Submit an AssemblyAI async transcription with speaker labels, sentiment and entity detection, then poll for results

Documented steps

1. Get an AssemblyAI API key from the dashboard.
2. Upload local audio: POST https://api.assemblyai.com/v2/upload with header authorization: <API_KEY> (no Bearer prefix) and the raw file bytes as the body (e.g. curl --data-binary @file.mp3).
3. Read upload_url from the JSON response. Alternatively skip the upload and pass any publicly reachable audio URL directly in the next step.
4. Submit the job: POST https://api.assemblyai.com/v2/transcript with header authorization: <API_KEY> and JSON body {"audio_url": "<upload_url or public URL>"}.
5. Add "speaker_labels": true for diarization — results come back as an utterances array of {speaker, text, start, end}.
6. Add "sentiment_analysis": true — results appear under sentiment_analysis_results as per-sentence {text, start, end, sentiment: POSITIVE|NEUTRAL|NEGATIVE, confidence}.
7. Add "entity_detection": true — results appear in a top-level entities array of {entity_type, text, start, end} ordered by first appearance.
8. Do not use summarization / summary_model / summary_type: these transcript-API parameters are deprecated (as is auto_chapters). For summaries, transcribe first then POST to the LLM Gateway at https://llm-gateway.assemblyai.com/v1/chat/completions with a prompt plus the transcript text.
9. The submit call returns immediately with the transcript object containing id and status: "queued".
10. Poll GET https://api.assemblyai.com/v2/transcript/{id} with the same authorization header at roughly 3-second intervals, the documented recommendation.
11. Stop when status becomes "completed" — then read text, utterances, sentiment_analysis_results and entities — or "error", in which case read the error field.
12. For EU data residency, use https://api.eu.assemblyai.com/v2/... in place of the default host throughout.
13. Verify against the official docs: https://www.assemblyai.com/docs/getting-started/transcribe-an-audio-file and https://www.assemblyai.com/docs/api-reference/transcripts/submit