diff --git a/ariadne/services/cluster_state.py b/ariadne/services/cluster_state.py
index 8d2037f..32607ee 100644
--- a/ariadne/services/cluster_state.py
+++ b/ariadne/services/cluster_state.py
@@ -1948,6 +1948,7 @@ def _namespace_context(
     namespace_pods: list[dict[str, Any]],
     namespace_nodes: list[dict[str, Any]],
     namespace_capacity: list[dict[str, Any]],
+    namespace_baseline: dict[str, dict[str, dict[str, float]]],
 ) -> list[dict[str, Any]]:
     node_map = {entry.get("namespace"): entry for entry in namespace_nodes if isinstance(entry, dict)}
     cap_map = {entry.get("namespace"): entry for entry in namespace_capacity if isinstance(entry, dict)}
@@ -1967,6 +1968,10 @@ def _namespace_context(
                 {"node": name, "pods": count}
                 for name, count in sorted(nodes.items(), key=lambda item: (-item[1], item[0]))[:3]
             ]
+        baseline = namespace_baseline.get(namespace, {}) if isinstance(namespace_baseline, dict) else {}
+        delta_cpu = _baseline_delta(cap_entry.get("cpu_usage"), baseline.get("cpu", {}))
+        delta_mem = _baseline_delta(cap_entry.get("mem_usage"), baseline.get("mem", {}))
+        baseline_delta = {k: v for k, v in (("cpu", delta_cpu), ("mem", delta_mem)) if v is not None}
         output.append(
             {
                 "namespace": namespace,
@@ -1983,12 +1988,49 @@ def _namespace_context(
                 "mem_usage": cap_entry.get("mem_usage"),
                 "mem_requests": cap_entry.get("mem_requests"),
                 "mem_ratio": cap_entry.get("mem_usage_ratio"),
+                "baseline_delta": baseline_delta,
             }
         )
     output.sort(key=lambda item: (-(item.get("pods_total") or 0), item.get("namespace") or ""))
     return output
 
 
+def _namespace_nodes_top(namespace_context: list[dict[str, Any]], limit: int = 5) -> list[dict[str, Any]]:
+    output: list[dict[str, Any]] = []
+    for entry in namespace_context[:limit]:
+        if not isinstance(entry, dict):
+            continue
+        output.append(
+            {
+                "namespace": entry.get("namespace"),
+                "pods_total": entry.get("pods_total"),
+                "primary_node": entry.get("primary_node"),
+                "nodes_top": entry.get("nodes_top") or [],
+            }
+        )
+    return output
+
+
+def _workload_nodes_top(workloads: list[dict[str, Any]], limit: int = 5) -> list[dict[str, Any]]:
+    output: list[dict[str, Any]] = []
+    entries = [w for w in workloads if isinstance(w, dict)]
+    entries.sort(
+        key=lambda item: (-(item.get("pods_total") or 0), item.get("namespace") or "", item.get("workload") or ""),
+    )
+    for entry in entries[:limit]:
+        output.append(
+            {
+                "namespace": entry.get("namespace"),
+                "workload": entry.get("workload"),
+                "source": entry.get("source"),
+                "pods_total": entry.get("pods_total"),
+                "pods_running": entry.get("pods_running"),
+                "primary_node": entry.get("primary_node"),
+            }
+        )
+    return output
+
+
 def _node_context(
     node_details: list[dict[str, Any]],
     node_load: list[dict[str, Any]],
@@ -2004,6 +2046,13 @@ def _node_context(
             continue
         load_entry = load_map.get(name, {})
         baseline = node_baseline.get(name, {}) if isinstance(node_baseline, dict) else {}
+        deltas: dict[str, float] = {}
+        for key in ("cpu", "ram", "net", "io", "disk"):
+            current = load_entry.get(key)
+            stats = baseline.get(key, {}) if isinstance(baseline, dict) else {}
+            delta = _baseline_delta(current, stats)
+            if delta is not None:
+                deltas[key] = delta
         output.append(
             {
                 "node": name,
@@ -2024,12 +2073,22 @@ def _node_context(
                 "io": load_entry.get("io"),
                 "load_index": load_entry.get("load_index"),
                 "baseline": baseline,
+                "baseline_delta": deltas,
             }
         )
     output.sort(key=lambda item: (-(item.get("load_index") or 0), item.get("node") or ""))
     return output
 
 
+def _baseline_delta(current: Any, stats: dict[str, Any]) -> float | None:
+    if not isinstance(current, (int, float)):
+        return None
+    avg = stats.get("avg")
+    if not isinstance(avg, (int, float)) or avg == 0:
+        return None
+    return round(((float(current) - float(avg)) / float(avg)) * 100, 2)
+
+
 def _build_anomalies(
     metrics: dict[str, Any],
     nodes_summary: dict[str, Any],
@@ -2401,6 +2460,7 @@ def collect_cluster_state() -> tuple[dict[str, Any], ClusterStateSummary]:
         namespace_pods,
         namespace_nodes,
         metrics.get("namespace_capacity", []),
+        metrics.get("namespace_baseline_map", {}),
     )
     node_context = _node_context(
         node_details,
@@ -2431,6 +2491,11 @@ def collect_cluster_state() -> tuple[dict[str, Any], ClusterStateSummary]:
             "workload_not_ready": _workload_not_ready_items(workload_health)[:5],
             "pod_restarts": _pod_restarts_top(metrics),
         },
+        "relationships": {
+            "namespace_nodes": _namespace_nodes_top(namespace_context, 5),
+            "node_namespaces": metrics.get("node_pods_top", []),
+            "workload_nodes": _workload_nodes_top(workloads, 5),
+        },
         "attention_ranked": _build_attention_ranked(metrics, node_context, pod_issues, workload_health),
         "anomalies": anomalies,
         "health_bullets": _health_bullets(metrics, node_summary, workload_health, anomalies),