Claude Code: Add /status endpoint with structured health checks (db, env, fs)

app.py

@@ -1,8 +1,10 @@
 import asyncio
 import os
 from datetime import datetime
-from fastapi import FastAPI, WebSocket, WebSocketDisconnect
+from pathlib import Path
+from fastapi import FastAPI, WebSocket, WebSocketDisconnect, HTTPException
 from fastapi.staticfiles import StaticFiles
+from fastapi.responses import JSONResponse
 
 # Initialize FastAPI application
 app = FastAPI()
@@ -27,6 +29,59 @@ async def health_check():
     }
 
 
+@app.get("/status")
+async def status_check():
+    """Comprehensive health check with db, env, and filesystem validation."""
+    checks = {"db": "ok", "env": "ok", "fs": "ok"}
+    overall_healthy = True
+
+    # 1. Database check - lightweight dataset list call
+    try:
+        from huggingface_hub import HfApi
+        api = HfApi()
+        # Lightweight call - just check if we can reach HF
+        repo_id = os.environ.get("OPENCLAW_DATASET_REPO", "")
+        if repo_id:
+            api.repo_info(repo_id=repo_id, repo_type="dataset")
+            checks["db"] = "ok"
+        else:
+            checks["db"] = "error: OPENCLAW_DATASET_REPO not set"
+            overall_healthy = False
+    except Exception as e:
+        checks["db"] = f"error: {str(e)}"
+        overall_healthy = False
+
+    # 2. Environment check - verify critical variables
+    critical_vars = ["HF_TOKEN", "OPENCLAW_DATASET_REPO"]
+    missing_vars = [v for v in critical_vars if not os.environ.get(v)]
+    if missing_vars:
+        checks["env"] = f"error: missing {', '.join(missing_vars)}"
+        overall_healthy = False
+    else:
+        checks["env"] = "ok"
+
+    # 3. Filesystem check - touch and delete test file
+    test_file = Path("/tmp/test_write")
+    try:
+        test_file.touch()
+        test_file.unlink()
+        checks["fs"] = "ok"
+    except Exception as e:
+        checks["fs"] = f"error: {str(e)}"
+        overall_healthy = False
+
+    status = "healthy" if overall_healthy else "degraded"
+
+    # Return HTTP 503 if critical checks fail
+    if not overall_healthy:
+        raise HTTPException(status_code=503, detail={
+            "status": status,
+            "checks": checks
+        })
+
+    return {"status": status, "checks": checks}
+
+
 @app.websocket("/ws/agents")
 async def websocket_agents(websocket: WebSocket):
     """WebSocket endpoint for real-time agent status updates."""