app.py

import io
import pickle
import numpy as np
import torch
from fastapi import FastAPI, UploadFile, File, Form
from fastapi.middleware.cors import CORSMiddleware
from PIL import Image
from transformers import AutoTokenizer, AutoModel
import open_clip
import re

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

# step 1: load the models
TEXT_MODEL_NAME = "indobenchmark/indobert-large-p1"
tokenizer = AutoTokenizer.from_pretrained(TEXT_MODEL_NAME)
text_model = AutoModel.from_pretrained(TEXT_MODEL_NAME).to(device)
text_model.eval()

clip_model, _, clip_preprocess = open_clip.create_model_and_transforms("EVA01-g-14-plus", pretrained="merged2b_s11b_b114k")
clip_model.to(device)
clip_model.eval()

with open("xgb_full.pkl", "rb") as f:
    xgb_model = pickle.load(f)

with open("k-means.pkl", "rb") as f:
    kmeans = pickle.load(f)

# step 2: preprocessing
def preprocess_text(text: str) -> str:
    text = str(text).lower()                                      
    text = re.sub(r'http\S+|www\.\S+', '', text)                  
    text = re.sub(r'@\w+|#\w+', '', text)                         
    text = re.sub(r'[^a-z\s]', ' ', text)                        
    text = re.sub(r'\s+', ' ', text).strip()
    return " ".join(text.split())

# step 3: feature encoding (text and image)
def encode_text(text: str):
    # step 3.1 preprocess text
    processed = preprocess_text(text)
    # step 3.2 tokenize text
    tokens = tokenizer(
        processed,
        return_tensors="pt",
        padding="max_length",
        truncation=True,
        max_length=128,
    )
    tokens = {k: v.to(device) for k, v in tokens.items()}

    with torch.no_grad():
        # take the [CLS] token
        out = text_model(**tokens).last_hidden_state[:, 0, :]
    return out.cpu().numpy()

def encode_image(image_bytes):
    # step 4.1 load the image
    img = Image.open(io.BytesIO(image_bytes)).convert("RGB")

    # step 4.2 encode the image into a tensor (embedding image)
    tensor = clip_preprocess(img).unsqueeze(0).to(device)

    with torch.no_grad():
        emb = clip_model.encode_image(tensor)
    return emb.cpu().numpy()

app = FastAPI(
    title="Multimodal Water Pollution Risk API",
    description=(
        "Input: text + image + geospatial + time\n"
        "Model: IndoBERT + EVA-CLIP + XGBoost\n"
    ),
    version="1.0.3",
)

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_methods=["*"],
    allow_headers=["*"],
)

@app.get("/")
def root():
    return {
        "status": "OK",
        "message": "Multimodal Water Pollution Risk API is running.",
        "info": "Use POST /predict with text, image, and features.",
    }

@app.post("/predict")
async def predict(
    text: str = Form(...),
    longitude: float = Form(...),
    latitude: float = Form(...),
    hour: int = Form(...),
    dayofweek: int = Form(...),
    month: int = Form(...),
    image: UploadFile = File(...),
):

    # 1. Encode text
    text_emb = encode_text(text)

    # 2. Encode image
    img_bytes = await image.read()
    img_emb = encode_image(img_bytes)

    # 3. Generate the location cluster
    location_cluster = int(kmeans.predict([[latitude, longitude]])[0])

    # 4. Create feature vector
    add_feats = np.array([[longitude, latitude, location_cluster, hour, dayofweek, month]], dtype=np.float32)

    # 5. Early Fusion 
    fused = np.concatenate([img_emb, text_emb, add_feats], axis=1)

    # 6. Predict
    proba = xgb_model.predict_proba(fused)[0]
    pred_idx = int(np.argmax(proba))

    label = "KRITIS" if pred_idx == 1 else "WASPADA"

    return {
        "prediction": label,
        "cluster_used": location_cluster,
        "probabilities": {
            "WASPADA": float(proba[0]),
            "KRITIS": float(proba[1])
        }
    }

@app.post("/predict_proba")
async def predict_proba(
    text: str = Form(...),
    longitude: float = Form(...),
    latitude: float = Form(...),
    hour: int = Form(...),
    dayofweek: int = Form(...),
    month: int = Form(...),
    image: UploadFile = File(...),
):
    text_emb = encode_text(text)
    img_bytes = await image.read()
    img_emb = encode_image(img_bytes)

    location_cluster = int(kmeans.predict([[latitude, longitude]])[0])

    add_feats = np.array([[longitude, latitude, location_cluster, hour, dayofweek, month]], dtype=np.float32)

    fused = np.concatenate([img_emb, text_emb, add_feats], axis=1)

    proba = xgb_model.predict_proba(fused)[0]

    return {
        "WASPADA": float(proba[0]),
        "KRITIS": float(proba[1]),
        "cluster_used": location_cluster,
    }

if __name__ == "__main__":
    import uvicorn

    uvicorn.run(app, host="0.0.0.0", port=7860)