diff --git a/services/comms/atlasbot-deployment.yaml b/services/comms/atlasbot-deployment.yaml
index e1ff2bb..4ac3582 100644
--- a/services/comms/atlasbot-deployment.yaml
+++ b/services/comms/atlasbot-deployment.yaml
@@ -16,7 +16,7 @@ spec:
       labels:
         app: atlasbot
       annotations:
-        checksum/atlasbot-configmap: manual-atlasbot-66
+        checksum/atlasbot-configmap: manual-atlasbot-67
         vault.hashicorp.com/agent-inject: "true"
         vault.hashicorp.com/role: "comms"
         vault.hashicorp.com/agent-inject-secret-turn-secret: "kv/data/atlas/comms/turn-shared-secret"
diff --git a/services/comms/scripts/atlasbot/bot.py b/services/comms/scripts/atlasbot/bot.py
index abdcbf2..0d0f92b 100644
--- a/services/comms/scripts/atlasbot/bot.py
+++ b/services/comms/scripts/atlasbot/bot.py
@@ -579,6 +579,10 @@ def _detect_entity(q: str) -> str | None:
         or "architecture" in q
         or "machine" in q
         or "machines" in q
+        or "host" in q
+        or "hosts" in q
+        or "hostname" in q
+        or "hostnames" in q
         or TITAN_NODE_RE.search(q)
     ):
         return "node"
@@ -1775,20 +1779,29 @@ def _format_insight_text(key: str, text: str) -> str:
             .strip()
             .strip(".")
         )
+        has_jetson = "jetson=" in counts
+        has_amd64 = "amd64=" in counts
         detail = f"mixed hardware stack ({counts})"
-        flavor = "It blends low-power Pis with Jetson accelerators and a couple of AMD64 nodes."
+        if has_jetson and has_amd64:
+            flavor = "It blends low-power Pis with Jetson accelerators and a couple of AMD64 boxes."
+        elif has_jetson:
+            flavor = "It pairs low-power Pis with Jetson accelerators for edge and AI workloads."
+        elif has_amd64:
+            flavor = "It mixes low-power Pis with a couple of heavier AMD64 nodes."
+        else:
+            flavor = "It is a pretty uniform hardware stack, which is rare for a homelab."
         return f"{detail}. {flavor}"
     if key == "postgres":
         detail = cleaned.replace("Postgres is at ", "")
-        return f"Postgres is at {detail}; that suggests moderate load."
+        return f"Postgres is at {detail}; that feels like healthy, steady load rather than strain."
     if key == "pods":
         detail = cleaned.replace("There are ", "")
-        return f"Pods look steady ({detail}); the workload mix looks healthy."
+        return f"Pods look steady ({detail}); nothing looks stuck or unhealthy."
     if key == "availability":
-        return cleaned + " That suggests the cluster is stable right now."
+        return cleaned + " That is the kind of stability I like to see."
     if key in ("cpu", "ram"):
         suffix = (
-            " If you're chasing hotspots, that's the busiest workload right now."
+            " If you're chasing hotspots, that's the node I'd watch first."
             if key == "cpu"
             else " That box is carrying the heaviest memory load right now."
         )
@@ -1799,19 +1812,19 @@ def _format_insight_text(key: str, text: str) -> str:
 def _insight_prefix(prompt: str) -> str:
     q = normalize_query(prompt)
     if "coolest" in q:
-        return "If I had to pick the coolest detail: "
+        return "If I had to pick the coolest detail, I'd say "
     if "favorite" in q or "favourite" in q:
-        return "My favorite detail: "
+        return "My favorite detail is "
     if "trivia" in q:
         return "A bit of trivia I like: "
     if "most interesting" in q:
-        return "The most interesting detail to me: "
+        return "The most interesting detail to me is "
     if any(word in q for word in ("another", "else", "different", "other")):
         return "Another interesting detail: "
     if any(word in q for word in ("unconventional", "weird", "odd", "unique", "surprising")):
-        return "What stands out is that "
+        return "What stands out to me is that "
     if any(word in q for word in ("interesting", "notable", "fun", "cool")):
-        return "One thing I'd highlight: "
+        return "One thing I'd call out is "
     return ""
 
 
@@ -2389,6 +2402,21 @@ def _normalize_reply(value: Any) -> str:
     return _ensure_confidence(text)
 
 
+def _history_payload_lines(history_payload: list[Any]) -> list[str]:
+    lines: list[str] = []
+    if not isinstance(history_payload, list):
+        return lines
+    for item in history_payload[-12:]:
+        if isinstance(item, dict):
+            for key in ("content", "message", "text", "prompt", "question", "body", "answer", "reply", "response"):
+                val = item.get(key)
+                if isinstance(val, str) and val.strip():
+                    lines.append(val.strip())
+        elif isinstance(item, str) and item.strip():
+            lines.append(item.strip())
+    return [line for line in lines if line]
+
+
 # Internal HTTP endpoint for cluster answers (website uses this).
 class _AtlasbotHandler(BaseHTTPRequestHandler):
     server_version = "AtlasbotHTTP/1.0"
@@ -2439,15 +2467,7 @@ class _AtlasbotHandler(BaseHTTPRequestHandler):
         inventory = _snapshot_inventory(snapshot) or node_inventory_live()
         workloads = _snapshot_workloads(snapshot)
         history_payload = payload.get("history") or []
-        history_lines: list[str] = []
-        if isinstance(history_payload, list):
-            for item in history_payload[-10:]:
-                if isinstance(item, dict):
-                    content = item.get("content") or item.get("message") or ""
-                    if isinstance(content, str) and content.strip():
-                        history_lines.append(content.strip())
-                elif isinstance(item, str) and item.strip():
-                    history_lines.append(item.strip())
+        history_lines = _history_payload_lines(history_payload)
         history_cluster = _history_mentions_cluster(
             history_lines,
             inventory=inventory,