swarmauri_evaluatorpool_accessibility 0.3.0.dev17

A package providing accessibility and readability evaluators for SwarmAuri

---

Swarmauri Evaluatorpool Accessibility

Accessibility-focused evaluator pool for Swarmauri programs. The pool bundles a set of classic readability metrics (Automated Readability Index, Coleman–Liau, Flesch–Kincaid Grade, Flesch Reading Ease, and Gunning Fog) and normalises their scores onto a single 0–1 scale where higher values mean easier-to-read content.

Installation

pip

pip install swarmauri_evaluatorpool_accessibility

Poetry

poetry add swarmauri_evaluatorpool_accessibility

uv

# Install into the current environment
uv pip install swarmauri_evaluatorpool_accessibility

# Or add it to the project pyproject.toml
uv add swarmauri_evaluatorpool_accessibility

Usage

AccessibilityEvaluatorPool is a registered EvaluatorPoolBase component. It expects a sequence of Swarmauri Program objects, runs the configured evaluators concurrently, then returns standard EvalResultBase instances with the aggregated score and detailed metadata for every evaluator.

Example

from swarmauri_evaluatorpool_accessibility.AccessibilityEvaluatorPool import AccessibilityEvaluatorPool
from swarmauri_standard.programs.Program import Program

program = Program(content={"example.txt": "This is a simple sentence. Here is another one."})

pool = AccessibilityEvaluatorPool()
pool.initialize()  # spin up the thread pool once before evaluating

results = pool.evaluate([program])
result = results[0]

print(f"Aggregate accessibility score: {result.score:.2f}")
print("Per-evaluator scores:", result.metadata["evaluator_scores"])
flesch_meta = result.metadata["evaluator_metadata"].get(
    "FleschReadingEaseEvaluator", {}
)
print(
    "Flesch interpretation:",
    flesch_meta.get("readability_interpretation", flesch_meta.get("error", "not available")),
)

pool.shutdown()  # release resources when finished

Result structure

Each entry returned from evaluate is an EvalResultBase with:

• score: the weighted mean of the evaluator scores mapped to the 0–1 range.
• metadata["evaluator_scores"]: raw scores keyed by evaluator name.
• metadata["evaluator_metadata"]: evaluator-specific metadata (e.g. sentence counts, syllable estimates, interpretation strings).
• metadata["aggregate_score"]: mirrors .score for convenience.

You can customise the weighting of each evaluator by passing a weights dictionary (name → float) when constructing the pool.

Notes

• The pool automatically registers the bundled evaluators when no explicit list is provided. You can pass your own evaluators that satisfy the IEvaluate interface to override or extend the defaults.
• FleschReadingEaseEvaluator downloads NLTK's punkt tokenizer and cmudict pronunciation dictionary on first use. Set NLTK_DATA_DIR to point to a writable cache location if required. When these resources are not available the evaluator reports an "error" entry in its metadata instead of raising an exception.

Want to help?

If you want to contribute to swarmauri-sdk, read up on our guidelines for contributing that will help you get started.