Move MIME type to metadata instead of its own field/column

ragna/core/_document.py

@@ -26,15 +26,15 @@ def __init__(
         name: str,
         metadata: dict[str, Any],
         handler: Optional[DocumentHandler] = None,
-        mime_type: str | None = None,
     ):
         self.id = id or uuid.uuid4()
         self.name = name
         self.metadata = metadata
         self.handler = handler or self.get_handler(name)
-        self.mime_type = (
-            mime_type or mimetypes.guess_type(name)[0] or "application/octet-stream"
-        )
+        if "mime_type" not in self.metadata:
+            self.metadata["mime_type"] = (
+                mimetypes.guess_type(name)[0] or "application/octet-stream"
+            )
 
     @staticmethod
     def supported_suffixes() -> set[str]:
@@ -81,11 +81,8 @@ def __init__(
         name: str,
         metadata: dict[str, Any],
         handler: Optional[DocumentHandler] = None,
-        mime_type: str | None = None,
     ):
-        super().__init__(
-            id=id, name=name, metadata=metadata, handler=handler, mime_type=mime_type
-        )
+        super().__init__(id=id, name=name, metadata=metadata, handler=handler)
         if "path" not in self.metadata:
             metadata["path"] = str(ragna.local_root() / "documents" / str(self.id))
 

ragna/deploy/_api.py

@@ -59,7 +59,7 @@ async def get_document_content(
 
         return StreamingResponse(
             io.BytesIO(core_document.read()),
-            media_type=core_document.mime_type,
+            media_type=core_document.metadata["mime_type"],
             headers=headers,
         )
 

ragna/deploy/_database.py

@@ -288,7 +288,6 @@ def document(
             user_id=user_id,
             name=document.name,
             metadata_=document.metadata,
-            mime_type=document.mime_type,
         )
 
     def source(self, source: schemas.Source) -> orm.Source:
@@ -358,7 +357,6 @@ def document(self, document: orm.Document) -> schemas.Document:
             id=document.id,
             name=document.name,
             metadata=document.metadata_,
-            mime_type=document.mime_type,
         )
 
     def source(self, source: orm.Source) -> schemas.Source:

ragna/deploy/_engine.py

@@ -286,7 +286,6 @@ def document(self, document: schemas.Document) -> core.Document:
             id=document.id,
             name=document.name,
             metadata=document.metadata,
-            mime_type=document.mime_type,
         )
 
     def source(self, source: schemas.Source) -> core.Source:
@@ -335,7 +334,6 @@ def document(self, document: core.Document) -> schemas.Document:
             id=document.id,
             name=document.name,
             metadata=document.metadata,
-            mime_type=document.mime_type,
         )
 
     def source(self, source: core.Source) -> schemas.Source:

ragna/deploy/_orm.py

@@ -103,7 +103,6 @@ class Document(Base):
     # Mind the trailing underscore here. Unfortunately, this is necessary, because
     # metadata without the underscore is reserved by SQLAlchemy
     metadata_ = Column(Json, nullable=False)
-    mime_type = Column(types.String, nullable=False)
     chats = relationship(
         "Chat",
         secondary=document_chat_association_table,

ragna/deploy/_schemas.py

@@ -84,7 +84,6 @@ class Document(BaseModel):
     id: uuid.UUID = Field(default_factory=uuid.uuid4)
     name: str
     metadata: dict[str, Any]
-    mime_type: str
 
 
 class Source(BaseModel):

tests/deploy/api/test_e2e.py

@@ -112,7 +112,12 @@ def test_e2e(tmp_local_root, multiple_answer_chunks, stream_answer, corpus_name)
         )
         assert corpus_name in corpuses_metadata[source_storage]
         metadata_keys = corpuses_metadata[source_storage][corpus_name].keys()
-        assert list(metadata_keys) == ["document_id", "document_name", "path"]
+        assert list(metadata_keys) == [
+            "document_id",
+            "document_name",
+            "mime_type",
+            "path",
+        ]
         for key in metadata_keys:
             assert corpuses_metadata[source_storage][corpus_name][key][0] == "str"
 
@@ -126,7 +131,12 @@ def test_e2e(tmp_local_root, multiple_answer_chunks, stream_answer, corpus_name)
         )
         assert corpus_name in corpuses_metadata[source_storage]
         metadata_keys = corpuses_metadata[source_storage][corpus_name].keys()
-        assert list(metadata_keys) == ["document_id", "document_name", "path"]
+        assert list(metadata_keys) == [
+            "document_id",
+            "document_name",
+            "mime_type",
+            "path",
+        ]
         for key in metadata_keys:
             assert corpuses_metadata[source_storage][corpus_name][key][0] == "str"