app.py

import os
import uvicorn
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from typing import List, Dict, Union
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch


# Definition of Pydantic data models
class ProblematicItem(BaseModel):
    text: str

class ProblematicList(BaseModel):
    problematics: List[str]

class PredictionResponse(BaseModel):
    predicted_class: str
    score: float

class PredictionsResponse(BaseModel):
    results: List[Dict[str, Union[str, float]]]

# FastAPI Configuration
app = FastAPI(
    title="Problematic Specificity Classification API",
    description="This API classifies problematics using a fine-tuned model hosted on Hugging Face.",
    version="1.0.0"
)

# Model environment variables
MODEL_NAME = os.getenv("MODEL_NAME", "votre-compte/votre-modele")
LABEL_0 = os.getenv("LABEL_0", "Classe A")
LABEL_1 = os.getenv("LABEL_1", "Classe B")

# Loading the model and tokenizer
tokenizer = None
model = None

def load_model():
    global tokenizer, model
    try:
        tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
        model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)
        return True
    except Exception as e:
        print(f"Error loading model: {e}")
        return False

# API state check route
@app.get("/")
def read_root():
    return {"status": "ok", "model": MODEL_NAME}

# Route for checking model status
@app.get("/health")
def health_check():
    global model, tokenizer
    if model is None or tokenizer is None:
        success = load_model()
        if not success:
            raise HTTPException(status_code=503, detail="Model not available")
    return {"status": "ok", "model": MODEL_NAME}

# Route to predict a single problem at a time
@app.post("/predict", response_model=PredictionResponse)
def predict_single(item: ProblematicItem):
    global model, tokenizer
    
    if model is None or tokenizer is None:
        success = load_model()
        if not success:
            raise HTTPException(status_code=503, detail="Model not available")
    
    try:
        # Tokenization
        inputs = tokenizer(item.text, padding=True, truncation=True, return_tensors="pt")
        
        # Prediction
        with torch.no_grad():
            outputs = model(**inputs)
            probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
            predicted_class = torch.argmax(probabilities, dim=1).item()
            confidence_score = probabilities[0][predicted_class].item()
        
        # Associate the correct label
        predicted_label = LABEL_0 if predicted_class == 0 else LABEL_1
        
        return PredictionResponse(predicted_class=predicted_label, score=confidence_score)
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error during prediction: {str(e)}")

# Route for predicting several problems at once
@app.post("/predict-batch", response_model=PredictionsResponse)
def predict_batch(items: ProblematicList):
    global model, tokenizer
    
    if model is None or tokenizer is None:
        success = load_model()
        if not success:
            raise HTTPException(status_code=503, detail="Model not available")
    
    try:
        results = []
        
        # Batch processing
        batch_size = 16
        for i in range(0, len(items.problematics), batch_size):
            batch_texts = items.problematics[i:i+batch_size]
            
            # Tokenization
            inputs = tokenizer(batch_texts, padding=True, truncation=True, return_tensors="pt")
            
            # Prediction
            with torch.no_grad():
                outputs = model(**inputs)
                probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
                predicted_classes = torch.argmax(probabilities, dim=1).tolist()
                confidence_scores = [probabilities[j][predicted_classes[j]].item() for j in range(len(predicted_classes))]
            
            # Converting numerical predictions into labels
            for j, (pred_class, score) in enumerate(zip(predicted_classes, confidence_scores)):
                predicted_label = LABEL_0 if pred_class == 0 else LABEL_1
                results.append({
                    "text": batch_texts[j],
                    "class": predicted_label,
                    "score": score
                })
        
        return PredictionsResponse(results=results)
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Error during prediction: {str(e)}")

# Model loading at startup
@app.on_event("startup")
async def startup_event():
    load_model()

# Entry point for uvicorn
if __name__ == "__main__":
    # Starting the server with uvicorn
    uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=True)