from transformers import pipeline

sentiment_model = pipeline(
    "sentiment-analysis", 
    model="cardiffnlp/twitter-roberta-base-sentiment-latest", 
    device=-1
)

def analyze_sentiment(reddit_data):
    sentiments = []
    if not reddit_data:
        return sentiments  # handle empty case

    comments_iter = []
    if isinstance(reddit_data, dict):
        for comments in reddit_data.values():
            comments_iter.extend(comments)
    elif isinstance(reddit_data, list):
        comments_iter = reddit_data

    for comment in comments_iter:
        body = comment.get("body", "")
        if body:
            result = sentiment_model(body[:512])[0]
            label = result["label"].lower()
            sentiments.append({"body": body, "sentiment": label})
    return sentiments