Skip to content

Features & Capabilities

The Liferay AI Commerce Accelerator provides a comprehensive suite of tools for data generation and management.

AI Data Generation

Rapidly create high-quality commerce entities using state-of-the-art generative AI.

Supported Entities

  • Products: Generates localized names, descriptions, and specifications.
  • Accounts: Creates realistic business accounts with multiple addresses (Billing, Shipping, Head Office). Choose Business, Individual, or Mixed — and with Mixed, set the business/individual split explicitly rather than leaving the proportion to the model. The channel dropdown shows each channel's commerce site type, and a run that would put business accounts into a B2C channel — or individual accounts into a B2B one — is refused rather than generating data Liferay will not let anyone order with. A channel whose site type has not been set is flagged but still allowed, since that is the normal state for one created through the API.
  • Orders: Generates historical order data linked to products and accounts. The date range and the open/processing/shipped/completed split are the operator's to set, and are applied to the generated orders rather than left to the model. When generating orders against existing data you can choose which accounts receive them — any customer, business only, or individual only. Guest and supplier accounts are never used: guest exists for anonymous checkout, and a supplier sells to the business rather than buying from it.
  • Warehouses: Creates inventory locations and manages stock distribution. Unchecking Create Warehouses means exactly that — the run creates none and uses the warehouses already in the instance instead, linking them to the run's channels and stocking their inventory as before.

Generation Options

Every option in the form gates the work it names, and a run does only what the form asks for. Create Warehouses and Generate Specifications each remove their steps from the workflow when unchecked, which is how an operator shortens a run and keeps unwanted entities out of the instance. Every option the form offers also reaches the run: the order date range, the order status split, the PDF content type and the seed pack are the operator's choices, not defaults chosen further down.

Chunked Generation

Large requests are split across multiple AI calls rather than asking for everything at once, which is what stops responses being truncated mid-schema when generating hundreds of entities.

  • Products, Accounts, Orders, Warehouses split the requested count: ask for 200 products at a chunk size of 10 and the service makes 20 calls and concatenates the results.
  • Pricing splits the input list instead, because it is driven by how many products it has been given rather than by a requested total. Its priceListName belongs to the list as a whole, so the first one is kept when the chunks are merged.
  • Chunk sizes are configurable per entity in the AI Chunk Sizes panel of the Configuration client extension, clamped to 1–50 with a default of 10.

Promotions are deliberately not chunked: they generate user segments alongside promotions from the whole catalogue, and splitting the input would fragment those segments across calls.

Visual Assets & Media

  • AI Images: Generates product visuals using DALL-E or Nano Banana.
  • AI PDFs: Creates detailed documentation like User Guides, Technical Specs, or Compliance sheets.
  • Placeholder Mode: Use lightweight mock assets for rapid prototyping without AI costs.

Prompt and Schema Portability

The Configuration client extension can move prompts and schemas between environments without touching the database.

  • Export All / Import All writes and reads a versioned JSON bundle, so a tuned set can be carried from a local instance to a demo one.
  • Copy All puts the same payload on the clipboard, for pasting into an issue or a review.
  • Imported schemas are compiled before being accepted. One that does not compile is rejected individually with the error shown against that entity, and the valid schemas in the same file still import — a bad schema would otherwise break generation for that entity.

Real-time Monitoring

The Dashboard provides granular feedback on every generation session.

  • Overall Progress Gauge: A striking visual summary of the entire workflow.
  • System Health Strip: Real-time status of Liferay connectivity, WebSockets, and AI providers. The /health endpoint also reports the last search reindex outcome, so content that was generated but never indexed is visible rather than silent.
  • Activity Log: Detailed audit trail of all microservice operations.
  • Unified Consoles: The Live Console and the Seeder Console are the same component, so both offer level filtering, real-time search, auto-scroll, clear, and copy-to-clipboard of whatever the filters currently show — useful for pasting a failure into an issue.
  • Batch Error Tracking: Specialized view for diagnosing failures at the External Reference Code (ERC) level.

Dynamic Configuration

Everything is manageable via the AI Commerce Accelerator Configuration UI.

  • Provider Agnosticism: Switch between OpenAI, Google Gemini, and Anthropic Claude for text generation. Claude has no image capability, so a dedicated media provider is required alongside it when generating images.
  • Provider-Aware Models: The default model list shows only the models the selected provider can actually run, so an OpenAI model can never be left selected against Anthropic Claude. A model added by hand that the build does not recognise stays selectable, and a mismatched pairing is rejected before a run starts rather than partway through it.
  • Dedicated Media Keys: Configure separate credentials for image generation (e.g., Nano Banana).
  • Custom Prompts & Schemas: Tailor the AI's output by editing the templates directly in the browser with a built-in CodeMirror editor. Prompts are Jinja2-compatible templates, so a condition or a loop can be written in the prompt itself rather than requiring a code change - see Prompt Templates.
  • Exclude Lists: Protect system data by specifying items that should never be modified or deleted.

System Administration & Troubleshooting

A dedicated Admin Dashboard provides high-level oversight and diagnostic tools for platform operators.

Session Management

  • Session Explorer: View a complete history of all generation and deletion runs.
  • Advanced Filtering: Filter by session name, date range, or outcome (Success, Failed, Cancelled).
  • History Sorting: Sort sessions to quickly find the latest runs or analyze past failures.
  • Targeted Export: Retrieve the exact AI-generated dataset from any previous successful run.
  • Failure Diagnostics: Immediate visibility into why a session failed, including the specific workflow step.

KPI Metrics

  • Success vs. Failure Rate: Real-time tracking of generation reliability.
  • Session Volume: Monitor total usage and throughput of the accelerator.

Configuration Doctor

  • Automated Diagnostics: Real-time alerts for missing AI API keys, invalid prompts, or broken Liferay connectivity.
  • Direct Resolution Links: Deep-links to the exact configuration screen in DXP required to fix an issue.
  • System Parity: Ensure environment consistency by identifying drifted settings across Liferay instances.

Data Portability

  • Export: Save your generated data set to a JSON file for backup or sharing.
  • Import: Restore a previously generated data set into a new environment.

Last Reviewed: 2026-09-18