app.py

from fastapi import FastAPI, Form, HTTPException, Request, File
from fastapi.concurrency import asynccontextmanager
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from typing import Annotated
from wtpsplit import SaT

sat_models = {}


@asynccontextmanager
async def lifespan(app: FastAPI):
    # Load the ML model
    sat_models["sat-3l-sm"] = SaT("sat-3l-sm")
    yield
    # Clean up the ML models and release the resources
    sat_models.clear()


app = FastAPI(lifespan=lifespan)
templates = Jinja2Templates(directory="templates")


@app.get("/", response_class=HTMLResponse)
def root(request: Request):
    return templates.TemplateResponse(request=request, name="index.html")


@app.post("/split", response_class=HTMLResponse)
async def split_text(request: Request, text: Annotated[str, Form()] = ""):
    sentences = sat_models["sat-3l-sm"].split(text)
    return templates.TemplateResponse(
        request=request, name="index.html", context={"sentences": sentences}
    )

@app.post("/api/split")
async def split_file(file: Annotated[bytes, File()]):
    if len(file) > 1.44 * 1024 * 1024:  # 1.44 MB
            raise HTTPException(status_code=413, detail="File too large")
    text = file.decode("utf-8")
    sentences = sat_models["sat-3l-sm"].split(text)
    return {"sentences": sentences}