Retrieve a Recall.ai bot's recording, video and transcript artifacts after the meeting ends

Documented steps

1. Wait for bot status 'done' (usually within seconds of call_ended) before fetching media. If the bot went 'fatal', media captured up to the failure is still recovered.
2. GET https://<region>.recall.ai/api/v1/bot/{id} against the SAME region the bot was created in, with 'Authorization: Token <key>'. Rate limit: 300 requests/min per workspace. Ref: https://docs.recall.ai/reference/bot_retrieve
3. Read the recordings array (one entry per recording session) and, inside each, media_shortcuts - a convenience map that avoids extra calls for common artifacts. Ref: https://docs.recall.ai/docs/media-shortcuts
4. media_shortcuts.video_mixed = combined video+audio; audio_mixed = mixed audio (if enabled); transcript = transcript artifact (only if a transcript provider was configured at create time); participant_events = join/leave/speaking activity; meeting_metadata = meeting info. Ref: https://docs.recall.ai/docs/receive-a-recording
5. Each shortcut entry carries id, recording_id, status, created_at, expires_at and data.download_url - a pre-signed, time-limited S3 URL. Download promptly; do not persist the URL.
6. Retention: accounts created after 2025-06-12 retain media indefinitely by default unless recording_config.retention is set; older accounts default to 7-day retention. Ref: https://docs.recall.ai/docs/storage-and-playback
7. If a download_url has expired, re-call GET /bot/{id} for a freshly signed URL rather than trying to refresh the old link.
8. The bot's meeting_url field is cleared 14 days after termination even though other metadata persists - copy it into your own store if you need it later.
9. For unmixed tracks, use the dedicated Retrieve Audio Separate / Retrieve Video Separate endpoints; those are not exposed through media_shortcuts.
10. Prefer the bot.done webhook over polling to know when artifacts are ready.