I have finally managed to work out how to get lindy to download a voice message from Telegram (via two agent steps and much more, which seems overkill but also the only way to extract the string I need from the json blob that is outputted (parseing doesn't work). However, now I can't find a transcription action that will work. The OpenAI ones seem to be broken and the deepgram one is also just giving me errors. When I tell it to just use the no URL and just the file path, as the file has been downloaded already, it gives me one error, and when I tell it to use a URL it gives me different ones; whether that is Telegrams URL (that works when I enter it in my browser) or the one for the downloaded file.
Any thoughts?
