app_model.py

import os
import re
from PIL import Image
from dotenv import load_dotenv
from fastapi import FastAPI, UploadFile, File, HTTPException, Header
from fastapi.middleware.cors import CORSMiddleware
from pdf2image import convert_from_bytes
import gradio as gr
from transformers import pipeline

# Load .env
load_dotenv()
API_KEY = os.getenv("API_KEY")
MODEL_ID = "scb10x/typhoon-ocr-7b"

ocr_pipeline = pipeline("image-to-text", model="scb10x/typhoon-ocr-7b")

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

# --- UTILS ---
def pdf_to_image(file_bytes: bytes) -> Image.Image:
    images = convert_from_bytes(file_bytes)
    return images[0]  # Only first page for now

def run_ocr(image: Image.Image) -> str:
    result = ocr_pipeline(image)
    return result[0]["generated_text"]

def preprocess_text(text: str) -> str:
    text = re.sub(r"</?(figure|table|tr|td|th|b|i|u|p|div|span)[^>]*>", "\n", text)
    text = re.sub(r"<.*?>", "", text)
    text = re.sub(r"\n+", "\n", text)
    text = re.sub(r"\s{2,}", " ", text)
    return text.strip()

def extract_fields_regex(text: str) -> dict:
    patterns = {
        "tax_id": r"(?:TAX\s*ID|เลขที่ผู้เสียภาษี)[\s:\-\.]*([0-9]{10,13})",
        "tax_invoice": r"(?:TAX\s*INV\.?|เลขที่ใบกำกับภาษี|ใบกำกับ)[\s:\-\.]*([0-9A-Z\-\/]{6,20})",
        "tax_date": r"(?:DATE|วันที่|ออกใบกำกับวันที่)?[\s:\-\.]*([0-9]{2,4}/[0-9]{1,2}/[0-9]{1,2})",
        "amount": r"(?:จำนวนเงิน(?:\s*บาทต่อลิตร)?|AMOUNT\s*THB|รวมเงิน)[\s:\-\.]*([0-9,]+\.[0-9]{2})",
        "baht_per_litre": r"(?:บาทต่อลิตร|ราคาต่อลิตร|Baht/Litr|Bath/Ltr)[\s:\-\.]*([0-9,]+\.[0-9]{2})",
        "litre": r"(?:ลิตร|Ltr\.?|Ltrs?\.?)[\s:\-\.]*([0-9,]+\.[0-9]{2,3})",
        "vat": r"(?:VAT|ภาษีมูลค่าเพิ่ม)[\s:\-\.]*([0-9,]+\.[0-9]{2})",
        "total": r"(?:TOTAL\s*THB|ยอดรวม|รวมทั้งสิ้น|รวมเงินทั้งสิ้น|ยอดเงินสุทธิ)[\s:\-\.]*([0-9,]+\.[0-9]{2})",
    }

    results = {}
    for field, pattern in patterns.items():
        match = re.search(pattern, text, re.IGNORECASE)
        results[field] = match.group(1).strip() if match else None
    return results

# --- API Endpoint ---
@app.post("/api/ocr_receipt")
async def ocr_receipt(
    file: UploadFile = File(...),
    x_api_key: str | None = Header(None),
):
    if API_KEY and x_api_key != API_KEY:
        raise HTTPException(status_code=401, detail="Invalid API key")

    content = await file.read()

    try:
        if file.filename.lower().endswith(".pdf"):
            image = pdf_to_image(content)
        else:
            image = Image.open(file.file).convert("RGB")

        text = run_ocr(image)
        text_cleaned = preprocess_text(text)
        extracted = extract_fields_regex(text_cleaned)

        return {
            "raw_ocr": text,
            "preprocessed_text": text_cleaned,
            "extracted_fields": extracted,
        }

    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

# --- Gradio UI ---
def gradio_interface(image_path: str | Image.Image):
    if isinstance(image_path, str) and image_path.lower().endswith(".pdf"):
        with open(image_path, "rb") as f:
            image = pdf_to_image(f.read())
    elif isinstance(image_path, str):
        image = Image.open(image_path).convert("RGB")
    else:
        image = image_path.convert("RGB")

    text = run_ocr(image)
    text_cleaned = preprocess_text(text)
    extracted = extract_fields_regex(text_cleaned)
    return text_cleaned, extracted

with gr.Blocks() as demo:
    gr.Markdown("## 🧾 Thai Receipt OCR (Typhoon 7B)")
    with gr.Row():
        img = gr.Image(type="filepath", label="📤 Upload receipt (Image or PDF)")
        out_text = gr.Textbox(label="📝 OCR Text", lines=12)
        out_fields = gr.JSON(label="🧠 Extracted Fields")
    gr.Button("🔍 Run OCR").click(fn=gradio_interface, inputs=img, outputs=[out_text, out_fields])

demo.launch()