From 60ab580f8b1f2c20cffa650b707948ff9792c6fc Mon Sep 17 00:00:00 2001
From: Xubin Ren <52506698+Re-bin@users.noreply.github.com>
Date: Wed, 22 Jul 2026 22:57:38 +0800
Subject: [PATCH] fix(documents): preserve DOCX table content

---
 nanobot/utils/document.py       | 30 ++++++++++++++++++++++++++----
 tests/test_context_documents.py | 20 ++++++++++++++++++++
 tests/test_document_parsing.py  | 23 +++++++++++++++++++++++
 3 files changed, 69 insertions(+), 4 deletions(-)

diff --git a/nanobot/utils/document.py b/nanobot/utils/document.py
index fec6ed4b..5570a1ee 100644
--- a/nanobot/utils/document.py
+++ b/nanobot/utils/document.py
@@ -210,6 +210,8 @@ def _extract_docx(path: Path) -> str:
     """Extract text from DOCX using python-docx."""
     try:
         from docx import Document as DocxDocument
+        from docx.table import Table
+        from docx.text.paragraph import Paragraph
     except ImportError:
         return "[error: python-docx not installed]"
     try:
@@ -217,10 +219,30 @@ def _extract_docx(path: Path) -> str:
             return error
         doc = DocxDocument(path)
         collector = _TextCollector(_MAX_TEXT_LENGTH)
-        for paragraph in doc.paragraphs:
-            text = paragraph.text.strip()
-            if text and not collector.add(text, separator="\n\n"):
-                break
+        for block in doc.iter_inner_content():
+            if isinstance(block, Paragraph):
+                text = block.text.strip()
+                if text and not collector.add(text, separator="\n\n"):
+                    break
+                continue
+            if not isinstance(block, Table):
+                continue
+            first_row = True
+            for row in block.rows:
+                cells: list[str] = []
+                seen_cells: set[int] = set()
+                for cell in row.cells:
+                    cell_id = id(cell._tc)
+                    if cell_id in seen_cells:
+                        continue
+                    seen_cells.add(cell_id)
+                    cells.append(" ".join(cell.text.split()))
+                if not any(cells):
+                    continue
+                separator = "\n\n" if first_row else "\n"
+                first_row = False
+                if not collector.add("\t".join(cells), separator=separator):
+                    return collector.render()
         return collector.render()
     except Exception as e:
         logger.exception("Failed to extract DOCX {}", path)
diff --git a/tests/test_context_documents.py b/tests/test_context_documents.py
index 7d9ac908..b90abb4d 100644
--- a/tests/test_context_documents.py
+++ b/tests/test_context_documents.py
@@ -93,6 +93,26 @@ def test_drain_pending_path_preserves_document_text(tmp_path: Path) -> None:
     assert "summarize" in result
 
 
+def test_drain_pending_path_preserves_docx_table_text(tmp_path: Path) -> None:
+    """Uploaded Word forms must retain content stored in table cells."""
+    from docx import Document
+
+    doc = Document()
+    table = doc.add_table(rows=2, cols=2)
+    table.cell(0, 0).text = "Applicant"
+    table.cell(0, 1).text = "Ada Lovelace"
+    table.cell(1, 0).text = "Research area"
+    table.cell(1, 1).text = "Analytical engines"
+    docx_path = tmp_path / "application.docx"
+    doc.save(docx_path)
+
+    content, image_only = extract_documents("summarize", [str(docx_path)])
+
+    assert image_only == []
+    assert "Applicant\tAda Lovelace" in content
+    assert "Research area\tAnalytical engines" in content
+
+
 def test_drain_pending_path_without_extract_loses_document(tmp_path: Path) -> None:
     """Demonstrates the BUG: if _drain_pending calls _build_user_content
     directly without extract_documents, document content is lost."""
diff --git a/tests/test_document_parsing.py b/tests/test_document_parsing.py
index be4e996f..5657e784 100644
--- a/tests/test_document_parsing.py
+++ b/tests/test_document_parsing.py
@@ -174,6 +174,29 @@ class TestExtractText:
         assert "This is paragraph one." in result
         assert "This is paragraph two." in result
 
+    def test_extract_text_docx_preserves_paragraph_and_table_order(self, tmp_path: Path):
+        """DOCX forms commonly keep nearly all meaningful content in tables."""
+        from docx import Document
+
+        docx_file = tmp_path / "form.docx"
+        doc = Document()
+        doc.add_paragraph("Applicant details")
+        table = doc.add_table(rows=2, cols=2)
+        table.cell(0, 0).text = "Name"
+        table.cell(0, 1).text = "Ada Lovelace"
+        table.cell(1, 0).text = "Project"
+        table.cell(1, 1).text = "Analytical Engine"
+        doc.add_paragraph("End of form")
+        doc.save(docx_file)
+
+        result = extract_text(docx_file)
+
+        assert result is not None
+        assert "Name\tAda Lovelace" in result
+        assert "Project\tAnalytical Engine" in result
+        assert result.index("Applicant details") < result.index("Name\tAda Lovelace")
+        assert result.index("Analytical Engine") < result.index("End of form")
+
     def test_extract_text_docx_empty(self, tmp_path: Path):
         """Test extracting text from an empty .docx file."""
         from docx import Document