Skip to main content

Dataset Discovery

This page is mainly for skill authors, delivery engineers, and Agent maintainers. Business users usually do not need to discover dataset codes manually. The Agent often performs this “discovery step” first.

What this step is for​

When an Agent receives a business request, it often needs to answer three questions first:

  1. Which business objects exist in this system?
  2. Which dataset matches the object we care about?
  3. Which fields are available for filtering and output?

dataset list and dataset detail exist to answer those questions.

When it is useful​

  • when you are exploring a business system for the first time
  • when you do not know where “customer,” “order,” “ticket,” or “inventory” data lives
  • when you need to confirm field names before building --params
  • when an Agent should understand the structure before moving on to reads or writes

If you already know the dataset code, you can skip this step.

Prerequisites​

Before using dataset discovery commands, make sure:

  • you have already logged in with an AccessKey
  • the current app is already resolved, for example through app use or --appcode

This is now an AccessKey-based runtime path, not the old Cookie model.

List datasets​

lovrabet dataset list
lovrabet dataset list --name order
lovrabet dataset list --code a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6

This step is basically “show me the business objects in this app.”

Common keyword-first usage:

lovrabet dataset list --name 客户
lovrabet dataset list --name 订单
lovrabet dataset list --name 工单

Inspect dataset details​

lovrabet dataset detail --code a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6

The most important things to read are:

  • name
  • code
  • table
  • fields
  • operations

For Agents and skill maintainers, fields is usually the critical section because later --params must use the correct field names.

How it usually fits into the next step​

The normal flow is:

  1. use dataset list to find the likely business object
  2. use dataset detail to inspect the fields
  3. move on to data filter, data aggregate, data create, or data update

Example:

  • the business question is “find paid but unshipped orders”
  • the Agent first identifies the order dataset
  • then checks whether fields like payStatus, deliveryStatus, and warehouseName exist
  • only then moves on to filtering and grouping

About dataset codes​

A dataset code is a 32-character hexadecimal string.

The safest approach is always:

  1. find it from dataset list
  2. copy it into the next command

Do not guess it and do not type it from memory.

A practical note for business-facing usage​

Business users usually should not ask:

“Find the dataset code for me.”

A more natural request is:

  • help me find where customer data lives
  • check whether the order dataset has a warehouse field
  • inspect the ticket object in this system first

Those requests often map to dataset list and dataset detail behind the scenes.