From cc3739efe4d93d6673aa57c0eda054b3a38e815d Mon Sep 17 00:00:00 2001
From: Brad Stein <Brad.Stein@gmail.com>
Date: Mon, 2 Feb 2026 02:16:00 -0300
Subject: [PATCH] atlasbot: enforce metric routing for namespace pods

---
 atlasbot/engine/answerer.py | 53 ++++++++++++++++++++++++++++++-------
 1 file changed, 44 insertions(+), 9 deletions(-)

diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index 120e877..a140651 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -227,11 +227,19 @@ class AnswerEngine:
             )
             if any(term in normalized.lower() for term in cluster_terms):
                 classify["needs_snapshot"] = True
+            lowered_norm = normalized.lower()
+            if (
+                ("namespace" in lowered_norm and ("pod" in lowered_norm or "pods" in lowered_norm))
+                or re.search(r"\bmost\s+pods\b", lowered_norm)
+                or re.search(r"\bpods\s+running\b", lowered_norm)
+            ):
+                classify["question_type"] = "metric"
+                classify["needs_snapshot"] = True
             if re.search(r"\b(how many|count|number of|list)\b", normalized.lower()):
                 classify["question_type"] = "metric"
             hottest_terms = ("hottest", "highest", "lowest", "most")
-            metric_terms = ("cpu", "ram", "memory", "net", "network", "io", "disk", "load", "usage")
-            lowered_question = normalized.lower()
+            metric_terms = ("cpu", "ram", "memory", "net", "network", "io", "disk", "load", "usage", "pod", "pods", "namespace")
+            lowered_question = f"{question} {normalized}".lower()
             if any(term in lowered_question for term in hottest_terms) and any(term in lowered_question for term in metric_terms):
                 classify["question_type"] = "metric"
 
@@ -312,15 +320,24 @@ class AnswerEngine:
                         if self._settings.debug_pipeline:
                             _debug_log("metric_facts_selected", {"facts": metric_facts})
                 if classify.get("question_type") in {"metric", "diagnostic"} and not metric_facts:
-                    for line in summary_lines:
-                        if "hardware_usage_top:" in line:
-                            metric_facts = [seg.strip() for seg in line.split(" | ") if seg.strip().startswith("hardware_usage_top:")]
-                            break
-                    if not metric_facts:
+                    lowered_q = f"{question} {normalized}".lower()
+                    if "namespace" in lowered_q and "pod" in lowered_q:
                         for line in summary_lines:
-                            if "hardware_usage_avg:" in line:
-                                metric_facts = [seg.strip() for seg in line.split(" | ") if seg.strip().startswith("hardware_usage_avg:")]
+                            if line.startswith("namespaces_top:"):
+                                metric_facts = [line]
                                 break
+                    if not metric_facts:
+                        hardware_tokens = ("hardware", "class", "type", "rpi", "jetson", "amd64", "arm64")
+                        if any(tok in lowered_q for tok in hardware_tokens):
+                            for line in summary_lines:
+                                if "hardware_usage_top:" in line:
+                                    metric_facts = [seg.strip() for seg in line.split(" | ") if seg.strip().startswith("hardware_usage_top:")]
+                                    break
+                            if not metric_facts:
+                                for line in summary_lines:
+                                    if "hardware_usage_avg:" in line:
+                                        metric_facts = [seg.strip() for seg in line.split(" | ") if seg.strip().startswith("hardware_usage_avg:")]
+                                        break
                     if metric_facts:
                         key_facts = _merge_fact_lines(metric_facts, key_facts)
                 if self._settings.debug_pipeline:
@@ -513,6 +530,19 @@ class AnswerEngine:
             if classify.get("question_type") in {"metric", "diagnostic"} and metric_facts:
                 reply = _metric_fact_guard(reply, metric_facts, keyword_tokens)
 
+            if classify.get("question_type") in {"metric", "diagnostic"}:
+                lowered_q = f"{question} {normalized}".lower()
+                if ("io" in lowered_q or "i/o" in lowered_q) and ("node" in lowered_q or "nodes" in lowered_q):
+                    io_facts = _extract_hottest_facts(summary_lines, lowered_q)
+                    io_line = next((fact for fact in io_facts if fact.startswith("hottest_io_node")), None)
+                    if io_line:
+                        reply = f"From the latest snapshot: {io_line}."
+                if "namespace" in lowered_q and "pod" in lowered_q:
+                    for line in summary_lines:
+                        if line.startswith("namespaces_top:"):
+                            reply = f"From the latest snapshot: {line}."
+                            break
+
             if classify.get("question_type") in {"metric", "diagnostic"}:
                 lowered_q = f"{question} {normalized}".lower()
                 if any(tok in lowered_q for tok in ("hardware", "class", "type", "rpi", "jetson", "amd64", "arm64")) and any(
@@ -1091,6 +1121,8 @@ def _extract_hottest_facts(lines: list[str], question: str) -> list[str]:
     if not facts:
         return []
     wanted = []
+    if ("io" in lowered or "i/o" in lowered) and ("disk" in lowered or "storage" in lowered):
+        return [fact for fact in facts if fact.startswith("hottest_io_node")]
     if any(term in lowered for term in ("cpu", "processor")):
         wanted.append("hottest_cpu_node")
     if any(term in lowered for term in ("ram", "memory")):
@@ -1156,6 +1188,9 @@ def _metric_candidate_lines(lines: list[str], keywords: list[str] | None, limit:
         "p95",
         "percent",
         "pressure",
+        "pod",
+        "pods",
+        "namespace",
     }
     candidates: list[str] = []
     for line in lines: