Importing / deploying datapacks

Deployment writes datapacks from disk into a Salesforce org. In Vlocode there is no separate "import" step — deployment is the import path: vlocode deploy reads datapack files, converts them to Salesforce records, resolves their dependencies, computes a safe order, and upserts them. The same engine handles Vlocity/OmniStudio activation (OmniScripts, LWC, matrices) as part of the run.

Command	vlocode deploy — full option reference in the CLI reference.
Engine	DatapackDeployer, DatapackDeployment.
Related	Customizing deployment with specs · Exporting datapacks

Prerequisites

• The Vlocode CLI installed and an authenticated org — see Installation and Authentication.
• A folder of datapacks to deploy (for example produced by vlocode export --expand).

Running a deployment

Point deploy at one or more folders or datapack files. Folders are scanned recursively for *_DataPack.json files:

# Deploy a folder tree using an SFDX alias
vlocode deploy ./datapacks --user my-org

# Deploy only changed datapacks to production
vlocode deploy ./datapacks --instance login.salesforce.com --delta

The most frequently used options are summarized below; see the CLI reference for the complete list with defaults.

Option	When to use
--delta	Deploy only datapacks that differ from the target org. Recommended for fast, repeatable deployments.
--retry-count <n>	Increase resilience to transient row-lock/timeout errors.
--strict-order	Enforce datapack-level ordering when record-level ordering is not enough.
--purge-dependencies	Force-replace embedded child collections (non-production).
--allow-unresolved	Continue when a dependency cannot be resolved (may produce inconsistent data).
-y, --continue-on-error	Continue with datapacks that loaded successfully even if others failed to load.

How deployment works

datapack files
       │
       ▼
1. load            DatapackLoader reads *_DataPack.json files and resolves
       │            external file references (HTML, images, ...). Datapack
       │            type is inferred from folder + SObject type.
       ▼
2. preprocess      spec `preprocess` hook may patch datapack data
       │
       ▼
3. convert         DatapackRecordFactory turns each datapack into one or more
       │            DatapackDeploymentRecords; embedded datapacks recurse;
       │            references become dependencies. spec `afterRecordConversion`.
       ▼
4. resolve deps    for each record, resolve:
       │              • internal deps  → other records in this deployment
       │              • external deps  → existing records in the org (lookup)
       ▼
5. resolve ids     look up existing records by matching key → upsert (update)
       │            or insert. `--delta` skips in-sync records.
       ▼
6. order & batch   repeatedly select records whose dependencies are all
       │            resolved, in chunks; detect circular dependencies.
       │            spec `beforeDeploy` / `beforeDeployRecord` hooks.
       ▼
7. deploy          RecordBatch upserts the chunk via REST (or Bulk) API.
       │            spec `afterDeployRecord` / `afterDeploy` hooks.
       ▼
8. activate        post-deploy work: verify Vlocity GlobalKeys, purge
       │            dependents, activate OmniScripts / LWC / matrices, ...
       ▼
9. retry/report    retry retryable failures up to --retry-count; collect
                    errors and warnings.

Records and record groups

A single datapack expands into one DatapackDeploymentRecord per SObject record. Records originating from the same datapack are tracked together as a DatapackDeploymentRecordGroup so that group-level hooks (activation, purge) run once the whole datapack is deployed.

Dependency resolution

Datapacks carry two kinds of dependency (see Concepts):

• Internal (VlocityMatchingKeyObject) — resolved against the other records in the same deployment by source key.
• External lookup (VlocityLookupMatchingKeyObject) — resolved against the target org by querying on the matching-key fields carried in the reference.

External resolution goes through DatapackLookupService, which fetches each SObject's matching-key definition, queries existing records, and caches the results. Requests are batched by DeferredDependencyResolver to minimize API calls.

Ordering and circular dependencies

The deployment loop repeatedly selects the records whose dependencies are all satisfied and deploys them in chunks. This record-level ordering allows optimal chunking across datapacks. If no record can make progress while pending records remain, the engine detects a circular dependency and fails the involved records. --strict-order additionally enforces whole-datapack ordering when record-level ordering is insufficient.

Matching and upsert

Each record is matched to an existing org record using its matching key fields — the same concept used on export. Matching-key definitions come from the org's %vlocity_namespace%__DRMatchingKey__mdt custom metadata, falling back to inferred keys.

• Match found → update the existing record.
• No match → insert a new record.
• Multiple matches → a warning is logged and the first match is used.
• With --delta, unchanged records are skipped entirely.

Vlocity triggers and GlobalKeys

Some Vlocity objects have triggers that overwrite fields (such as GlobalKey fields) during insert/update. After deploying, the engine re-fetches the affected records, compares the GlobalKey fields against the expected values, and corrects them if a trigger changed them — disabling triggers during the corrective update where needed.

What the command reports

After a run, vlocode deploy prints a summary with the elapsed time and the number of errors and warnings, then lists per-datapack messages (errors always; warnings under --verbose). Namespace prefixes are stripped from source keys for readability.

Important for automation. A deployment that completes but reports per-record errors still exits 0. To gate a pipeline on record-level failures, inspect the summary or parse the NDJSON log file. The process exits 1 only on an unhandled error (see Exit codes).

Troubleshooting

Symptom	Cause and resolution
No datapacks found in specified paths	The paths contain no *_DataPack.json files. Check the folder, or that the export produced expanded datapacks.
Circular dependency failure	Two or more records depend on each other. Review the datapack references; try --strict-order, or split the cycle.
Dependency cannot be resolved	A referenced record is missing from both the deployment and the org. Deploy the dependency first, use --lookup-failed, or (with caution) --allow-unresolved.
Field exists in datapack but not deployed	The field is absent in the target org; Vlocode matches datapack fields against the org schema and reports unmatched fields as errors. Review the org's object/field setup.
Deleted child records are not removed	By design Vlocode does not delete records unless asked. Use --purge-dependencies (non-production) to replace embedded child collections.

Raise the log level with -v/--debug to see detailed per-record diagnostics in the NDJSON log file.

Programmatic deployment

The engine is usable directly as a library; the convenience entry point is the deploy function in datapackDeploy.ts:

import { container } from '@vlocode/core';
import { DatapackDeployer } from '@vlocode/vlocity-deploy';
import { DatapackLoader } from '@vlocode/vlocity';

const datapacks = await container.get(DatapackLoader)
  .loadDatapacksFromFolder('./datapacks');

const deployment = await container.new(DatapackDeployer)
  .createDeployment(datapacks, {
    maxRetries: 1,
    deltaCheck: true,
    strictOrder: false,
  });

await deployment.start();

// Inspect results
const messages = deployment.getMessages();
const byDatapack = deployment.getMessagesByDatapack();

DatapackDeploymentOptions (source) exposes the same switches as the CLI plus internal tuning such as chunkSize, bulkApiThreshold, purgeMatchingDependencies, and disableTriggers.

Next steps

To add behaviour for a specific datapack type — patch data before conversion, add dependencies, or run post-deploy activation — see Customizing deployment with specs.