Add .docx support to RAG document loader

fenn/agents/rag/loader.py

@@ -6,6 +6,7 @@
     ".txt",
     ".md",
     ".pdf",
+    ".docx",
     ".json",
     ".py",
     ".js",
@@ -95,6 +96,8 @@ def _read_file(path):
     try:
         if path.suffix == ".pdf":
             return _read_pdf(path)
+        if path.suffix == ".docx":
+            return _read_docx(path)
         return path.read_text(encoding="utf-8")
     except (OSError, UnicodeDecodeError, ValueError, RuntimeError) as e:
         logger.error(f"[cofone] read error {path.name}: {e}")
@@ -124,6 +127,33 @@ def _read_pdf(path):
         )
 
 
+def _read_docx(path):
+    """
+    Extract text from a Word (.docx) file using python-docx.
+    Reads both paragraphs and table cells.
+    Requires: pip install "cofone[docx]"  or  pip install python-docx
+    """
+    try:
+        import docx
+
+        document = docx.Document(str(path))
+        parts = [p.text for p in document.paragraphs]
+        for table in document.tables:
+            for row in table.rows:
+                parts.append("\t".join(cell.text for cell in row.cells))
+        text = "\n".join(parts).strip()
+        if not text:
+            logger.warning(
+                f"[cofone] warning: DOCX '{path.name}' returned no text (empty document)"
+            )
+        return text or None
+    except ImportError:
+        raise ImportError(
+            "[cofone] python-docx not installed.\n"
+            'Run: pip install "cofone[docx]"  or  pip install python-docx'
+        )
+
+
 def _load_url(url):
     """
     Fetch a web page and return its visible text content.

pyproject.toml

@@ -107,11 +107,13 @@ llms = [
 ]
 
 rag-pdf   = ["pypdf>=4.0.0"]
+rag-docx  = ["python-docx>=1.0.0"]
 rag-faiss = ["faiss-cpu>=1.7.0", "sentence-transformers>=2.0.0"]
 rag-web   = ["wikipedia>=1.4.0", "youtube-transcript-api>=0.6.0"]
 
 rag-all  = [
     "pypdf>=4.0.0",
+    "python-docx>=1.0.0",
     "faiss-cpu>=1.7.0",
     "sentence-transformers>=2.0.0",
     "wikipedia>=1.4.0",

tests/unit/agents/test_docx_loader.py

@@ -0,0 +1,72 @@
+"""Tests for the .docx loader in fenn/agents/rag/loader.py
+
+Place at: tests/unit/agents/test_docx_loader.py
+(or merge these cases into tests/unit/agents/test_rag_components.py).
+"""
+
+from unittest.mock import patch
+
+import pytest
+
+from fenn.agents.rag.loader import (
+    _read_docx,
+    _read_file,
+    load_documents,
+)
+
+
+def _make_docx(path, paragraphs=("Hello from docx", "second line"), table=None):
+    docx = pytest.importorskip("docx")
+    d = docx.Document()
+    for p in paragraphs:
+        d.add_paragraph(p)
+    if table:
+        t = d.add_table(rows=len(table), cols=len(table[0]))
+        for r, row in enumerate(table):
+            for c, val in enumerate(row):
+                t.rows[r].cells[c].text = val
+    d.save(str(path))
+    return path
+
+
+class TestReadDocx:
+    def test_reads_paragraphs(self, tmp_path):
+        f = _make_docx(tmp_path / "doc.docx")
+        assert _read_docx(f) == "Hello from docx\nsecond line"
+
+    def test_reads_table_cells(self, tmp_path):
+        f = _make_docx(
+            tmp_path / "t.docx",
+            paragraphs=("Prices:",),
+            table=[["Item", "Price"], ["Coffee", "3.00"]],
+        )
+        text = _read_docx(f)
+        assert "Item\tPrice" in text
+        assert "Coffee\t3.00" in text
+
+    def test_empty_docx_returns_none(self, tmp_path):
+        docx = pytest.importorskip("docx")
+        d = docx.Document()
+        d.save(str(tmp_path / "empty.docx"))
+        assert _read_docx(tmp_path / "empty.docx") is None
+
+    def test_read_file_delegates_docx(self, tmp_path):
+        f = _make_docx(tmp_path / "d.docx")
+        assert _read_file(f) == "Hello from docx\nsecond line"
+
+    def test_load_documents_reads_docx(self, tmp_path):
+        f = _make_docx(tmp_path / "d.docx")
+        assert load_documents(str(f)) == ["Hello from docx\nsecond line"]
+
+    def test_missing_python_docx_raises_importerror(self, tmp_path):
+        f = _make_docx(tmp_path / "d.docx")
+        real_import = __import__
+
+        def fake_import(name, *args, **kwargs):
+            if name == "docx":
+                raise ImportError("no docx")
+            return real_import(name, *args, **kwargs)
+
+        with patch("builtins.__import__", side_effect=fake_import):
+            with pytest.raises(ImportError, match="python-docx"):
+                _read_docx(f)