()
+        command_template = "owner_supplied_service_specific_command_after_evidence_review"
+        rollback_template = "owner_supplied_rollback_or_safe_stop_after_evidence_review"
+        route = "owner_review_required"
+        verifier_plan = [
+            "rerun_mcp_evidence_collection",
+            "confirm_fingerprint_recurrence_stops_or_decreases",
+            "write_execution_result_and_verifier_outcome",
+            "update_km_and_playbook_trust_after_owner_review",
+        ]
+
+        if target_kind == "k8s_workload":
+            workload = target.replace("deployment/", "").replace("pod/", "").replace("svc/", "")
+            workload_ref = workload or "<deployment>"
+            namespace_ref = ns or "<namespace>"
+            command_template = (
+                f"kubectl rollout restart deployment/{workload_ref} -n {namespace_ref}"
+            )
+            rollback_template = (
+                f"kubectl rollout undo deployment/{workload_ref} -n {namespace_ref}"
+            )
+            route = "k8s_rollout_after_owner_review"
+            verifier_plan.extend([
+                f"kubectl rollout status deployment/{workload_ref} -n {namespace_ref}",
+                f"kubectl get events -n {namespace_ref} --field-selector involvedObject.name={workload_ref}",
+            ])
+        elif target_kind == "host_service":
+            service_ref = target or "<service>"
+            command_template = f"systemctl restart {service_ref}"
+            rollback_template = f"systemctl status {service_ref}; journalctl -u {service_ref} -n 120"
+            route = "host_service_route_after_owner_review"
+            verifier_plan.extend([
+                f"systemctl is-active {service_ref}",
+                f"journalctl -u {service_ref} -n 120 --no-pager",
+            ])
+        elif target_kind == "service":
+            service_ref = target or "<service>"
+            command_template = f"service-specific repair for {service_ref}"
+            rollback_template = f"service-specific rollback for {service_ref}"
+            route = "service_specific_route_after_owner_review"
+            verifier_plan.append(f"confirm service health for {service_ref}")
+
+        owner_review_checklist = [
+            "確認 alertname 與 target selector 沒有誤配",
+            "確認 MCP evidence refs 足以支持修復假設",
+            "確認 repair command 不是純診斷命令或通用兜底命令",
+            "確認 rollback command 能在失敗時安全收斂",
+            "確認 verifier plan 能判斷成功、失敗與是否升級人工",
+            "確認 blast radius、maintenance window 與 owner 已填寫",
+            "確認 runtime_execution_authorized 仍為 false，直到正式 approval gate 通過",
+        ]
+
+        return {
+            "schema_version": "service_specific_playbook_draft_template_v1",
+            "status": "prefilled_owner_review_required",
+            "lane": lane,
+            "coverage_key": coverage_gap.get("coverage_key"),
+            "blocking_stage": coverage_gap.get("blocking_stage"),
+            "blockers": list(dict.fromkeys(blockers)),
+            "alert_selector": {
+                "alertname": alertname or None,
+                "namespace": ns or None,
+                "target_resource": target or None,
+                "target_kind": target_kind,
+            },
+            "mcp_evidence_refs": coverage_gap.get("required_mcp_evidence_refs") or [],
+            "matched_playbook": {
+                "playbook_id": playbook.playbook_id if playbook else None,
+                "name": playbook.name if playbook else None,
+                "status": playbook.status.value if playbook else None,
+                "trust_score": float(playbook.trust_score) if playbook else None,
+            },
+            "suggested_route": route,
+            "repair_command_template": command_template,
+            "rollback_command_template": rollback_template,
+            "verifier_plan_template": list(dict.fromkeys(verifier_plan)),
+            "owner_review_checklist": owner_review_checklist,
+            "evidence_snapshot_id": evidence.snapshot_id if evidence else None,
+            "template_is_executable": False,
+            "approval_required_before_execution": True,
+            "runtime_execution_authorized": False,
+            "writes_runtime_state": False,
+            "telegram_send_authorized": False,
+        }
+
     def _build_coverage_gap(
         self,
         *,