fix(tree_renderer): render real per-node source text, not just the summary

openkb/tree_renderer.py

@@ -2,8 +2,27 @@
 
 from __future__ import annotations
 
+import re
+
 from openkb import frontmatter
 
+# PageIndex's own include_text extraction embeds image links into its private
+# cache (.openkb/files/{doc_id}/images/...), not wiki/sources/images/ where
+# OpenKB's own page extraction saves them — those paths don't resolve from a
+# wiki page and aren't part of the Obsidian vault. Strip them from the quoted
+# source text; the real per-page images are referenced from
+# wiki/sources/{doc_name}.json instead.
+_PAGEINDEX_IMAGE_RE = re.compile(r"!\[[^\]]*\]\([^)]*\)\n?")
+
+
+def _strip_internal_image_refs(text: str) -> str:
+    return _PAGEINDEX_IMAGE_RE.sub("", text)
+
+
+def _quote_block(text: str) -> str:
+    """Render ``text`` as a Markdown blockquote, one ``>`` per line."""
+    return "\n".join(f"> {line}" if line else ">" for line in text.strip().splitlines())
+
 
 def _yaml_frontmatter(source_name: str, doc_id: str, description: str = "") -> str:
     """Return a YAML frontmatter block for a PageIndex wiki page."""
@@ -16,19 +35,24 @@ def _yaml_frontmatter(source_name: str, doc_id: str, description: str = "") -> s
 
 
 def _render_nodes_summary(nodes: list[dict], depth: int) -> str:
-    """Recursively render nodes for the *summary* view (summaries only)."""
+    """Recursively render nodes for the *summary* view (summary + source text)."""
     lines: list[str] = []
     heading_prefix = "#" * min(depth, 6)
     for node in nodes:
         title = node.get("title", "")
         start = node.get("start_index", "")
         end = node.get("end_index", "")
         summary = node.get("summary", "")
+        text = node.get("text", "")
         children = node.get("nodes", [])
 
         lines.append(f"{heading_prefix} {title} (pages {start}–{end})\n")
         if summary:
             lines.append(f"Summary: {summary}\n")
+        stripped_text = _strip_internal_image_refs(text).strip() if text else ""
+        if stripped_text:
+            lines.append("Source text:\n")
+            lines.append(_quote_block(stripped_text) + "\n")
         if children:
             lines.append(_render_nodes_summary(children, depth + 1))
 

tests/test_tree_renderer.py

@@ -31,12 +31,45 @@ def test_page_range_included(self, sample_tree):
         assert "(pages 0–120)" in output
         assert "(pages 121–200)" in output
 
-    def test_summary_included_not_text(self, sample_tree):
+    def test_summary_and_source_text_both_included(self, sample_tree):
         output = render_summary_md(sample_tree, "Sample Document", "doc-abc")
         assert "Summary: Overview of the document topic." in output
         assert "Summary: Historical context." in output
-        # Raw text should NOT appear in summary view
-        assert "This document introduces the core concepts of the system." not in output
+        # The real per-node source text is now quoted too, not just a
+        # paraphrase — IndexConfig(if_add_node_text=True) already fetches
+        # it, the old renderer just silently discarded it.
+        assert "Source text:" in output
+        assert "> This document introduces the core concepts of the system." in output
+
+    def test_node_without_text_has_no_source_text_block(self):
+        tree = {
+            "structure": [
+                {"title": "Intro", "start_index": 1, "end_index": 2, "summary": "x", "nodes": []}
+            ]
+        }
+        output = render_summary_md(tree, "my-doc", "doc-123")
+        assert "Source text:" not in output
+
+    def test_internal_pageindex_image_refs_are_stripped_from_source_text(self):
+        # PageIndex's own image refs point into its private
+        # .openkb/files/{doc_id}/images/... cache, which never resolves from
+        # a wiki page, so they're stripped rather than quoted verbatim.
+        tree = {
+            "structure": [
+                {
+                    "title": "Intro",
+                    "start_index": 1,
+                    "end_index": 2,
+                    "summary": "x",
+                    "text": "Some text.\n![fig](/private/cache/img.png)\nMore text.",
+                    "nodes": [],
+                }
+            ]
+        }
+        output = render_summary_md(tree, "my-doc", "doc-123")
+        assert "![fig]" not in output
+        assert "> Some text." in output
+        assert "> More text." in output
 
 
 def test_summary_md_has_type_and_description():