Data Operations in Practice — CRUD, Filtering, and Aggregation

This page is mainly for Skill authors, delivery teams, and Agent maintainers. Business users generally just describe "what I want to query or change" — the actual data commands are usually executed by the Agent in the background.

---

What the data commands do

data is the most-used command family in the Lovrabet Runtime CLI, split into two categories:

• Read: list records, fetch a single record, compute statistics
• Write: create, correct, delete

The commands are:

• data filter
• data getOne
• data aggregate
• data create
• data batchCreate
• data update
• data delete

---

What all data commands share

They all revolve around two parameters:

• --code: the dataset code, required
• --params: a JSON parameter that drives filters, pagination, write payloads, and so on

In other words, the Agent typically knows (or first discovers) a dataset code, then executes the actual read or write with --params.

---

Read operations: list, single record, statistics

1. data filter

Good for:

• Fetching a batch of records
• Conditional filtering
• Pulling detail lists

lovrabet data filter --code <datasetCode> --params '{"where":{"status":{"$eq":"active"}},"currentPage":1,"pageSize":20}'

Typical scenarios:

• Unshipped orders from the last 7 days
• Customers added in the last 30 days with no first order yet
• All tickets in a given status

2. data getOne

Good for:

• Fetching one record by id or a single primary key
• Checking what state a specific record is actually in

lovrabet data getOne --code <datasetCode> --params '{"id":123}'

3. data aggregate

Good for:

• Summarizing by dimension
• Producing operational stats, reconciliation figures, or management dashboard inputs

lovrabet data aggregate --code <datasetCode> --params '{
  "aggregate":[{"field":"amount","type":"SUM","alias":"total"}],
  "groupBy":["status"]
}'

---

Write operations: create, correct, delete

1. data create

Creates a new record:

lovrabet data create --code <datasetCode> --params '{"name":"test"}'

2. data batchCreate

Creates multiple records at once:

lovrabet data batchCreate --code <datasetCode> --params '[{"name":"a"},{"name":"b"}]'

3. data update

Modifies an existing record:

lovrabet data update --code <datasetCode> --params '{"id":123,"status":"completed"}'

4. data delete

Deletes records:

lovrabet data delete --code <datasetCode> --params '{"id":123}' --yes

---

Before any write, remember two things

First: run --dry-run

Preview write commands first:

lovrabet data update --code <datasetCode> --params '{"id":123,"status":"completed"}' --dry-run

--dry-run shows you how the API would be called and with what parameters — without executing anything.

Second: check the risk level

• create / batchCreate / update are write
• delete is high-risk-write

If the current riskLevel isn't sufficient, the CLI blocks the command. For high-risk deletes in non-interactive mode, --yes is also mandatory.

---

Common filter syntax

data filter most commonly uses where conditions:

Operator	Meaning
$eq	equals
$ne	not equals
$gt / $gte	greater than / greater than or equal
$lt / $lte	less than / less than or equal
$in	in a set
$contain	contains a substring
$startWith / $endWith	starts with / ends with
$and / $or	combine conditions

For example:

lovrabet data filter --code <datasetCode> --params '{
  "where":{
    "$and":[
      {"amount":{"$gte":100}},
      {"status":{"$eq":"active"}}
    ]
  }
}'

---

The most common aggregation types

Type	Meaning
SUM	sum
COUNT	count
AVG	average
MAX / MIN	maximum / minimum

---

How business users should phrase requests

Instead of saying:

Run data aggregate for me

Say directly:

Summarize the unshipped order count for the last 7 days by warehouse

Or:

Change requirement 241 to "processing" — preview first, then execute

This lets the Agent decide naturally whether to use filter, aggregate, update, or start with dataset discovery.

---

One-sentence summary

The data commands carry the actual business actions:

• List records → filter
• Fetch one record → getOne
• Compute summaries → aggregate
• Create / correct → create / batchCreate / update
• Delete → delete

For business users, the point isn't memorizing these names — it's stating the business object, the conditions, the expected result, and the risk boundary clearly.