Deliver your archive to your own bucket

Blob export continuously delivers your team’s sealed archive into storage you own and control - an Amazon S3 bucket or an Azure Blob container. It runs alongside Comma’s own retention: your archive keeps living in Comma, and a second, verifiable copy lands in your bucket that your own auditors, e-discovery tools, and data pipelines can read directly.

This is the “bring your own bucket” path for firms that want the archive to physically reside in their own cloud tenancy - for data-residency, defense-in-depth, or downstream analytics.

What gets delivered

Every sealed record for your team is written out as a line-delimited JSON envelope plus content-addressed object bytes:

• Messages, reactions, edits/revisions, and poll votes - for business conversations only
• Calls and voicemails (telephony compliance records)
• Attachments and media, content-addressed by SHA-256
• Journaled inbound email (the RFC822 original ships as a referenced object)
• Captured site-scan / file-version content

Personal content never leaves

Business classification is enforced in a single database filter anchored on your team boundary. A personal (non-business) message is never delivered to your bucket, and one team’s delivery can never contain another team’s data.

How it works

• Go-forward from activation. Delivery starts at the moment your destination is activated. Content sealed before activation is not back-filled in v1 (a historical backfill is a planned follow-up).
• Deterministic and idempotent. Records are delivered in daily-partitioned batches under object keys derived from the data itself. A retry re-ships the same records under the same keys with byte-identical content, so a re-run is always safe.
• Completeness you can prove. Each batch writes a manifest.json listing every data object with its SHA-256 and byte size, plus per-record-type counts. A zero-length _SUCCESS marker is written last - its presence means the batch is complete.
• Export duplicates, never destroys. Blob export only ever writes a copy into your bucket. It does not delete or alter your Comma archive, and legal holds are unaffected.

Object layout

Everything lands under an object-key prefix you choose (blank means the bucket/container root):

<prefix>dt=<YYYYMMDD>/batch-<batch_key>/records.jsonl.gz   # gzipped JSON Lines, one record per line
<prefix>dt=<YYYYMMDD>/batch-<batch_key>/manifest.json      # per-batch inventory + SHA-256s
<prefix>dt=<YYYYMMDD>/batch-<batch_key>/_SUCCESS           # written LAST = batch complete
<prefix>attachments/<content_hash>                         # content-addressed bytes, deduped across batches

Each line of records.jsonl.gz is a self-describing envelope (schema: comma-blob-export/1) carrying the record type, content hash, archived-at timestamp, custodian identity, the canonical payload (inline, or referenced for large/raw payloads), and its attachments. Lines are recursively key-sorted so a redelivery is byte-identical.

Provisioning (v1)

In v1, destinations are provisioned by your Comma account team. There is no self-serve configuration screen yet - so your job is to prepare the bucket and a scoped credential, then hand the details to Comma:

1. You create the bucket/container and a least-privilege credential (the guides below give you an exact script).
2. You verify that credential works in the shape Comma needs (a one-command probe).
3. You send Comma the connection details over a secure channel.
4. Comma runs a one-time Verify & activate check (it writes, reads back, and deletes a single tiny probe object under your prefix) and turns on delivery.

Once active, a team admin sees a read-only Bucket delivery status line - provider, an Active badge, and the last delivery time - on the Retention, storage & records settings page. No credentials are ever shown there.

Set up your destination

Pick your cloud and follow the exact steps - including a copy-paste setup script and a verification script:

• Amazon S3 setup →
• Azure Blob setup →

Egress IP allowlisting

All delivery leaves Comma from a single static egress IP. If you restrict your bucket to specific source addresses, your account team will give you the current IP to allowlist - confirm it with them before locking a bucket policy, as it can change with an infrastructure move.

Download the setup guide (PDF)

Prefer a printable version to hand to your cloud admin? Each setup guide is also available as a PDF:

• Amazon S3 setup (PDF)
• Azure Blob setup (PDF)