Initial curated ITWorx_MCP_Hub source from 368fba691ea8d23403b4d56b53b5ec670d7c5020

packages/hub_runtime/supervisor.py

@@ -0,0 +1,103 @@
+"""Process-local MCP runtime supervision and recovery telemetry."""
+
+from __future__ import annotations
+
+import hashlib
+import os
+import socket
+import time
+from datetime import UTC, datetime
+from threading import Lock
+from typing import Any
+
+from mcp.server.context import CallNext, ServerRequestContext
+from mcp.shared.exceptions import MCPError
+
+
+class RuntimeSupervisor:
+    """Track transport health without turning tool failures into process failures."""
+
+    def __init__(self) -> None:
+        self._started_monotonic = time.monotonic()
+        self._started_at = datetime.now(UTC).isoformat()
+        material = f"{socket.gethostname()}:{os.getpid()}:{self._started_at}".encode()
+        self.worker_id = hashlib.sha256(material).hexdigest()[:16]
+        self.generation = os.environ.get(
+            "HUB_RUNTIME_GENERATION",
+            os.environ.get("GITEA_CONNECTOR_BUILD_SHA", "development"),
+        )[:64]
+        self.restart_count = max(int(os.environ.get("HUB_RUNTIME_RESTART_COUNT", "0")), 0)
+        self._lock = Lock()
+        self._in_flight = 0
+        self._requests = 0
+        self._failures = 0
+        self._last_error_code: str | None = None
+        self._last_fatal_error_code: str | None = None
+        self._circuit_state = "CLOSED"
+        self._recovery_count = 0
+
+    def started(self) -> None:
+        with self._lock:
+            self._in_flight += 1
+            self._requests += 1
+
+    def completed(self) -> None:
+        with self._lock:
+            self._in_flight = max(self._in_flight - 1, 0)
+            if self._circuit_state == "DEGRADED":
+                self._circuit_state = "CLOSED"
+                self._recovery_count += 1
+
+    def failed(self, code: str, *, fatal: bool) -> None:
+        with self._lock:
+            self._in_flight = max(self._in_flight - 1, 0)
+            self._failures += 1
+            self._last_error_code = code[:128]
+            if fatal:
+                self._last_fatal_error_code = code[:128]
+                self._circuit_state = "DEGRADED"
+
+    def snapshot(self) -> dict[str, Any]:
+        with self._lock:
+            return {
+                "workerId": self.worker_id,
+                "sessionId": None,
+                "sessionMode": "stateless-resumable-http",
+                "resumable": True,
+                "startedAt": self._started_at,
+                "uptimeSeconds": max(int(time.monotonic() - self._started_monotonic), 0),
+                "restartCount": self.restart_count,
+                "generation": self.generation,
+                "circuitState": self._circuit_state,
+                "recoveryCount": self._recovery_count,
+                "requests": self._requests,
+                "inFlight": self._in_flight,
+                "failures": self._failures,
+                "lastErrorCode": self._last_error_code,
+                "lastFatalErrorCode": self._last_fatal_error_code,
+            }
+
+
+SUPERVISOR = RuntimeSupervisor()
+
+
+class SupervisorMiddleware:
+    """Keep per-operation errors observable while preserving the next request."""
+
+    async def __call__(self, ctx: ServerRequestContext[Any, Any], call_next: CallNext) -> Any:
+        SUPERVISOR.started()
+        try:
+            result = await call_next(ctx)
+        except MCPError as exc:
+            data = exc.data if isinstance(exc.data, dict) else {}
+            SUPERVISOR.failed(str(data.get("code", "MCP_ERROR")), fatal=False)
+            raise
+        except Exception as exc:
+            SUPERVISOR.failed(type(exc).__name__, fatal=True)
+            raise
+        SUPERVISOR.completed()
+        return result
+
+
+def runtime_snapshot() -> dict[str, Any]:
+    return SUPERVISOR.snapshot()