diff --git a/services/comms/scripts/atlasbot/bot.py b/services/comms/scripts/atlasbot/bot.py
index 6f18b9e..96765b1 100644
--- a/services/comms/scripts/atlasbot/bot.py
+++ b/services/comms/scripts/atlasbot/bot.py
@@ -1037,6 +1037,11 @@ def facts_context(
         nodes_list = by_hardware.get(key) or []
         if nodes_list:
             lines.append(f"- {key}: {', '.join(nodes_list)}")
+    if by_hardware:
+        counts = {key: len(nodes_list) for key, nodes_list in by_hardware.items() if nodes_list}
+        if counts:
+            parts = [f"{key}={count}" for key, count in sorted(counts.items())]
+            lines.append(f"- nodes_by_hardware_count: {', '.join(parts)}")
     non_rpi = sorted(set(by_hardware.get("jetson", [])) | set(by_hardware.get("amd64", [])))
     if non_rpi:
         lines.append(f"- non_raspberry_pi: {', '.join(non_rpi)}")
@@ -1096,6 +1101,25 @@ def facts_context(
         value = metrics.get(key)
         if value is not None:
             lines.append(f"- {key}: {value}")
+    if workloads:
+        ns_counts: dict[str, int] = collections.defaultdict(int)
+        for entry in workloads:
+            if not isinstance(entry, dict):
+                continue
+            ns = entry.get("namespace") or ""
+            pods = entry.get("pods_running")
+            if pods is None:
+                pods = entry.get("pods_total")
+            try:
+                pods_val = int(pods)
+            except (TypeError, ValueError):
+                pods_val = 0
+            if ns:
+                ns_counts[ns] += pods_val
+        if ns_counts:
+            top_ns = sorted(ns_counts.items(), key=lambda item: item[1], reverse=True)[:5]
+            parts = [f"{ns}={count}" for ns, count in top_ns]
+            lines.append(f"- pods_by_namespace: {', '.join(parts)}")
 
     top_restarts = metrics.get("top_restarts_1h") if isinstance(metrics.get("top_restarts_1h"), list) else []
     if top_restarts:
@@ -2725,6 +2749,8 @@ def _fact_line_tags(line: str) -> set[str]:
         tags.add("database")
     if "pods_" in text or "pod phases" in text or "restarts" in text:
         tags.add("pods")
+    if "namespace" in text:
+        tags.add("workloads")
     if "workloads" in text or "primary_node" in text or "workload_" in text:
         tags.add("workloads")
     if "node_details" in text:
@@ -2900,6 +2926,8 @@ def _open_ended_system() -> str:
         "If the question asks for a list, embed the list inline in a sentence (comma-separated). "
         "If the question is ambiguous, pick a reasonable interpretation and state it briefly. "
         "Avoid repeating the exact same observation as the last response if possible; vary across metrics, workload, or hardware details. "
+        "When the fact pack includes hottest_cpu/ram/net/io lines, use them to answer hottest/busiest node questions. "
+        "When the fact pack includes postgres_hottest_db, use it for questions about the busiest database. "
         "Do not invent numbers or facts. "
         "End with lines: Confidence, Relevance (0-100), Satisfaction (0-100), HallucinationRisk (low|medium|high)."
     )