Skip empty doc files and prune dead type-index entries

- render_header() returns "" when a header has no documented content
  (no /** */ comments on any class, property, function, enum, or delegate)
- generate.py skips writing those files and tracks which were written
- type-index.txt is filtered to only include types from written files,
  preventing dead entries that would cause get_class_overview to fail
- Summary line now reports how many files were skipped

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

CLAUDE.md

@@ -55,6 +55,7 @@ Typical query flow: `search_types` → `get_class_overview` → `get_member` (~3
 - Deprecated functions excluded from output (`not f.is_deprecated` in visibility filter).
 - Enums with no value descriptions use compact inline format (`Values: A, B, C`) instead of a table.
 - Placeholder descriptions (`------//`) filtered by `_PLACEHOLDER_RE` in `_fn_body()`.
+- Headers with no documented content: `render_header()` returns `""`, `generate.py` skips writing the file and filters those types from `type-index.txt` to avoid dead entries.
 
 ## Critical Gotchas
 

README.md

@@ -5,8 +5,8 @@ Generates compact, agent-readable Markdown documentation from Unreal Engine C++
 ## How it works
 
 1. **Parse** — `docgen/ue_parser.py` scans UE headers into dataclasses using a position-based scanner (handles nested braces, macros, delegates, namespaces).
-2. **Render** — `docgen/ue_markdown.py` emits one `.md` per header: only items with C++ doc comments, no deprecated functions, compact enum format.
-3. **Index** — `docgen/generate.py` produces `type-index.txt`: a flat `TypeName: path/to/File.md` lookup for instant type resolution.
+2. **Render** — `docgen/ue_markdown.py` emits one `.md` per header: only items with C++ doc comments, no deprecated functions, compact enum format. Headers with no documented content produce no output file.
+3. **Index** — `docgen/generate.py` produces `type-index.txt`: a flat `TypeName: path/to/File.md` lookup for instant type resolution. Only types from files that were actually written are indexed.
 4. **Serve** — `ue_mcp_server.py` exposes the docs to Claude as callable MCP tools.
 
 ## Usage

docgen/generate.py

@@ -135,6 +135,8 @@ def main():
 
     # --- Pass 2: render all ---
     success = 0
+    skipped = 0
+    written_mds: set[str] = set()
     for h, base, parsed in parsed_list:
         print(f"Rendering {h} ...")
         current_md = _md_rel(h, base)
@@ -143,13 +145,20 @@ def main():
 
         try:
             md = render_header(parsed, type_index=type_index, current_md=current_md)
+            if not md:
+                skipped += 1
+                continue
             out_path.write_text(md, encoding='utf-8')
+            written_mds.add(current_md)
             success += 1
         except Exception as exc:
             print(f"  ERROR rendering {h}: {exc}", file=sys.stderr)
 
+    # Remove type-index entries whose files were not written (no documented content)
+    type_index = {name: path for name, path in type_index.items() if path in written_mds}
     write_type_index(type_index, output_dir)
-    print(f"\nGenerated {success}/{len(parsed_list)} files + type-index.txt in {output_dir}/")
+    print(f"\nGenerated {success}/{len(parsed_list)} files "
+          f"({skipped} skipped — no documented content) + type-index.txt in {output_dir}/")
 
 
 if __name__ == '__main__':

docgen/ue_markdown.py

@@ -361,14 +361,13 @@ def _render_namespace(ns: NamespaceInfo) -> str:
 def render_header(parsed: ParsedHeader,
                   type_index: dict[str, str] = None,
                   current_md: str = "") -> str:
+    """
+    Render a ParsedHeader to Markdown. Returns empty string if the header
+    has no documented content (so callers can skip writing the file).
+    """
     if type_index is None:
         type_index = {}
 
-    lines = []
-    lines.append(f"# `{parsed.filename}`")
-    lines.append(f"**Module**: `{parsed.module_name}`")
-    lines.append("")
-
     sections = []
 
     d_sec = _render_delegates(parsed.delegates)
@@ -400,6 +399,13 @@ def render_header(parsed: ParsedHeader,
                 ff_lines.append(_render_ff_compact(fn))
             sections.append('\n'.join(ff_lines))
 
-    lines.append('\n\n---\n\n'.join(sections))
+    if not sections:
+        return ""
 
+    lines = [
+        f"# `{parsed.filename}`",
+        f"**Module**: `{parsed.module_name}`",
+        "",
+        '\n\n---\n\n'.join(sections),
+    ]
     return '\n'.join(lines)