Support using .docx files (#281)

Co-authored-by: Philip Meier <github.pmeier@posteo.de>

pyproject.toml

@@ -59,6 +59,7 @@ all = [
     "lancedb>=0.2",
     "pyarrow",
     "pymupdf>=1.23.6",
+    "python-docx",
     "tiktoken",
 ]
 
@@ -141,6 +142,7 @@ module = [
     "lancedb",
     "param",
     "pyarrow",
+    "docx",
     "sentence_transformers",
 ]
 ignore_missing_imports = true

ragna/core/__init__.py

@@ -5,6 +5,7 @@
     "Document",
     "DocumentHandler",
     "DocumentUploadParameters",
+    "DocxDocumentHandler",
     "EnvVarRequirement",
     "LocalDocument",
     "Message",
@@ -33,6 +34,7 @@
     Document,
     DocumentHandler,
     DocumentUploadParameters,
+    DocxDocumentHandler,
     LocalDocument,
     Page,
     PdfDocumentHandler,

ragna/core/_document.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import abc
+import io
 import os
 import secrets
 import time
@@ -269,3 +270,34 @@ def extract_pages(self, document: Document) -> Iterator[Page]:
         ) as document:
             for number, page in enumerate(document, 1):
                 yield Page(text=page.get_text(sort=True), number=number)
+
+
+@DOCUMENT_HANDLERS.load_if_available
+class DocxDocumentHandler(DocumentHandler):
+    """Document handler for `.docx` documents.
+
+    !!! note
+
+        This does *not* extract text from headers or footers.
+
+    !!! info "Package requirements"
+
+        - [`python-docx`](https://github.com/python-openxml/python-docx)
+    """
+
+    @classmethod
+    def requirements(cls) -> list[Requirement]:
+        return [PackageRequirement("python-docx")]
+
+    @classmethod
+    def supported_suffixes(cls) -> list[str]:
+        return [".docx"]
+
+    def extract_pages(self, document: Document) -> Iterator[Page]:
+        import docx
+
+        document_docx = docx.Document(io.BytesIO(document.read()))
+        for paragraph in document_docx.paragraphs:
+            text = paragraph.text
+            if len(text) > 0:
+                yield Page(text=text)

tests/core/test_document.py

@@ -0,0 +1,22 @@
+import docx
+
+from ragna.core import DocxDocumentHandler, LocalDocument
+
+
+def get_docx_document(tmp_path, docx_text):
+    document = docx.Document()
+    document.add_heading(docx_text)
+    document.add_paragraph(docx_text)
+    path = tmp_path / "test_document.docx"
+    document.save(path)
+    return LocalDocument.from_path(path)
+
+
+def test_docx(tmp_path):
+    docx_text = "ragna is neat!"
+    tmp_docx_document = get_docx_document(tmp_path, docx_text)
+    assert isinstance(tmp_docx_document.handler, DocxDocumentHandler)
+    pages = list(tmp_docx_document.extract_pages())
+    assert len(pages) == 2
+    for page in pages:
+        assert page.text == docx_text