Improve /transcribe endpoint compatibility and robustness

[Forchapeatl]

Problem

Currently, the /transcribe endpoint uses request.stream.read(),

susi_translator/flask/transcribe_server.py

Line 301 in 96848b5

chunk = request.stream.read(2048000)

it assumes a continuous stream of data. This makes it difficult to interface with standard JSON-based clients

susi_translator/flask/audio_grabber.html

Lines 171 to 186 in 96848b5

	const data = { chunk_id: chunkId, audio_b64: base64data };
	// construct the URL from the host and port
	const serverhost = document.getElementById('serverhost').value;
	const serverport = document.getElementById('serverport').value;
	const transcribeurl = `http://${serverhost}:${serverport}/transcribe`;
	const csrftoken = getCookie('csrftoken');
	fetch(transcribeurl, {
	method: 'POST',
	headers: { 'Content-Type': 'application/json' },
	body: JSON.stringify(data)
	})
	.then(response => {
	if (response.ok) {
	console.log(`Sent chunk ${chunkId} with ${buffer.length} samples`);

and can lead to JSONDecodeError if the stream chunks don't align perfectly with JSON boundaries.

Solution

Update the endpoint to handle standard application/json POST requests using Flask’s get_json(). This ensures immediate response to the client while the background thread processes the queue.