ue-sdk-helper

Utilities for the Uncertainty Engine ecosystem.

Table of contents

• ue-sdk-helper
  • Table of contents
  • Overview
  • Installation
  • Repository structure
  • Development workflow
    • Branch naming
    • Commits
    • Pull requests
      • PR template (paste this into the PR body)
    • PR checklist
  • Testing expectations
  • Code style
  • Local dev quickstart

---

Overview

ue-sdk-helper is a small helper library for the Uncertainty Engine stack. It collects frequently used utilities (resource interaction, workflow rappers, plotting helpers, data generators, notebook ergonomics) so downstream projects can depend on one place instead of re-implementing the same snippets.

Installation

pip install git+https://github.com/digiLab-ai/ue-sdk-helper.git@main
# or, from source
pip install -e ".[dev]"

The project depends on the Uncertainty Engine:

pip install "uncertainty-engine" matplotlib numpy pandas seaborn

Repository structure

ue-sdk-helper/
├─ src/
│  └─ ue_sdk_helper/
│     ├─ __init__.py
│     ├─ plotting.py
│     ├─ workflows.py
│     └─ notebook.py
├─ tests/
│  ├─ test_plotting.py
├─ notebooks/
│  └─ 00_client.ipynb
├─ pyproject.toml
├─ README.md
├─ LICENSE
├─ .gitignore
├─ .pre-commit-config.yaml
└─ Makefile

Development workflow

Branch naming

Create a short, descriptive branch that starts with your initials (or GitHub handle) and the objective:

• Feature: dg/feat-dataset-generator
• Fix: dg/fix-NaN-handling-stats
• Chore/CI: dg/chore-ci-precommit
• Issue-scoped: dg/123-fix-bounds-checking (where 123 is the issue number)

Tip: keep branches focused; one logical change per branch.

Commits

• Conventional, present-tense, and scoped where helpful:
  • feat(stats): add weighted_mean with tests
  • fix(io): handle missing .env gracefully
  • docs(readme): add dev workflow
• Keep commits small and meaningful; include rationale when non-obvious.

Pull requests

Open a PR early as a draft; convert to “Ready for review” when:

• All tests pass locally
• Lint/format checks pass
• PR description explains objective, approach, and trade-offs
• Added/updated tests demonstrate the change

PR template (paste this into the PR body)

## Objective
(What problem are we solving? Link issues if applicable.)

## Approach
(What did you implement? Why this design?)

## Tests
(What unit tests were added/updated? Any coverage notes?)

## Breaking changes?
(Yes/No — if yes, describe migration.)

## Screenshots / Notes
(Optional: plots, logs, or snippets useful for reviewers.)

PR checklist

• Branch name uses initials/...
• New code has unit tests
• make format lint test passes locally
• Public APIs documented (docstrings / README / examples)
• No large, unrelated refactors mixed in

Testing expectations

• Unit tests live in tests/ and must cover new code paths.
• Coverage target: ≥ 85% for changed files (keep simple utilities near 100%).
• Use pytest; prefer property-based checks for tiny helpers when sensible.
• Plotting functions: test pure parts (e.g., data preparation) and smoke-test figure creation (no exceptions).

Example test:

# tests/test_stats.py
import numpy as np
from ue_sdk_helper.stats import weighted_mean

def test_weighted_mean_basic():
    x = np.array([1.0, 2.0, 3.0])
    w = np.array([1.0, 1.0, 2.0])
    assert np.isclose(weighted_mean(x, w), 2.25)

def test_weighted_mean_handles_empty():
    assert np.isnan(weighted_mean(np.array([]), np.array([])))

Code style

• Black for formatting
• Ruff for linting (ruff check + ruff format or let Black format)
• Pre-commit hooks run Black, Ruff, end-of-file fixes, and trailing whitespace
• Optional: mypy if/when we add types

Local dev quickstart

# 1) Set up environment
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

# 2) Install pre-commit
pre-commit install

# 3) Run quality gates
make format lint test
# or the raw commands:
# ruff check .
# black --check .
# pytest -q --maxfail=1 --disable-warnings

Example Makefile:

.PHONY: format lint test all
format:
	black src tests
	ruff format src tests
lint:
	ruff check src tests
test:
	pytest -q --maxfail=1 --disable-warnings
all: format lint test