Skip to content
Documentation

RTC bring your own storage

RTC can copy every finished recording to a bucket you own. CHAMPREP keeps its own copy as well, so the recording stays downloadable from RTC whether or not the delivery succeeds.

Delivery to your own storage is included in the Growth and Business tiers. The free trial, the Starter tier and the allowance included with paid CHAMPREP plans do not include it. Without it, the storage routes answer 403 RTC_FEATURE_NOT_IN_PLAN with the tiers that include it in offers. See plans and billing.

If your plan stops including it, RTC turns your destinations off. After you upgrade again, turn each one back on; RTC runs a new test-write first.

TypeUse it forWhat RTC holds
aws-role (preferred)An AWS S3 bucketNothing secret. RTC assumes an IAM role you create, with an external ID.
s3-compatibleCloudflare R2, MinIO, Wasabi, Backblaze B2, or AWS with access keysYour access key ID and secret key, encrypted

A destination belongs to your organization. It serves one project, or every project when projectId is null. An organization can have up to 5 destinations; one more answers 409 DESTINATION_LIMIT.

Create, change, test and remove destinations in the RTC console, or through the API with your signed-in console session. API keys cannot create or change destinations. A key with recordings:read can list them; a project key sees its project’s destinations and those for every project.

Method and pathCredentialResult
GET /v1/storage/destinationsConsole session, or a key with recordings:readThe destinations, oldest first
GET /v1/storage/destinations/{destinationId}Console session, or a key with recordings:readOne destination
GET /v1/storage/aws-setupConsole sessionYour external ID and the IAM policies to paste (AWS role)
POST /v1/storage/destinationsConsole session201, the destination after its test-write
PATCH /v1/storage/destinations/{destinationId}Console sessionChange settings, rotate keys, or turn it on or off
POST /v1/storage/destinations/{destinationId}/testConsole sessionRuns the test-write again
DELETE /v1/storage/destinations/{destinationId}Console session{ "id": "…", "deleted": true }
GET /v1/recordings/{recordingId}/deliveriesrecordings:readThe recording’s deliveries (follow deliveries)
POST /v1/recordings/{recordingId}/redeliverrecordings:write202, delivers the recording again
DELETE /v1/recordings/{recordingId}recordings:writeSends the recording to Trash (delete a recording)
GET /v1/recordings?trashed=truerecordings:readThe recordings in Trash
POST /v1/recordings/{recordingId}/restorerecordings:writeTakes the recording out of Trash
POST /v1/recordings/{recordingId}/purgerecordings:writeDeletes the recording permanently from Trash

Credentials are never returned. Responses are sent with Cache-Control: no-store.

FieldBoth typesNotes
type✓aws-role or s3-compatible
name✓Your label, up to 80 characters
projectId✓One project, or null for every project
bucket, region✓The bucket name and its region
pathTemplate✓Optional; see path templates
keepCopyDays✓Optional, 0 to 3650, default 7; see retention
roleArn, kmsKeyIdaws-roleYour role, and optionally your KMS key for SSE-KMS
provider, endpoint, forcePathStyle, accessKeyId, secretAccessKeys3-compatibleforcePathStyle defaults to true

A destination also reports its status (active, unverified or disabled), lastCheck (the latest test-write) and lastDelivery.

RTC never holds your AWS keys. You create a role in your account that RTC’s delivery role may assume, limited by an external ID unique to your organization.

  1. Call GET /v1/storage/aws-setup?bucket=<your-bucket> (add pathTemplate and kmsKeyArn if you use them), or open the same setup in the console.
  2. Create an IAM role. Paste the returned trustPolicy: it trusts principalArn only when the request carries your externalId.
  3. Attach the returned permissionsPolicy. It allows s3:PutObject and s3:AbortMultipartUpload under your path, s3:DeleteObject only on the validation folder, and kms:GenerateDataKey and kms:Encrypt on your KMS key when you gave one.
  4. Create the destination with type: "aws-role" and the role’s roleArn:
{
"type": "aws-role",
"name": "Recordings archive",
"projectId": null,
"bucket": "acme-rtc-recordings",
"region": "us-east-1",
"roleArn": "<your-delivery-role-arn>"
}

The external ID is created once for your organization and does not change. When available is false in the setup answer, AWS role destinations are not available yet and creating one answers 503 DELIVERY_NOT_CONFIGURED; use access keys meanwhile.

