# models/skill_extractor.py
from huggingface_hub import snapshot_download
import spacy

class SkillExtractor:
    def __init__(self):
        # Load Hugging Face pretrained skill extractor
        model_path = snapshot_download("amjad-awad/skill-extractor")
        self.nlp = spacy.load(model_path)

    def extract_skills(self, text):
        """
        Extract skills from text and merge multi-word skills.
        """
        doc = self.nlp(text)
        skills = []
        current_skill = []
        prev_end = None

        for ent in doc.ents:
            if "SKILL" in ent.label_:
                if prev_end is not None and ent.start_char == prev_end + 1:
                    current_skill.append(ent.text)
                else:
                    if current_skill:
                        skills.append(" ".join(current_skill))
                    current_skill = [ent.text]
                prev_end = ent.end_char
            else:
                if current_skill:
                    skills.append(" ".join(current_skill))
                    current_skill = []
                prev_end = None

        if current_skill:
            skills.append(" ".join(current_skill))

        # Remove duplicates
        skills = list(dict.fromkeys(skills))
        return skills