docling-project / docling-project/docling-core
It looks like this PR and 2.83.0 release might have broken the docling-tools model download command:
- Dominant language
- HTML
- Stars
- 282
- Forks
- 214
- Avg merge
- 2d 15h
- Merged PRs (30d)
- 21
Description
It looks like this PR and 2.83.0 release might have broken the docling-tools model download command:
```
Traceback (most recent call last):
File "/var/lang/bin/docling-tools", line 5, in
from docling.cli.tools import app
File "/var/lang/lib/python3.12/site-packages/docling/cli/tools.py", line 3, in
from docling.cli.models import app as models_app
File "/var/lang/lib/python3.12/site-packages/docling/cli/models.py", line 13, in
from docling.utils.model_downloader import download_models
File "/var/lang/lib/python3.12/site-packages/docling/utils/model_downloader.py", line 17, in
from docling.models.stages.chart_extraction.granite_vision import (
File "/var/lang/lib/python3.12/site-packages/docling/models/stages/chart_extraction/granite_vision.py", line 25, in
from docling.models.base_model import BaseItemAndImageEnrichmentModel
File "/var/lang/lib/python3.12/site-packages/docling/models/base_model.py", line 22, in
from docling.datamodel.document import ConversionResult
File "/var/lang/lib/python3.12/site-packages/docling/datamodel/document.py", line 38, in
from docling_core.types.legacy_doc.base import (
ModuleNotFoundError: No module named 'docling_core.types.legacy_doc'
```
Unless there's a suggested upgrade/workaround?
_Originally posted by @sgottuso-bh in https://github.com/docling-project/docling-core/issues/644#issuecomment-4744310937_
Confirmed that docling-tools models download continues to work if you install docling and pin docling-core to < 2.83.0
Contributor guide
Research direction
Start by reproducing the docling-tools model download command with docling-core 2.83.0 and compare it with a version below 2.83.0. Read docling/cli/tools.py, docling/cli/models.py, docling/utils/model_downloader.py, and the import path in docling/datamodel/document.py. Done means the command starts without the ModuleNotFoundError and completes model downloading without requiring the reported pin.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- cli
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100