Use type: "s3-compatible" for Cloudflare R2, MinIO, Wasabi, Backblaze B2, or AWS when you prefer access keys. Name your service in the provider field, with one of these values: r2, minio, wasabi, backblaze, aws or other.

  • endpoint must be an HTTPS origin on port 443, with no path, query or user name, outside CHAMPREP domains, and must resolve to public addresses. RTC checks the address again on every connection. Otherwise RTC answers 400 ENDPOINT_NOT_ALLOWED.
  • Give the keys write access to the bucket. Delete access on the champrep-rtc-validation/ folder lets RTC remove its test object.
  • Keys are encrypted at rest and never returned. A destination shows only accessKeyIdHint, the last four characters of the key ID.
  • To rotate keys, PATCH accessKeyId and secretAccessKey together.
  • When storing keys is not available, RTC answers 503 CREDENTIALS_UNAVAILABLE; use an AWS role meanwhile.

RTC tests a destination when you create it, when you change its region, path template, role, KMS key, endpoint, path style or keys, when you turn it back on, and when you call POST /v1/storage/destinations/{destinationId}/test.

The test writes a 64-byte object to <path prefix>/champrep-rtc-validation/<uuid>.txt and then removes it. A refused removal is only a warning: allow s3:DeleteObject on that folder, or delete the object yourself.

When the test fails, the destination is saved as unverified and receives nothing. lastCheck holds a code and a sentence:

CodeMeaning
ROLE_ASSUME_DENIEDRTC could not assume the role (trust policy or external ID)
WRITE_DENIEDThe write was refused
BUCKET_NOT_FOUNDThe bucket does not exist
WRONG_REGIONThe bucket is in another region
INVALID_ACCESS_KEY, INVALID_SECRET_KEYThe keys were refused
KMS_DENIEDThe KMS key refused the encryption
ENDPOINT_NOT_ALLOWED, ENDPOINT_NOT_FOUND, ENDPOINT_UNREACHABLEThe endpoint is refused, unknown or not answering
DELIVERY_NOT_CONFIGUREDAWS role destinations are not available yet
STORAGE_ERRORAnother storage error

Fix the cause, then test again. A destination becomes active when a test succeeds. A destination turned off by CHAMPREP support cannot be turned back on by you (DESTINATION_ADMIN_DISABLED).

The template sets the object key of each file. The default is:

{project}/{yyyy}/{mm}/{dd}/{room}/{recording_id}/{file}
VariableValue
{project}Project ID
{room}Room ID
{recording_id}Recording ID
{external_ref}The room’s externalRef; none when it is absent or unusual
{track_id}Track ID
{artifact}Which file of the recording
{file}<trackId>.mkv, recording.mp4, audio.m4a, or for a transcript transcript.json and transcript.vtt
{yyyy}, {mm}, {dd}Date parts

The aliases project_id, projectId, room_id, roomId, recordingId, externalRef and trackId are also accepted.

A template must contain {recording_id} and end with {file}. Literal text may use letters, digits and ! _ . * ' ( ) / -. It cannot start with / or contain //, . or .. folders. A key can be at most 900 bytes.

  • When a recording is ready, RTC copies each of its files (every track, and the composite files when there are any) to each active destination of its project. A finished transcript is delivered the same way, as transcript.json and transcript.vtt.
  • Only recordings that finish after a destination becomes active are delivered automatically. Deliver earlier ones with redeliver.
  • Files above 16 MiB are sent in parts. RTC reports the SHA-256 of the bytes it sent.
  • A failed attempt is retried with backoff (1, 2, 4 … minutes, at most an hour apart) for 72 hours. After that the delivery fails and recording.delivery_failed is sent.
  • Delivered bytes count toward your plan’s download allowance.
  • Turning a destination off cancels its queued deliveries. Removing it does the same and drops its credentials.

A worker stopped in the middle of an upload can leave incomplete parts in your bucket. Add an AbortIncompleteMultipartUpload lifecycle rule (1 day) to the bucket.

Subscribe your webhook endpoints to these events. Each file delivery sends its own event.

