import re
import json
from bertopic import BERTopic
from sentence_transformers import SentenceTransformer
from transformers import pipeline

# IMPORTANT: must match training embedding model
EMBEDDING_NAME = "paraphrase-multilingual-MiniLM-L12-v2"

# Load embedding + topic model
embedding_model = SentenceTransformer(EMBEDDING_NAME)
topic_model = BERTopic.load("afcon_topic_model", embedding_model=embedding_model)

# Load topic mapping
with open("topic_mapping.json", "r") as f:
    topic_mapping = {int(k): v for k, v in json.load(f).items()}

# Sentiment pipeline
sentiment_pipe = pipeline("sentiment-analysis", model="ImaneAz/youtube-roberta-sentiment")
LABEL_MAP = {"LABEL_0": "negative", "LABEL_1": "neutral", "LABEL_2": "positive"}

# Same cleaning used in notebook
def clean_for_topics(text: str) -> str:
    if not isinstance(text, str):
        return ""
    text = text.lower()
    text = re.sub(r"http\S+", "", text)
    text = re.sub(r"[^a-z0-9\s]", " ", text)
    text = re.sub(r"\s+", " ", text).strip()
    return text

def analyze_text(text: str):
    norm = clean_for_topics(text)

    # topic prediction
    topic_id, _ = topic_model.transform([norm])
    topic_id = int(topic_id[0])
    topic_final = topic_mapping.get(topic_id, "Other")

    # sentiment prediction
    try:
        s = sentiment_pipe(text[:512])[0]["label"]
        sentiment = LABEL_MAP.get(s, s)
    except:
        sentiment = "UNKNOWN"

    return {
        "topic_id": topic_id,
        "topic_final": topic_final,
        "sentiment": sentiment,
        "clean_text": norm
    }