fix(ai): improve docker repair verification signals

apps/api/src/services/post_execution_verifier.py

@@ -255,6 +255,8 @@ class PostExecutionVerifier:
             "pod_name": labels.get("pod", labels.get("name", "")),
             "deployment": labels.get("deployment", ""),
             "host": labels.get("instance", "").split(":")[0] or labels.get("host", ""),
+            "container_name": _extract_container_name(labels),
+            "filter_name": _extract_container_name(labels),
             "query": _build_prometheus_query(alertname, labels),
         }
 
@@ -495,7 +497,13 @@ def _assess_recovery(
         "'success': true",
         '"success": true',
     ]
-    if any(sig in post_str for sig in success_signals):
+    has_success_signal = any(sig in post_str for sig in success_signals)
+    if not has_success_signal and _docker_state_indicates_running(post_str):
+        has_success_signal = True
+
+    if has_success_signal:
+        if _is_observe_only_action(action_taken):
+            return "degraded"
         # 但如果 pre_state 已經是 running，可能是無效操作
         if pre_str and any(sig in pre_str for sig in success_signals):
             # 如果執行的是 restart，即使 pre/post 都 Running 也算 success
@@ -533,6 +541,53 @@ def _get_labels(incident: "Incident") -> dict[str, Any]:
     return {}
 
 
+def _extract_container_name(labels: dict[str, Any]) -> str:
+    """Resolve Docker container target labels for post-execution SSH sensors."""
+    for key in ("filter_name", "container_name", "container", "resource", "name"):
+        value = str(labels.get(key) or "").strip()
+        if value and "{" not in value and "}" not in value:
+            return value
+    return ""
+
+
+def _is_observe_only_action(action_taken: str) -> bool:
+    """Return true when the executed step collected evidence but did not mutate state."""
+    lowered = (action_taken or "").lower()
+    observe_tokens = (
+        "mcp:ssh_diagnose",
+        "ssh_diagnose",
+        "docker stats",
+        "ps aux",
+        "free -h",
+        "df -h",
+    )
+    mutation_tokens = (
+        "restart",
+        "delete",
+        "rollout",
+        "scale",
+        "patch",
+        "apply",
+        "prune",
+        "truncate",
+        "clear",
+    )
+    return any(token in lowered for token in observe_tokens) and not any(
+        token in lowered for token in mutation_tokens
+    )
+
+
+def _docker_state_indicates_running(post_str: str) -> bool:
+    """Recognize Docker-specific healthy/running output without matching uptime."""
+    if not any(token in post_str for token in ("docker ps", "docker inspect", "docker stats")):
+        return False
+    return bool(
+        re.search(r'\bup\s+\d', post_str)
+        or re.search(r'["\']status["\']\s*:\s*["\']running["\']', post_str)
+        or re.search(r'["\']running["\']\s*:\s*true', post_str)
+    )
+
+
 def _build_prometheus_query(alertname: str, labels: dict[str, Any]) -> str:
     """Build a non-empty PromQL probe for post-execution metric sensors."""