Eventdata
recording.deliveredrecordingId, roomId, destinationId, artifact, trackId (for a track), bucket, key, size, sha256, checksum, etag, deliveredAt
recording.delivery_failedrecordingId, roomId, destinationId, artifact, trackId (for a track), bucket, key, attempts, error (code, message)

A recording carries a delivery summary: delivery.state (pending, delivered, partial or failed) and the pending, delivered and failed counts.

GET /v1/recordings/{recordingId}/deliveries returns recordingId, that summary and deliveries: each job with its target bucket and key, its result (size, SHA-256 and ETag) and its last error.

POST /v1/recordings/{recordingId}/redeliver with an optional { "destinationId": "…" } starts failed, cancelled and delivered jobs over with a new 72-hour window and creates any missing ones.

HTTPCodeMeaning
409RECORDING_NOT_READYOnly a ready recording can be delivered
409NO_ACTIVE_DESTINATIONNo active destination applies to this recording
404NO_ACTIVE_DESTINATIONNo active destination has this destinationId

Deleting a recording sends it to Trash. It is auto-deleted after 30 days, and you can restore it until then. Every call below needs a key with recordings:write or your console session.

DELETE /v1/recordings/{recordingId} sends the recording to Trash:

{ "id": "…", "status": "trashed", "trashedAt": "2026-11-08T12:00:00.000Z", "reason": "customer", "purgeAt": "2026-12-08T12:00:00.000Z" }
  • The recording leaves the list and GET /v1/recordings/{recordingId}. GET /v1/recordings?trashed=true lists the recordings in Trash.
  • Its downloads stop working, including download links already issued.
  • Its queued deliveries are cancelled.
  • It still counts toward recording storage, and is billed, while it is in Trash.
  • purgeAt is when it is auto-deleted: 30 days after it was sent to Trash. It is fixed when the recording is sent to Trash: a later change of the Trash window applies only to recordings sent to Trash after it. purgeAt is null while the Trash auto-delete is switched off.
  • recording.deleted is sent.
  • A recording that is still running or still being processed answers 409 RECORDING_ACTIVE; stop it or wait.

POST /v1/recordings/{recordingId}/restore takes the recording out of Trash. The answer is the recording, back in your lists and downloads as before it was deleted. recording.restored is sent. You can restore a recording until it is auto-deleted or deleted permanently. A restore that comes too late answers 409 RECORDING_PURGED.

Deliveries to your own storage that were cancelled when the recording went to Trash are sent again. A restored recording is kept until you delete it again: retention does not send it back to Trash.

POST /v1/recordings/{recordingId}/purge deletes a recording in Trash for good. RTC accepts it at once and answers 202; the deletion finishes shortly:

{ "id": "…", "status": "deleting", "purgeRequestedAt": "2026-11-08T12:05:00.000Z" }
  • It can never be restored, from the moment RTC accepts the request.
  • When the deletion is done, its files are erased, it stops counting toward recording storage, and recording.purged is sent.
  • Repeating the request is safe: it answers 202 again while the deletion runs, then 200 once it is done: { "id": "…", "status": "purged", "purgedAt": "2026-11-08T12:05:30.000Z" }.
  • Only a recording in Trash can be deleted permanently. Any other recording answers 409 RECORDING_NOT_IN_TRASH: send it to Trash first.

A recording is deleted permanently at its purgeAt (30 days in Trash), the same way, and recording.purged is sent with reason: "expired". CHAMPREP support can also delete a recording in Trash earlier at your request.

Copies already delivered to your bucket are yours. Sending a recording to Trash, deleting it permanently or its auto-delete never removes them.

CHAMPREP keeps its own copy of a recording for your plan’s retention: 7 days on the trial, 30 days on Starter, 90 days on Growth and 365 days on Business. When that time has passed, the recording is sent to Trash the same way as above, with reason: "retention": you can restore it for 30 days, and it is auto-deleted after that.

Retention applies only to recordings made after RTC started applying it. Recordings made before then are not expired by it. Retention never touches the copies in your own bucket.

keepCopyDays on a destination (0 to 3650, default 7) records how long you want CHAMPREP to keep its copy after a delivery. Today CHAMPREP’s copy follows your plan’s retention; this setting does not remove it earlier.