diff --git a/atlasbot/engine/answerer.py b/atlasbot/engine/answerer.py
index 1ef22e6..6aa82eb 100644
--- a/atlasbot/engine/answerer.py
+++ b/atlasbot/engine/answerer.py
@@ -335,6 +335,12 @@ class AnswerEngine:
                 if observer:
                     observer("retrieve", "scoring chunks")
                 chunks = _chunk_lines(summary_lines, plan.chunk_lines)
+                kb_lines = self._kb.chunk_lines() if self._kb else []
+                if kb_lines:
+                    kb_chunks = _chunk_lines(kb_lines, plan.chunk_lines)
+                    for idx, chunk in enumerate(kb_chunks):
+                        chunk["id"] = f"k{idx}"
+                    chunks.extend(kb_chunks)
                 metric_keys: list[str] = []
                 must_chunk_ids: list[str] = []
                 metric_task = None
@@ -2092,6 +2098,21 @@ def _format_direct_metric_line(line: str) -> str:
         key = key.strip().replace("_", " ")
         value = value.strip()
         if value:
+            if key == "nodes":
+                parts = [p.strip() for p in value.split(",") if p.strip()]
+                total = None
+                rest: list[str] = []
+                for part in parts:
+                    if part.startswith("total="):
+                        total = part.split("=", 1)[1]
+                    else:
+                        rest.append(part.replace("_", " "))
+                if total:
+                    if rest:
+                        return f"Atlas has {total} total nodes ({'; '.join(rest)})."
+                    return f"Atlas has {total} total nodes."
+            if key in {"nodes total", "nodes_total"}:
+                return f"Atlas has {value} total nodes."
             return f"{key} is {value}."
     return line
 
diff --git a/atlasbot/knowledge/loader.py b/atlasbot/knowledge/loader.py
index a6d544b..ae86693 100644
--- a/atlasbot/knowledge/loader.py
+++ b/atlasbot/knowledge/loader.py
@@ -71,3 +71,46 @@ class KnowledgeBase:
             if path:
                 paths.append(str(path))
         return paths[:limit]
+
+    def chunk_lines(self, *, max_files: int = 20, max_chars: int = 6000) -> list[str]:
+        self.load()
+        lines: list[str] = []
+        if not self._base:
+            return lines
+        # Prefer curated catalog JSON if present.
+        if self._atlas:
+            try:
+                atlas_json = json.dumps(self._atlas, indent=2)
+                lines.append("KB: atlas.json")
+                lines.extend(atlas_json.splitlines())
+            except Exception:
+                pass
+        if self._runbooks:
+            lines.append("KB: runbooks.json")
+            for entry in self._runbooks:
+                if not isinstance(entry, dict):
+                    continue
+                title = entry.get("title")
+                path = entry.get("path")
+                if title and path:
+                    lines.append(f"- {title} ({path})")
+        # Include markdown/text sources as additional chunks.
+        if len(lines) >= max_chars:
+            return lines
+        files = sorted(self._base.rglob("*.md")) + sorted(self._base.rglob("*.txt"))
+        for path in files:
+            if len(lines) >= max_chars:
+                break
+            if len(lines) > max_files * 50:
+                break
+            try:
+                text = path.read_text(encoding="utf-8", errors="ignore")
+            except Exception:
+                continue
+            if not text:
+                continue
+            lines.append(f"KB File: {path.relative_to(self._base)}")
+            lines.extend(text.splitlines())
+            if sum(len(line) for line in lines) >= max_chars:
+                break
+        return lines