-
Choose a method for Autopilot and Citus
A native Autopilot or Citus integration is worth evaluating when its current object coverage, direction, and account requirements match your use case. Check both vendors' current listings; this page does not assume a native connector exists.
Evaluate Stacksync with a compatibility demo covering the exact objects, required operations, and data volume. Ask for a working example before choosing the managed sync route.
A custom API integration gives the team control over transformations and orchestration. Include authentication renewal, pagination, request quotas, duplicate prevention, retry behavior, and reconciliation in its maintenance cost. A one-time file export can fit a migration or snapshot when ongoing updates are unnecessary.
-
Define record identity and field ownership
Record the stable source ID and destination ID for each Autopilot–Citus record relationship. Matching object names do not prove matching business meaning: choose an explicit identity rule and check duplicates before the first load.
Stacksync does not automatically merge pre-existing duplicates when two-way sync starts. Use an empty destination for a clean initial load where possible, or agree on a reconciliation plan for existing data.
For each editable field, decide whether Autopilot, Citus, or an approved two-way rule owns changes. Managed IDs, formulas, and other read-only fields cannot become writable by mapping them. Test simultaneous edits with the integration team instead of assuming a conflict-resolution policy.
Sources: Stacksync two-way sync: fields, views, and initial data
-
Build and verify the mapping
Use a mapping worksheet with source object, source field, destination object, destination field, direction, identity key, transformation, required-field rule, and a sample value. Check allowed values, timestamp zones, null behavior, precision, and maximum lengths.
For example, if one field represents an amount in cents and the other uses currency units, document the conversion and verify the currency and rounding rule. This is a mapping design example, not a claim about either connector schema.
Map related records in dependency order and verify that foreign IDs or association tables refer to the intended parent record. Map business entities deliberately; unrelated objects must not be paired by their position in a list.
Sources: Stacksync two-way sync: fields, views, and initial data
-
Validate the initial load and ongoing changes
Start with a small set of non-production records in Autopilot and Citus. Reconcile counts within the same filters, verify stable IDs, and compare critical fields after the first load.
Run create and update checks only in confirmed directions. Then test a rejected value, a missing required field, a repeated update, a relationship change, and concurrent edits. Confirm deletion and archival behavior separately before testing either operation.
Record source-change time and destination-observed time for normal traffic and a burst. Use those measurements to set the accepted delay and backlog alert; CDC or webhooks alone do not establish an end-to-end speed guarantee.
Sources: Stacksync two-way sync: fields, views, and initial data
-
Troubleshoot and operate the sync
If an object is missing, check account eligibility, permissions, object support, primary keys, and change-tracking setup before changing mappings. If an update is rejected, inspect the specific destination validation error and correct the data or rule.
The Stacksync issues dashboard documents retry and revert actions. Retry reads the current source value again; it is not an immutable replay of an earlier payload. Review the intended record state before retrying or reverting.
Assign an owner for credential renewal, rejected records, backlog, and schema changes. Re-run the pilot checks after changing permissions, fields, account plans, or mappings, and schedule reconciliation at a frequency suited to the business process.
Sources: Stacksync issues dashboard: retry and revert
