Spaces:
Sleeping
Sleeping
| spec_version: 1 | |
| name: HyperBrickCaseOps | |
| env_name: supportdesk_env | |
| type: space | |
| runtime: fastapi | |
| app: server.app:app | |
| port: 8000 | |
| entry_point: server.supportdesk_environment:SupportDeskEnvironment | |
| description: Enterprise support operations environment with SLA pressure, business-impact aware triage, and primary-vs-secondary issue prioritization. | |
| endpoints: | |
| reset: /reset | |
| step: /step | |
| state: /state | |
| tasks: /tasks | |
| tasks: | |
| - id: billing_refund_easy | |
| difficulty: easy | |
| max_steps: 6 | |
| grader: | |
| type: llm | |
| prompt_template: | | |
| You are a QA judge. Score the agent's handling of a duplicate-charge refund ticket from 0.0 to 1.0. | |
| Full score if: queue=billing_ops, priority=high, issue_type=duplicate_charge, status=resolved, | |
| reply confirms refund with reference/timeline, and internal note mentions duplicate charge/refund. | |
| - id: account_takeover_medium | |
| difficulty: medium | |
| max_steps: 7 | |
| grader: | |
| type: llm | |
| prompt_template: | | |
| You are a QA judge. Score 0.0–1.0 for account-compromise recovery. | |
| Full score if: account locked, sessions invalidated, ownership verification requested, queue=trust_and_safety, | |
| status escalated/waiting_on_customer, and reply is security-appropriate (no promises). | |
| - id: api_incident_hard | |
| difficulty: hard | |
| max_steps: 8 | |
| grader: | |
| type: llm | |
| prompt_template: | | |
| You are a QA judge. Score 0.0–1.0 for production API incident triage. | |
| Full score if: request IDs/time window collected, status page mentioned, escalation to platform_engineering, | |
| queue=platform_engineering, and reply acknowledges incident. | |
| - id: regulated_export_exception_hard | |
| difficulty: hard | |
| max_steps: 8 | |
| grader: | |
| type: llm | |
| prompt_template: | | |
| You are a QA judge. Score 0.0–1.0 for regulated export exception handling. | |
| Full score if: compliance review required, no promise of approval, recipient identity/duration/safeguards requested, | |
| queue=compliance_ops, status=waiting_on_customer. | |