StandardPdfPipeline.download_models_hf() method not found - Incompatibility with Docling >= 2.28.4

[KabeerThockchom]

Bug Description

When running ilab data generate --pipeline simple, the process fails with the following error:

failed to generate data with exception: type object 'StandardPdfPipeline' has no attribute 'download_models_hf'

Environment

InstructLab version: 0.26.1
instructlab-sdg version: 0.8.3
Docling version: 2.61.1 (required >= 2.28.4)
Platform: macOS (also affects other platforms)

Root Cause

The StandardPdfPipeline.download_models_hf() method has been deprecated and removed from StandardPdfPipeline in newer versions of Docling (>= 2.28.4). The method now only exists in LegacyStandardPdfPipeline with a deprecation warning.

According to Docling's deprecation notice, the recommended approach is to use:

docling.utils.model_downloader.download_models() (programmatic)
docling-tools models download (CLI)

Affected Code

File: chunkers.py (line ~142)

from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline

if self.docling_model_path is None:
    logger.info("Docling models not found on disk, downloading models...")
    self.docling_model_path = StandardPdfPipeline.download_models_hf()

Proposed Fix
Option 1 (Recommended): Use the new Docling API

from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline
from docling.utils.model_downloader import download_models

if self.docling_model_path is None:
    logger.info("Docling models not found on disk, downloading models...")
    self.docling_model_path = download_models(output_dir=None, force=False, progress=False)

Option 2 (Temporary): Use LegacyStandardPdfPipeline

from docling.pipeline.legacy_standard_pdf_pipeline import LegacyStandardPdfPipeline

if self.docling_model_path is None:
    logger.info("Docling models not found on disk, downloading models...")
    self.docling_model_path = LegacyStandardPdfPipeline.download_models_hf()

Steps to Reproduce
Install InstructLab with Docling >= 2.28.4
Set up a taxonomy with knowledge documents
Run ilab data generate --pipeline simple
Observe the error

Impact
This bug prevents users from generating synthetic training data, blocking the entire InstructLab workflow for knowledge-based model fine-tuning.

Workaround

Manually patch the file at: chunkers.py with the proposed fix above.