Add Space landing endpoint and hub checkpoint push

.gitignore

@@ -13,3 +13,7 @@ outputs/*
 wandb/
 *.egg-info/
 .env
+
+# Keep huge raw HF logs local; commit summaries instead.
+outputs/logs/hf_job_*_logs.txt
+

incident_commander_env/server/app.py

@@ -1,4 +1,4 @@
-from __future__ import annotations
+from __future__ import annotations
 
 from typing import Any
 
@@ -10,6 +10,26 @@ def create_incident_commander_environment() -> IncidentCommanderEnvironment:
     return IncidentCommanderEnvironment()
 
 
+def root_landing() -> dict[str, Any]:
+    return {
+        "name": "Incident Commander RL Arena",
+        "status": "running",
+        "description": "OpenEnv environment where LLM agents investigate incidents, choose safe mitigations, and write stakeholder updates.",
+        "endpoints": {
+            "health": "/health",
+            "reset": "/reset",
+            "step": "/step",
+            "state": "/state",
+            "docs": "/docs",
+        },
+        "submission": {
+            "training_job": "https://huggingface.co/jobs/Dar3devil/69ed03a7d2c8bd8662bce22d",
+            "best_logged_reward": 0.23,
+            "final_logged_reward": 0.2125,
+        },
+    }
+
+
 try:
     from openenv.core.env_server import create_app
 
@@ -20,12 +40,17 @@ try:
         env_name="incident_commander_env",
         max_concurrent_envs=64,
     )
+    app.get("/")(root_landing)
 except Exception:
     from fastapi import FastAPI
 
     app = FastAPI(title="Incident Commander RL Arena")
     _env = create_incident_commander_environment()
 
+    @app.get("/")
+    def root() -> dict[str, Any]:
+        return root_landing()
+
     @app.get("/health")
     def health() -> dict[str, str]:
         return {"status": "healthy"}
@@ -49,4 +74,3 @@ def main() -> None:
     import uvicorn
 
     uvicorn.run("server.app:app", host="0.0.0.0", port=8000)
-

incident_commander_env/training.py

@@ -1,4 +1,4 @@
-from __future__ import annotations
+from __future__ import annotations
 
 import argparse
 from pathlib import Path
@@ -144,6 +144,8 @@ def main() -> None:
     parser.add_argument("--gradient-accumulation-steps", type=int, default=16)
     parser.add_argument("--enable-thinking", action="store_true")
     parser.add_argument("--no-vllm", action="store_true", help="Fallback for T4/Colab installs where vLLM fails.")
+    parser.add_argument("--push-to-hub", action="store_true", help="Push final model/checkpoint to the Hugging Face Hub.")
+    parser.add_argument("--hub-model-id", default=None, help="Target Hub repo id for --push-to-hub.")
     args = parser.parse_args()
 
     from trl import GRPOConfig, GRPOTrainer
@@ -167,6 +169,8 @@ def main() -> None:
         vllm_mode="colocate",
         chat_template_kwargs={"enable_thinking": args.enable_thinking},
         report_to=[],
+        push_to_hub=args.push_to_hub,
+        hub_model_id=args.hub_model_id,
     )
     trainer = GRPOTrainer(
         model=args.model,
@@ -177,6 +181,8 @@ def main() -> None:
     )
     trainer.train()
     trainer.save_model(str(output_dir / "final"))
+    if args.push_to_hub:
+        trainer.push_to_hub(commit_message="Add incident commander GRPO checkpoint")
 
 
 if __name__ == "__main__":