ticket-support-env / openenv.yaml
khushiii02's picture
Upload 16 files
82f89f0 verified
name: support-ticket-agent
version: "1.0.0"
description: >
Real-world customer support ticket triage environment.
An AI agent reads incoming support tickets and must classify the department,
assign priority, and draft a professional first reply.
Powered by the Tobi-Bueck/customer-support-tickets dataset (HuggingFace).
tags:
- openenv
- support
- triage
- nlp
- classification
author: "The Avengers"
tasks:
- id: task1
name: "Department Classification"
description: "Classify the support ticket into the correct department (Easy)."
difficulty: easy
max_steps: 20
reward_range: [0.0, 1.0]
- id: task2
name: "Classification + Priority"
description: "Classify department AND assign priority 1/2/3 (Medium)."
difficulty: medium
max_steps: 20
reward_range: [0.0, 1.0]
- id: task3
name: "Triage + Draft Reply"
description: "Classify, assign priority, AND write a professional first reply (Hard)."
difficulty: hard
max_steps: 20
reward_range: [0.0, 1.0]
observation:
type: object
fields:
ticket_id: string
subject: string
body: string
customer_name: string
task_id: string
step: integer
max_steps: integer
valid_departments: array
instructions: string
action:
type: object
fields:
department:
type: string
options: [Technical, Billing, Product, IT, Returns, Sales, HR]
priority:
type: integer
options: [1, 2, 3]
reply:
type: string
description: "Required for task3 only"
reward:
type: float
range: [0.0, 1.0]
description: >
Task1: binary department match (1.0 or 0.0).
Task2: weighted department (0.6) + priority (0.4).
Task3: weighted department (0.4) + priority (0.3) + reply quality (0.3).
endpoints:
health: GET /health
reset: POST /reset
step: POST /step
state: GET /state
tasks: GET /tasks
dataset:
name: "Tobi-Bueck/customer-support-tickets"
source: "https://huggingface.co/datasets/Tobi-Bueck/customer-support-tickets"
license: "open"