feat: Add DocX parsing strategy to enable configuring convert-to-PDF (#173)

Adds an optional (None-default) `docx_parsing_strategy` to the
ParserConfig to mirror `docx_parsing_strategy` in `ParserConfig` in
[compass-parser](https://github.com/cohere-ai/compass-parser/blob/e0ea790c040de8dc2246c96de63d2598264692dd/parser/compass_parser/parser_types.py#L410).

In conformance with the pattern established by
`presentation_parsing_strategy`, by default the SDK will not pass an
explicit value for `docx_parsing_strategy` and let the server decide.

**Related PRs**
- cohere-ai/compass-parser#247 - Corresponding
compass-parser PR
- #165 - Similar but
for adding convert-to-PDF presentation strategy

**Blocking**
- Once this PR is merged, version bumped & released, [this PR adding e2e
docx parsing eval can be
merged](cohere-ai/compass-eval#21).

Author: jarrah-cohere

cohere_compass/models/config.py

@@ -113,6 +113,19 @@ def _missing_(cls, value: Any):
         return cls.Unstructured
 
 
+class DocxParsingStrategy(str, Enum):
+    """Enum for specifying the parsing strategy for DOCX files."""
+
+    # Uses https://github.com/microsoft/markitdown
+    MarkItDown = "MarkItDown"
+    # Converts the DOCX to PDF and uses the PDF parsing strategy
+    ConvertToPDF = "ConvertToPDF"
+
+    @classmethod
+    def _missing_(cls, value: Any):
+        return cls.MarkItDown
+
+
 class ParsingStrategy(str, Enum):
     """Enum for specifying the parsing strategy to use."""
 
@@ -193,12 +206,12 @@ class ParserConfig(BaseModel):
     vertical_table_crop_margin: int = 100
     horizontal_table_crop_margin: int = 100
 
+    pdf_parsing_config: PDFParsingConfig = PDFParsingConfig()
     pdf_parsing_strategy: PDFParsingStrategy = PDFParsingStrategy.QuickText
     tabular_parsing_strategy: TabularParsingStrategy = TabularParsingStrategy.Granular
-
-    pdf_parsing_config: PDFParsingConfig = PDFParsingConfig()
-
     presentation_parsing_strategy: PresentationParsingStrategy | None = None
+    docx_parsing_strategy: DocxParsingStrategy | None = None
+
     enable_assets_returned_as_base64: bool = True
 
 

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "cohere-compass-sdk"
-version = "2.2.1"
+version = "2.2.2"
 authors = []
 description = "Cohere Compass SDK"
 readme = "README.md"
@@ -93,4 +93,4 @@ omit = [
 ]
 
 [tool.coverage.html]
-directory = "coverage_html"
+directory = "coverage_html"