app.py

import gradio as gr
import torch
import joblib
import numpy as np
from transformers import DistilBertTokenizerFast, DistilBertForSequenceClassification
import re
import contractions

# Load model and tokenizer
tokenizer = DistilBertTokenizerFast.from_pretrained("distilbert")
model     = DistilBertForSequenceClassification.from_pretrained("distilbert")
model.eval()

le = joblib.load("label_encoder.pkl")
label_map = {i: cls for i, cls in enumerate(le.classes_)}

# Class display config
class_config = {
    "Normal":               {"emoji": "🟢", "color": "#2ecc71"},
    "Anxiety":              {"emoji": "🟡", "color": "#f39c12"},
    "Stress":               {"emoji": "🟠", "color": "#e67e22"},
    "Depression":           {"emoji": "🔵", "color": "#3498db"},
    "Bipolar":              {"emoji": "🟣", "color": "#9b59b6"},
    "Personality disorder": {"emoji": "⚪", "color": "#95a5a6"},
    "Suicidal":             {"emoji": "🔴", "color": "#e74c3c"},
}

CRISIS_MESSAGE = """
⚠️ CRISIS SUPPORT — You are not alone.

If you or someone you know is in crisis, please reach out immediately:
- International Association for Suicide Prevention: https://www.iasp.info/resources/Crisis_Centres/
- Crisis Text Line (US): Text HOME to 741741
- Nigeria: NEEM Foundation — +234 806 210 6493
- Befrienders Worldwide: https://www.befrienders.org
"""

def clean_text(text):
    if not isinstance(text, str):
        return ""
    text = contractions.fix(text)
    text = text.lower()
    text = re.sub(r"http\S+|www\S+", "", text)
    text = re.sub(r"[^a-z\s]", "", text)
    text = re.sub(r"\s+", " ", text).strip()
    return text

def predict(statement):
    if not statement.strip():
        return "Please enter a statement.", "", ""

    clean = clean_text(statement)

    inputs = tokenizer(
        clean,
        return_tensors="pt",
        truncation=True,
        padding=True,
        max_length=128
    )

    with torch.no_grad():
        outputs = model(**inputs)
        probs   = torch.softmax(outputs.logits, dim=1).squeeze()

    pred_idx    = torch.argmax(probs).item()
    pred_label  = label_map[pred_idx]
    confidence  = round(probs[pred_idx].item() * 100, 1)

    cfg    = class_config[pred_label]
    result = f"{cfg['emoji']} {pred_label}  ({confidence}% confidence)"

    # Top 3 probabilities
    top3_idx  = torch.argsort(probs, descending=True)[:3]
    breakdown = "Probability breakdown:\n"
    for idx in top3_idx:
        lbl = label_map[idx.item()]
        pct = round(probs[idx].item() * 100, 1)
        breakdown += f"  {class_config[lbl]['emoji']} {lbl}: {pct}%\n"

    crisis = CRISIS_MESSAGE if pred_label == "Suicidal" else ""

    return result, breakdown, crisis

# UI
with gr.Blocks(title="MentalPulse") as demo:
    gr.Markdown("""
    # 🧠 MentalPulse — Mental Health Signal Classifier
    ### Detect mental health signals from text using fine-tuned DistilBERT
    *Built by Samuel Yaula Dutse | Mental Health Awareness Month — June 2026*

    > ⚠️ This tool is for research and educational purposes only.
    > It is not a clinical diagnostic tool. If you are in crisis, please seek professional help.
    """)

    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(
                label="Enter a statement",
                placeholder="How are you feeling? Describe your thoughts or emotions...",
                lines=5
            )
            submit_btn = gr.Button("Analyze", variant="primary")

        with gr.Column():
            pred_output      = gr.Textbox(label="Detected Signal", interactive=False)
            breakdown_output = gr.Textbox(label="Confidence Breakdown", lines=5, interactive=False)
            crisis_output    = gr.Textbox(label="Crisis Resources", lines=8, interactive=False, visible=True)

    gr.Examples(
        examples=[
            ["I have been feeling so hopeless lately, nothing makes sense and I just want it to stop"],
            ["My heart races constantly and I cannot stop worrying about everything"],
            ["Today was actually a good day. I felt present and enjoyed time with my family"],
            ["My mood swings are so extreme, one day I feel invincible and the next I cannot get out of bed"],
            ["Work has been overwhelming and I feel like I am stretched too thin"],
        ],
        inputs=[text_input]
    )

    submit_btn.click(
        fn=predict,
        inputs=[text_input],
        outputs=[pred_output, breakdown_output, crisis_output]
    )

demo.launch()