app.py

import gradio as gr
from transformers import pipeline

# Load SST sentiment model (adjust model name if different)
sst_classifier = pipeline(
    "text-classification",
    model="SamanthaStorm/tether-sst",  # replace with your actual SST model name
    top_k=None,
    truncation=True
)

def classify_sentiment(text):
    result = sst_classifier(text)[0]
    label = result["label"]
    score = round(result["score"] * 100, 2)

    if label == "LABEL_0":
        return f"Supportive ({score}%)"
    elif label == "LABEL_1":
        return f"Undermining ({score}%)"
    else:
        return f"Unknown label ({label})"

iface = gr.Interface(
    fn=classify_sentiment,
    inputs=gr.Textbox(lines=4, placeholder="Paste message here..."),
    outputs="text",
    title="Tether SST Sentiment Analyzer",
    description="Classifies language as Supportive or Undermining based on tone and intent. Part of the Tether project."
)

iface.launch()