Upload folder using huggingface_hub

README.md

@@ -150,16 +150,17 @@ The environment has 5 mechanisms that prevent reward hacking:
 
 ## Baseline Scores
 
-| Strategy | Task 1 | Task 2 | Task 3 |
-|----------|--------|--------|--------|
-| **Oracle (rule-based)** | **0.79** | **0.81** | **0.70** |
-| Do-Nothing (grid only) | 0.58 | 0.51 | 0.45 |
-| Always-Discharge | 0.59 | 0.51 | 0.45 |
-| Always-Diesel | 0.42 | 0.42 | 0.44 |
+| Strategy | Task 1 | Task 2 | Task 3 | What it does |
+|----------|--------|--------|--------|-------------|
+| **Oracle (rule-based)** | **0.79** | **0.81** | **0.70** | Time-of-day + price + SOC aware |
+| Do-Nothing (grid only) | 0.58 | 0.51 | 0.45 | Grid covers everything it can |
+| Always-Discharge | 0.59 | 0.51 | 0.45 | Drains battery, empty by evening |
+| Always-Diesel | 0.42 | 0.42 | 0.44 | Rs 25/kWh burns money |
 
 - **Deterministic**: identical scores across 3 runs (seeded RNG)
-- **Oracle ceiling is < 1.0**: proves the environment has real physics constraints, not inflated scores
-- **Clear separation**: oracle >> heuristics on every task
+- **Oracle ceiling < 1.0**: real physics constraints, not inflated scores
+- **Clear separation**: oracle >> heuristics on every task (0.20-0.35 gap)
+- **Task 3 hardest**: grid outage drops oracle from 0.81 to 0.70
 
 ---
 

gridops/server/app.py

@@ -79,19 +79,22 @@ def list_tasks():
                 "id": "task_1_normal",
                 "name": "Normal Summer",
                 "difficulty": "Easy",
-                "description": "Clear skies, standard demand, Rs 3-12 price range.",
+                "description": "Clear skies, ~100 kW avg demand, Rs 3-12 prices. Tests basic battery arbitrage.",
+                "oracle_score": 0.79,
             },
             {
                 "id": "task_2_heatwave",
-                "name": "Heatwave + Clouds",
+                "name": "Heatwave + Price Spike",
                 "difficulty": "Medium",
-                "description": "Day 2-3 heatwave, +30% demand, price spikes to Rs 18.",
+                "description": "Day 2-3 heatwave (+30% demand), Rs 20 price spike. Tests temporal planning via forecast.",
+                "oracle_score": 0.81,
             },
             {
                 "id": "task_3_crisis",
-                "name": "Extreme Crisis",
+                "name": "Extreme Crisis + Grid Outage",
                 "difficulty": "Hard",
-                "description": "Full 3-day heatwave, -30% solar, prices Rs 8-20, limited diesel.",
+                "description": "Full 3-day heatwave, -30% solar, +50% demand, limited diesel, 6-hour grid outage. Tests islanding.",
+                "oracle_score": 0.70,
             },
         ]
     }

gridops/server/environment.py

@@ -168,10 +168,19 @@ class GridOpsEnvironment(Environment):
         )
 
     def get_metadata(self) -> EnvironmentMetadata:
+        readme = None
+        try:
+            from pathlib import Path
+            readme_path = Path(__file__).parent.parent.parent / "README.md"
+            if readme_path.exists():
+                readme = readme_path.read_text()
+        except Exception:
+            pass
         return EnvironmentMetadata(
             name="GridOps",
             description="Community microgrid bridge operator — balance solar, battery, diesel, and grid across 3-day episodes.",
             version="0.2.0",
+            readme_content=readme,
         )
 
     def _make_observation(self, reward: float, done: bool, narration: str) -> GridOpsObservation:

openenv.yaml

@@ -1,6 +1,6 @@
 name: gridops
-description: "Community microgrid bridge operator — balance solar, battery, diesel, and grid trade across 3-day summer episodes in India."
-version: "0.1.0"
+description: "Community microgrid bridge operator — balance solar, battery, diesel, and grid across 3-day summer episodes in India."
+version: "0.2.0"
 
 environment:
   module: gridops.server.app
@@ -19,21 +19,23 @@ tasks:
   - id: task_1_normal
     name: "Normal Summer"
     difficulty: easy
-    description: "Clear skies, standard demand curve, smooth IEX prices Rs 4-8."
+    description: "Clear skies, ~100 kW avg demand, 250 kW evening peak, IEX prices Rs 3-12. Tests basic battery arbitrage."
 
   - id: task_2_heatwave
-    name: "Heatwave + Clouds"
+    name: "Heatwave + Price Spike"
     difficulty: medium
-    description: "Day 2-3 heatwave (+30% demand), intermittent clouds, price spikes to Rs 18."
+    description: "Day 2-3 heatwave (+30% demand), intermittent clouds, Rs 20 evening price spike. Tests temporal planning via 4h forecast."
 
   - id: task_3_crisis
-    name: "Extreme Crisis"
+    name: "Extreme Crisis + Grid Outage"
     difficulty: hard
-    description: "Full 3-day heatwave, -30% solar, Rs 8-20 prices, limited diesel fuel."
+    description: "Full 3-day heatwave, -30% solar, +50% demand, limited diesel (33%), 6-hour grid outage on Day 2. Tests islanding and constraint management."
 
 metadata:
   action_space: "3D continuous: battery_dispatch [-1,1], diesel_dispatch [0,1], demand_shedding [0,1]"
-  observation_space: "12 fields: hour, demand, solar, battery_soc, price, fuel, 3×4h forecasts, cumulative metrics"
+  observation_space: "30+ fields: hour, demand, solar, battery_soc, price, fuel, diesel_is_on, 3x4h noisy forecasts, cumulative metrics, detailed energy flows"
   episode_length: 72
   step_duration: "1 hour"
+  episode_start: "6 AM Day 1"
   grading: "Composite: 50% cost efficiency + 25% reliability + 25% green score. VoLL Rs 150/kWh blackout penalty."
+  anti_gaming: "Shedding Rs 40/kWh + 100% rebound, battery Rs 2.5/kWh degradation, diesel Rs 100 startup, smooth VoLL, grid cap 200 kW"