Recording Downloader

Fetches the meeting recording from the meeting provider and stores it, ready for everything else. This is always the first step: no recording, no pipeline.

At a glance

Needs	A recording ID on the conversation (put there when the meeting was booked)
Produces	The recording file, stored and linked to the conversation
Waits when	The provider hasn't finished uploading the recording yet
Re-running	Safe — skips if the file is already there

What it does

Downloads the recording from the meeting provider (Cloudflare RealtimeKit) and prepares it for streaming, so a video player can start it without downloading the whole file first.

It does not fetch the transcript. That's a separate step (Transcript Importer), specifically so that a missing or slow transcript never holds up — or breaks — getting the recording safely stored.

How it behaves

• The usual wait. Recordings move through invoked → recording → uploading → uploaded at the provider, and only an uploaded recording can be downloaded. Just after a meeting ends it will be in one of the earlier states, so the step waits and tries again. A few minutes is normal.
• It won't download twice. If the recording is already recorded on the conversation and the file is still on disk, the step does nothing and moves on.
• Fallback on preparation. If the streaming-preparation pass fails, the original file is kept and used rather than the step failing.

Troubleshooting

Symptom	Likely cause	What to do
Sitting here for a few minutes after the meeting, no error	The recording isn't uploaded at the provider yet	Normal. Wait for the next run.
Sitting here for hours, no error	The meeting may never have been recorded, or the conversation has no recording ID	Check the meeting was actually recorded; check the conversation carries a recording reference from booking
Error mentioning credentials or authorisation	The meeting-provider credentials on the agent are missing or wrong	An administrator needs to correct them, then re-run the step
Error from the provider's API	The recording ID doesn't resolve, or the provider rejected the request	Confirm the recording exists at the provider for that session
Downloaded, but a later step says the file is missing	The file was archived to cloud storage and the local copy deleted	Expected after the Cloud Storage Migrator runs — steps that need the media use the cloud copy

Technical reference

Step type	realtimekit_downloader
Runs after	(nothing — first step)
Feeds	transcript_importer, google_transcribe, video_editor
Reads	config["enter-coherence"]["recordingId"]
Writes	config["iris.downloads"]["recording"]
Needs on the agent	cloudflare.cloudflare_account_id, cloudflare.app_id, cloudflare.api_token
Needs on the server	ffmpeg