src/inference.py

import os
import numpy as np
import joblib
import scipy.sparse
from textblob import TextBlob
import nltk

# Download NLTK punkt tokenizer if not already present
try:
    nltk.data.find('tokenizers/punkt')
except LookupError:
    nltk.download('punkt')

from src.preprocessing import clean_text
from src.features import create_features
from src.model import load_model
from src.entity_extraction import extract_entities

# Define the path to the models directory
BASE_PATH = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "models"))

# Load models and vectorizer
model_issue = load_model(os.path.join(BASE_PATH, "model_issue_type.pkl"))
model_urgency = load_model(os.path.join(BASE_PATH, "model_urgency_level.pkl"))
tfidf = joblib.load(os.path.join(BASE_PATH, "tfidf.pkl"))

def predict_ticket(ticket_text):
    # Preprocess the input ticket text
    clean = clean_text(ticket_text)
    
    # TF-IDF transformation
    X_tfidf = tfidf.transform([clean])
    
    # Additional features
    ticket_length = len(clean.split())
    sentiment = TextBlob(clean).sentiment.polarity

    # Combine sparse TF-IDF with dense features
    X_features = scipy.sparse.hstack([
        X_tfidf,
        np.array([[ticket_length]]),
        np.array([[sentiment]])
    ])

    # Make predictions
    issue_pred = model_issue.predict(X_features)[0]
    urgency_pred = model_urgency.predict(X_features)[0]
    entities = extract_entities(ticket_text)

    return {
        "issue_type": issue_pred,
        "urgency_level": urgency_pred,
        "entities": entities
    }