⬇️(markitdown) enforce old 0.0.2 version

Our base docker image (alpine) does not support magika dependency
because there is no wheel for onnxruntime for alpine...
It's complex to build and a bit hackish.

Switching to a slim (debian) base image, is a solution but trivy
scan detects CVEs. We need to check the CVEs don't regard our stack
but it's bit of work also.

For the time being, I pin the markitdown version to an old one...
This is sad, since a lot of bugs has been fixed.
This commit is contained in:
Quentin BEY
2025-08-06 15:31:11 +02:00
parent a9883a1a31
commit aafaaeed3e
2 changed files with 2 additions and 2 deletions
@@ -12,7 +12,7 @@ class DocumentConverter:
def __init__(self):
"""Initialize the DocumentConverter with MarkItDown."""
self.converter = MarkItDown(enable_plugins=False)
self.converter = MarkItDown()
def convert_raw( # pylint: disable=unused-argument
self,
+1 -1
View File
@@ -50,7 +50,7 @@ dependencies = [
"jsonschema==4.24.0",
"lxml==5.4.0",
"markdown==3.8",
"markitdown[pptx,docx,xlsx,xls,pdf,outlook]==0.1.2",
"markitdown==0.0.2",
"mozilla-django-oidc==4.0.1",
"nested-multipart-parser==1.5.0",
"pydantic==2.11.7",