app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

# -------------------------------------------------
# CONFIG
# -------------------------------------------------
MODEL_NAME = "j-hartmann/emotion-english-distilroberta-base"

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# -------------------------------------------------
# LOAD MODEL (ONCE AT STARTUP)
# -------------------------------------------------
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)

model.to(device)
model.eval()

# -------------------------------------------------
# FASTAPI APP
# -------------------------------------------------
app = FastAPI(title="Emotion Detection API")

# ✅ ROOT ROUTE (CRITICAL FOR HF SPACES)
@app.get("/")
def health():
    """
    Health / wake-up endpoint.
    Hugging Face uses this to wake the Space.
    """
    return {"status": "ok"}

# -------------------------------------------------
# REQUEST SCHEMA
# -------------------------------------------------
class EmotionRequest(BaseModel):
    text: str

# -------------------------------------------------
# EMOTION ENDPOINT
# -------------------------------------------------
@app.post("/emotion")
def classify_emotion(payload: EmotionRequest):
    text = payload.text.strip()
    if not text:
        raise HTTPException(status_code=400, detail="Text cannot be empty")

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

    inputs = {k: v.to(device) for k, v in inputs.items()}

    with torch.no_grad():
        outputs = model(**inputs)

    probs = torch.softmax(outputs.logits, dim=-1)
    pred_id = torch.argmax(probs, dim=-1).item()

    return {
        "emotion": model.config.id2label[pred_id],
        "confidence": round(probs[0][pred_id].item(), 4),
    }