Spaces:
Running
Running
| spec_version: 1 | |
| name: gov-workflow-openenv | |
| version: "0.3.0" | |
| type: space | |
| runtime: fastapi | |
| app: app.main:app | |
| port: 7860 | |
| description: > | |
| A real-world OpenEnv environment for government-service workflow control. | |
| The agent optimizes avoidable administrative delay via priority control, | |
| document resolution, staffing, escalation, and fairness balancing. | |
| entrypoint: | |
| module: app.main | |
| object: app | |
| inference_script: inference.py | |
| environment: | |
| class: server.gov_environment.GovWorkflowOpenEnv | |
| observation_model: app.models.ObservationModel | |
| action_model: app.models.ActionModel | |
| reward_model: app.models.RewardModel | |
| state_model: app.models.EpisodeStateModel | |
| step_info_model: app.models.StepInfoModel | |
| tasks: | |
| - id: district_backlog_easy | |
| seed: 11 | |
| description: > | |
| Small district office with 3 services and generous SLA windows. | |
| Tests baseline queue control and document handling. | |
| grader: app.graders.grade_easy | |
| - id: mixed_urgency_medium | |
| seed: 22 | |
| description: > | |
| Mid-sized office with mixed urgency and tighter fairness requirements. | |
| Tests urgency prioritization and staffing trade-offs. | |
| grader: app.graders.grade_medium | |
| - id: cross_department_hard | |
| seed: 33 | |
| description: > | |
| Large cross-department office with high arrivals and strict fairness. | |
| Tests escalation discipline and multi-queue balancing. | |
| grader: app.graders.grade_hard | |
| api: | |
| endpoints: | |
| - method: GET | |
| path: /health | |
| description: Server and session health check | |
| - method: POST | |
| path: /reset | |
| description: Create a new episode session and return initial observation | |
| - method: POST | |
| path: /step | |
| description: Apply one action and advance simulation state | |
| - method: GET | |
| path: /state | |
| description: Return current episode state (query param session_id) | |
| - method: POST | |
| path: /state | |
| description: Return current episode state (body with session_id) | |
| - method: POST | |
| path: /grade | |
| description: Run deterministic task grader for the current episode | |
| metadata: | |
| domain: government-services | |
| real_world: true | |
| reward_type: dense | |
| action_space: discrete | |
| observation_space: structured | |
| deterministic_tasks: true | |
| deterministic_graders: true | |
| num_tasks: 3 | |
| framework: fastapi | |
| language: python | |
| deployment: | |
| host: 0.0.0.0 | |
| port: 7860 | |
| dockerfile: Dockerfile | |
| platform: huggingface-spaces | |
| runtime: docker | |