feat: add rotation detection parameter (#110)

Add to `ParseConfig` the `enable_rotation_detection: bool` parameter, to
automatically detect if document pages are rotated or not.

This also adds to the response `metadata` which pages were actually
detected as rotate, and which rotation was detected.

Author: pappacena

agentic_doc/common.py

@@ -134,6 +134,9 @@ class DocumentMetadata(BaseModel):
     processing_time_ms: Optional[int] = None
     pages_processed: Optional[int] = None
     user_id: Optional[str] = None
+    pages_rotation_angles: Optional[dict[str, float]] = Field(
+        default_factory=dict,  # type: ignore[arg-type]
+    )
 
 
 class ParsedDocument(BaseModel, Generic[T]):

agentic_doc/config.py

@@ -33,6 +33,7 @@ def __init__(
         extraction_schema: Optional[dict[str, Any]] = None,
         split_size: Optional[int] = None,
         extraction_split_size: Optional[int] = None,
+        enable_rotation_detection: Optional[bool] = None,
     ) -> None:
         self.api_key = api_key
         self.include_marginalia = include_marginalia
@@ -41,6 +42,7 @@ def __init__(
         self.extraction_schema = extraction_schema
         self.split_size = split_size
         self.extraction_split_size = extraction_split_size
+        self.enable_rotation_detection = enable_rotation_detection
 
 
 class SettingsOverrides:

agentic_doc/parse.py

@@ -781,6 +781,8 @@ def _send_parsing_request(
                 "include_marginalia": include_marginalia,
                 "include_metadata_in_markdown": include_metadata_in_markdown,
             }
+            if config and config.enable_rotation_detection is not None:
+                data["enable_rotation_detection"] = config.enable_rotation_detection
 
             def resolve_refs(obj: Any, defs: Dict[str, Any]) -> Any:
                 if isinstance(obj, dict):