Create openenv.yaml

openenv.yaml

@@ -0,0 +1,56 @@
+# ─────────────────────────────────────────────────────────────────────────────
+# openenv.yaml — Disaster Grid Environment Metadata
+# Conforms to the OpenEnv 1.0 specification
+# ─────────────────────────────────────────────────────────────────────────────
+
+name: disaster-grid-env
+version: "1.0.0"
+description: >
+  A 5x5 fully-observable grid simulation where an RL agent (Llama-3 trained via GRPO) 
+  acts as an Autonomous Emergency Manager. The agent must manage its energy 
+  to navigate and repair decaying city sectors before critical failure.
+
+author: "Sanjith & Varsha"
+license: MIT
+
+# ── interface ─────────────────────────────────────────────────────────────────
+interface:
+  # This tells the framework exactly where our Python class lives!
+  step:   "src.disaster_grid.environment.CityGrid.step"
+  reset:  "src.disaster_grid.environment.CityGrid.reset"
+
+# ── observation space (From our models.py) ────────────────────────────────────
+observation_space:
+  type: object
+  fields:
+    step_number:           {type: integer}
+    agent_position:        {type: integer, description: "Index 0-24"}
+    agent_energy:          {type: integer, description: "Scale 0-100"}
+    current_sector_health: {type: integer}
+    critical_sectors:      
+      type: array
+      items: {type: integer}
+    average_city_health:   {type: number, format: float}
+
+# ── action space (From our models.py) ─────────────────────────────────────────
+action_space:
+  type: object
+  fields:
+    action:    {type: string, enum: [MOVE_N, MOVE_S, MOVE_E, MOVE_W, REPAIR, RECHARGE, WAIT]}
+    reasoning: {type: string, description: "Chain of thought explanation"}
+
+# ── runtime ───────────────────────────────────────────────────────────────────
+runtime:
+  python: "3.10"
+  entrypoint: "app.py"
+  port: 7860
+  framework: fastapi
+
+# ── tags ──────────────────────────────────────────────────────────────────────
+tags:
+  - disaster-management
+  - grid-world
+  - reinforcement-learning
+  - grpo
+  - llama-3
+  - huggingface-spaces