Publish to RabbitMQ

Use a RabbitMQ destination to publish rows from a Lake table, a Warehouse table or a query to an exchange on your broker. Each table publishes under its own routing key, and you choose whether each run sends every row or only recent ones.

Before you start

• A RabbitMQ broker that DataMind OS can reach on its AMQP port.
• An exchange that already exists on the broker. DataMind OS publishes into it and never declares it.
• A broker user that may publish to that exchange. Publishing needs no other broker rights.
• The Management API port must be reachable while Writeback lists the routing keys bound to your exchange. If it is not reachable, define your tables in Message Schemas instead; the picker then uses them.
• The Add/Modify WriteBack Objects permission to add a destination. See Writeback permissions.

Add a RabbitMQ destination

1. Open WriteBack in the sidebar, click Add destination, then choose RabbitMQ.
2. In General, enter Name, which needs at least 3 characters, and Code, which needs at least 2 characters and may contain only letters and numbers.
3. In Broker, enter Host, Port, Virtual host, Management API port, Username and Password. Turn on TLS to connect over AMQPS.
4. In Test connection, click Test connection. The test is optional. The card reads Connection test passed, or Connection test failed with the reason. The footer of this form has no test button of its own.
5. In Exchange & Queues, enter Exchange and choose an Exchange Type. Change Persistent messages and Max rows per message if the defaults do not suit you.
6. Optional: in Message Schemas (optional), add a JSON list of table entries. See Choose routing keys and message format below.
7. Click Add.
8. Import the tables you want to fill, as described in Import destination tables.

For editing, duplicating and deleting a destination, see Add and manage destinations.

Settings for the broker and exchange

To read from RabbitMQ into the Lake instead, see Webhook and RabbitMQ.

Choose routing keys and message format

Leave Message Schemas empty and Writeback discovers the tables from the exchange's existing bindings: each routing key bound to a queue becomes a table. A routing key with no bound queue does not appear; add an entry for it to make it appear.

Message Schemas is an optional JSON list of table entries. Each entry can set:

• name: the table name. Required.
• routingKey: the routing key to publish to. Defaults to the name.
• messageMode: row (the default) or envelope.
• columns: the columns of the table, if you list them. "pk": true marks the key column; leave the list out and the columns are detected from sample messages.

When the field has entries, only those tables are offered. Clear the field to go back to discovery.

Row mode (the default) sends each message as one JSON object per row, with the mapped target column names as property names. When Max rows per message is above 1, several rows share one message, one JSON object per line. Your consumer must then split each message on new lines.

Envelope mode sends one message per row, wrapped like this:

{"entity": "Orders", "emittedAt": "2026-10-15T09:30:00Z", "data": {"OrderID": 1042, "Status": "shipped"}}

emittedAt is the time the message is published. If the table is built from a query that returns one text column, envelope mode publishes that text as the message body, without the wrapper.

The fields that are published come from the column mapping on Columns. See Map columns.

Publish only recent rows

By default, Sync mode is Full snapshot, so every run publishes all rows of the table. To publish only recent rows:

1. Open the Writeback table (see Configure a Writeback table) and stay on General.
2. In Publish Window, set Sync mode to Incremental.
3. Choose an Increment column from the table's columns. Use a datetime column that changes when a row changes, such as an update timestamp.
4. Set Minus days, a whole number from 0 to 9999, with no decimals or negative values. Press Enter or click outside the field to save it.

Each change saves at once, and DataMind OS confirms it with Publish window updated. If the save fails, Failed to update publish window appears; try again.

On every run, the table publishes the rows whose Increment column falls between the start of the run day minus Minus days, and the date and time of the run. For a run you start with Synchronize, that is the date and time you choose, recorded as its Snap Date. See Synchronize a table and read the Sync Log.

The Publish Window appears only on tables of RabbitMQ destinations. Switch back to Full snapshot and every run publishes all rows again.

Each run publishes its rows again, so a repeated or a retried run can publish the same rows twice. Keep the window on Incremental, or limit the rows your source returns.

Permissions

• Add, edit or duplicate a destination: Add/Modify WriteBack Objects.
• Delete a destination: Remove WriteBack Objects.
• Import its routing-key tables: Add/Modify Lake Objects.
• Run a table with Synchronize, or on its workflow: Execute/Abort WriteBack Objects.

See Writeback permissions for the full list.

Limits and edge cases

• Exchange is required. Saving without it stops with the message Exchange is required - messages are published into it.
• Invalid JSON in Message Schemas stops saving, with the message Message schemas: invalid JSON.
• Max rows per message applies to row mode only. Envelope mode always sends one row per message.
• The Publish Window is not offered for database destinations.
• DataMind OS never declares or changes the exchange, its queues or their bindings.
• To copy a RabbitMQ destination, right-click it and choose Duplicate. See Add and manage destinations.
• For other messages, see Writeback troubleshooting.

Questions, answered

Which routing key does a table publish to?

It uses the routingKey of the Message Schemas entry whose name matches the table. Without a matching entry, the table's name is the routing key.

Do I have to fill in Message Schemas?

No. Leave it empty and Writeback offers every routing key bound to a queue on the exchange as a table. Add entries when you need to control the routing key or the message mode, list columns, or include a routing key that has no bound queue yet. While the field has entries, only those tables are offered.

My consumer gets several rows in one message. Why?

Row mode batches rows into one message, up to Max rows per message, which defaults to 1000. Set it to 1 for one message per row, or split each message on new lines in your consumer.

Does DataMind OS create the exchange or the queues?

No. The exchange must already exist on the broker, and the broker user must be allowed to publish to it. DataMind OS does not declare the exchange, its queues or their bindings, so no other broker rights are needed.

Are my messages persistent?

Yes, by default. Persistent messages is on, so messages survive a broker restart. Turn it off to send Transient messages.

How do I stop publishing old rows on every run?

Set Sync mode to Incremental in the table's Publish Window, then choose an Increment column and a Minus days value. Full snapshot publishes every row on every run.

What does Minus days 0 mean?

Each run publishes the rows from the start of the run day up to the time of the run.

Are integers sent as numbers or as strings?

Integers are sent as JSON numbers, including 64-bit values. Text columns are sent as strings.