Datasets:

CJJones
/

ServiceNow_Search_NLP_2_JSON_LLM_Training_Sample

license: cc-by-sa-4.0
language:
  - en
pretty_name: ServiceNow Incident Search Conversations
size_categories:
  - n<1K
task_categories:
  - question-answering
tags:
  - llm
  - tool-calling
  - function-calling
  - servicenow
  - it-service-management
  - itsm
  - enterprise-ai
  - support-automation
  - incident-management
  - structured-output
  - api-generation
  - ai-assistant
  - enterprise-automation
  - rag-compatible
  - synthetic-data
  - prompt-engineering
  - dataset-generation
  - conversational-ai
  - devops
  - helpdesk
annotations_creators:
  - machine-generated
language_creators:
  - machine-generated
multilinguality:
  - monolingual
source_datasets:
  - synthetic
domain:
  - enterprise
  - information-technology
  - customer-support
  - devops
task_domain:
  - enterprise-ai
  - workflow-automation
author:
  - C. J. Jones
dataset_type:
  - conversational
  - structured-query
version: 1.0.0

This dataset contains structured User → Bot conversations demonstrating how a natural language request can be translated into a structured ServiceNow incident search API call.

The full CJ Jones' synthetic dataset catalog is available at: https://datadeveloper1.gumroad.com

Each record consists of a user requesting incident data from an IT service management system and a bot responding with a JSON query specification compatible with the ServiceNow Table API.

The dataset is designed for training and evaluating LLM tool-use capabilities, specifically:

Natural language → API query translation

Incident ticket search automation

IT service desk assistant systems

Enterprise workflow copilots

Retrieval query generation

The bot responses strictly follow a structured schema that represents a ServiceNow incident table search operation.

This dataset preview includes 10 example conversation pairs.

Supported Tasks

tool-use

function calling

information extraction

enterprise automation

natural language query translation

retrieval query generation

Languages

English

Dataset Structure

Each sample contains two fields:

{ "user": string, "bot": string } Field Descriptions Field Description user Natural language request asking for incident or ticket history bot Structured JSON tool call specifying the ServiceNow search parameters Example User: Could you get trouble tickets that mention API integration failures regarding Seattle office staff ranking 2 for post-mortem review.

Bot: { "action": "servicenow.table.search", "parameters": { "tableName": "incident", "query": { "short_description": "CONTAINS: API integration failures", "location": "Seattle office", "priority": "2" }, "sysparm_limit": "21", "sysparm_fields": "number,short_description,description,priority,sys_created_on,assignment_group,location,state", "sysparm_display_value": "false" } } Data Instances

Example dataset record:

{ "user": "Locate employee reports about wireless access point outages in Marketing group for Sao Paulo office region Critical priority incidents for compliance purposes.", "bot": "{ "action": "servicenow.table.search", "parameters": { "tableName": "incident", "query": { "description": "CONTAINS: wireless access point outages", "assignment_group": "Marketing", "location": "Sao Paulo office", "priority": "1" }, "sysparm_limit": "39", "sysparm_fields": "number,short_description,description,priority,sys_created_on,assignment_group,location,state", "sysparm_display_value": "true" }}" } Dataset Creation Source

The conversations were synthetically generated using programmatic generation techniques designed to simulate realistic enterprise IT service desk queries.

Generation introduces variation across:

issue categories

office locations

organizational departments

priority levels

request phrasing

reporting contexts

Example issue types include:

API integration failures

wireless access point outages

network switch port errors

single sign-on errors

firewall configuration errors

power supply failures

password reset requests

email delivery issues

Generation Strategy

Records were generated to maximize variation in:

user intent phrasing

department references

location mentions

priority terminology

reporting context

The output schema enforces consistent structure compatible with ServiceNow incident table queries.

Intended Use

This dataset is intended for:

training LLM agents that interact with enterprise systems

benchmarking tool-calling accuracy

developing AI service desk assistants

research on structured query generation

👤 Creator C.J. Jones AI engineer and developer of domain-specific synthetic datasets for reasoning and diagnostic training in LLMs.

Disclaimer: There are no implied guarrantees and user must accept all risk and resposibilities regarding the use of this and any other datasets provided by CJ Jones.

If you liked this, you may also be interested in: