handler.py

import spacy

from typing import Any


def _load_spacy_model(name: str) -> spacy.Language:
    if not hasattr(_load_spacy_model, "nlp"):
        # pipeline info https://spacy.io/models/en#en_core_web_lg
        all_except_ner: list[str] = [
            "tok2vec",
            "tagger",
            "parser",
            "attribute_ruler",
            "lemmatizer"]

        nlp = spacy.load(name=name, exclude=all_except_ner)
        _load_spacy_model.nlp = nlp
        print(f"Loaded {nlp.meta.get('name', 'unknown')} model from {nlp.path}")

    return _load_spacy_model.nlp


class EndpointHandler:
    def __init__(self, path: str = ""):
        print(f"EndpointHandler(path='{path}')")
        self._nlp: spacy.Language = _load_spacy_model(name="en_core_web_lg")

    def __call__(self, data: dict[str, Any]) -> list[dict[str, Any]]:
        inputs: str = data.pop("inputs", "")
        if not inputs:
            return []

        outputs: list[dict[str, Any]] = []

        doc = self._nlp(text=inputs)
        for ent in doc.ents:
            if ent.label_ != "PERSON":
                continue

            entity: dict = {
                "entity_group": ent.label_,
                "score": 1,
                "word": ent.text,
                "start": ent.start_char,
                "end": ent.end_char}

            outputs.append(entity)

        return outputs