spec_version: 1
name: customer_support_env
type: space
runtime: fastapi
app: server.app:app
port: 8000

# Environment metadata
title: "Customer Support Ticket Management Environment"
description: "Real-world OpenEnv environment for training AI agents on customer support tasks"
version: "1.0.0"

# Tags for Hugging Face Spaces
tags:
  - openenv
  - customer-support
  - reinforcement-learning
  - agent-training

# Task configurations
tasks:
  - id: easy
    name: "Ticket Classification"
    description: "Categorize support tickets into the correct category (billing, technical, account, shipping, general)"
    difficulty: easy
    max_steps: 1
    success_threshold: 0.8

  - id: medium
    name: "Priority Assignment & Routing"
    description: "Categorize tickets, assign correct priority level, and route to the appropriate support team"
    difficulty: medium
    max_steps: 1
    success_threshold: 0.75

  - id: hard
    name: "Complete Ticket Resolution"
    description: "Fully resolve tickets: correct categorization, priority, routing, and draft a professional, helpful response"
    difficulty: hard
    max_steps: 1
    success_threshold: 0.70

# Environment configuration
config:
  action_space:
    category:
      - billing
      - technical
      - account
      - shipping
      - general
    priority:
      - low
      - medium
      - high
      - critical
    assigned_team:
      - tier1
      - tier2
      - billing
      - technical
      - management
    response_draft: "string (min 10 characters)"
    internal_notes: "string (optional)"
    escalate: "boolean"

  observation_space:
    ticket_metadata: "TicketMetadata (ID, timestamp, customer_id, channel)"
    customer_message: "string (support request)"
    customer_history: "CustomerHistory (account_age, tickets, satisfaction, premium status, lifetime value)"
    previous_interactions: "List[str]"
    attachments: "List[str]"

  reward_function:
    category_correctness: 0.25
    priority_correctness: 0.20
    team_routing_correctness: 0.25
    response_quality: 0.20
    efficiency_bonus: "up to 0.15"
    penalties: "up to -0.15"