Turn images into one PDF.

Your browser collects the images

The page keeps your files in an items list, in whatever order you set by dragging.

Create PDF packages and sends them

An HTTPS POST with multipart/form-data, each image a separate part of the body.

const formData = new FormData();
items.forEach(it => formData.append('files', it.file));
xhr.open('POST', `${API_URL}/convert`);
xhr.send(formData);

nginx receives the request

Terminates TLS (decrypts the HTTPS traffic), checks client_max_body_size (150M) and rejects anything larger with a 413, then proxies the request to the FastAPI container on localhost:8000.

FastAPI (main.py) receives it

For each file, in order: checks the extension against ALLOWED_EXTENSIONS, reads the bytes into memory, checks the per-file size and running total, then writes the bytes to a RAM-backed temp file. A numbered prefix keeps page order regardless of the original filenames.

async def convert(files: list[UploadFile]):

ImageMagick does the conversion

The magick CLI reads and decodes each image, turns it upright using its orientation data, renders each one onto its own PDF page in order, and writes the multi-page PDF into the same temp directory.

cmd = ["magick", *paths, "-auto-orient", output_path]
subprocess.run(cmd, ...)

FastAPI reads the result into memory

Then the temp directory context exits, and the whole folder, original images and PDF included, is deleted immediately.

with open(output_path, "rb") as f:
    pdf_bytes = f.read()

The response streams back

It travels back through nginx, re-encrypted over HTTPS, to your browser.

return Response(content=pdf_bytes, media_type="application/pdf", ...)

Your browser saves the file

The page makes a temporary in-browser object URL and clicks a download link with your chosen file name. The URL is released when you change the pages or clear them.

pdfUrl = URL.createObjectURL(xhr.response);
a.download = sanitizeFilename(filenameInput.value) + '.pdf';
a.click();