Nl blend adjuster

[PavanRaghavendraKulkarni]

Pull Request

Description

Adjuster Blend Pass

Adds an “adjuster” step to the NL blend pipeline. After the main blend completes, the same blend is run again using each model’s {model_name}_adjust forecaster on the Data Platform. The result is saved under {forecaster_name}_adjust.

Changes

• blend/app.py
  • Extracted _run_blend_pass() coroutine to handle the full pipeline:
    fetch weights → blend → save
  • run_blend_app() now calls this twice when use_adjuster=True:
    • once for main models
    • once for _adjust models
  • During the adjuster pass, weight column names are suffixed with _adjust so the Data Platform fetches the correct forecasters
• blend/config.py
  Added:
  • use_adjuster: bool (default: False)
  • adjuster_forecaster_name property in NlBlendConfig

Fixes #

How Has This Been Tested?

Unit tests were run locally with uv run pytest tests/unit/test_nl_adjuster.py tests/unit/test_nl_app.py

Checklist:

• My code follows OCF's coding style guidelines
• I have performed a self-review of my own code
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• I have checked my code and corrected any misspellings