forgexa-cli 1.62.2__tar.gz → 1.62.3__tar.gz

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: forgexa-cli
-Version: 1.62.2
+Version: 1.62.3
 Summary: Forgexa CLI — command-line client and AI agent runtime for the Forgexa platform
 Author-email: Jason Sun <dev.winds@gmail.com>
 License-Expression: MIT

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/forgexa_cli/__init__.py

@@ -1,2 +1,2 @@
 """forgexa-cli — Forgexa command-line client."""
-__version__ = "1.62.2"
+__version__ = "1.62.3"

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/forgexa_cli/agent_core.py

@@ -3626,7 +3626,7 @@ def is_copilot_model_discovery_auth_error(*messages: str | None) -> bool:
 
 
 def is_runtime_resource_exhaustion_error(*messages: str | None) -> bool:
-    """Return whether the OS refused to create an agent process or thread."""
+    """Return whether the OS refused an agent process, thread, or file watch."""
     detail = " ".join(str(message or "").lower() for message in messages)
     return any(
         pattern in detail
@@ -3635,6 +3635,9 @@ def is_runtime_resource_exhaustion_error(*messages: str | None) -> bool:
             "failed to load ca certificates off thread",
             "resource temporarily unavailable (os error 11)",
             "pthread_create failed",
+            "os file watch limit reached",
+            "inotify watch limit reached",
+            "inotify instance limit reached",
         )
     )
 
@@ -3644,7 +3647,7 @@ def collect_runtime_resource_metrics(
     proc_root: str | Path = "/proc",
     cgroup_root: str | Path = "/sys/fs/cgroup",
 ) -> dict[str, Any]:
-    """Collect Linux process/thread headroom without optional dependencies."""
+    """Collect Linux process/thread, memory, and file-watch capacity."""
     metrics: dict[str, Any] = {}
     if not sys.platform.startswith("linux"):
         return metrics
@@ -3714,11 +3717,36 @@ def collect_runtime_resource_metrics(
                 break
     except (OSError, ValueError, IndexError):
         pass
+
+    probe_fd: int | None = None
+    try:
+        for name in ("max_user_watches", "max_user_instances"):
+            metrics[f"inotify_{name}"] = int(
+                (proc_path / "sys" / "fs" / "inotify" / name).read_text().strip()
+            )
+        import ctypes
+
+        library = ctypes.CDLL(None, use_errno=True)
+        library.inotify_init1.argtypes = [ctypes.c_int]
+        library.inotify_init1.restype = ctypes.c_int
+        library.inotify_add_watch.argtypes = [ctypes.c_int, ctypes.c_char_p, ctypes.c_uint32]
+        library.inotify_add_watch.restype = ctypes.c_int
+        probe_fd = library.inotify_init1(os.O_CLOEXEC | os.O_NONBLOCK)
+        if probe_fd < 0:
+            metrics["inotify_probe_errno"] = ctypes.get_errno()
+        else:
+            watch = library.inotify_add_watch(probe_fd, os.fsencode(proc_path), 0x00000004)
+            metrics["inotify_probe_errno"] = ctypes.get_errno() if watch < 0 else 0
+    except (AttributeError, OSError, ValueError):
+        pass
+    finally:
+        if probe_fd is not None and probe_fd >= 0:
+            os.close(probe_fd)
     return metrics
 
 
 def runtime_resource_exhaustion_reason(metrics: dict[str, Any]) -> str | None:
-    """Return a blocking reason when an agent lacks safe process headroom."""
+    """Return a blocking reason when the OS cannot safely start an agent."""
     process_remaining = metrics.get("process_remaining")
     if isinstance(process_remaining, int) and process_remaining < 32:
         return (
@@ -3729,9 +3757,41 @@ def runtime_resource_exhaustion_reason(metrics: dict[str, Any]) -> str | None:
     memory_available_mb = metrics.get("memory_available_mb")
     if isinstance(memory_available_mb, int) and memory_available_mb < 256:
         return f"Only {memory_available_mb} MiB memory is available"
+    import errno
+
+    probe_errno = metrics.get("inotify_probe_errno")
+    if probe_errno == errno.ENOSPC:
+        return (
+            "OS file watch limit reached (inotify ENOSPC; "
+            f"fs.inotify.max_user_watches={metrics.get('inotify_max_user_watches')})"
+        )
+    if probe_errno in {errno.EMFILE, errno.ENFILE}:
+        return (
+            "Inotify instance or file descriptor limit reached "
+            f"({errno.errorcode[probe_errno]}; "
+            f"fs.inotify.max_user_instances={metrics.get('inotify_max_user_instances')})"
+        )
     return None
 
 
+def runtime_resource_exhaustion_remediation(metrics: dict[str, Any]) -> str:
+    """Suggest recovery for the specific resource rejected by the kernel."""
+    import errno
+
+    probe_errno = metrics.get("inotify_probe_errno")
+    if probe_errno == errno.ENOSPC:
+        return (
+            "Inspect and stop unneeded file watchers owned by the daemon user, "
+            "or raise fs.inotify.max_user_watches, then retry"
+        )
+    if probe_errno in {errno.EMFILE, errno.ENFILE}:
+        return (
+            "Close unused watcher processes or file descriptors; check "
+            "fs.inotify.max_user_instances and the file descriptor limits, then retry"
+        )
+    return "Stop orphaned agent processes or raise the user/cgroup PID limit, then retry"
+
+
 def copilot_auth_state_signature(copilot_home: str | Path) -> str | None:
     """Return a content signature for Copilot's persisted authentication state."""
     config_file = Path(copilot_home) / "config.json"

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/forgexa_cli/daemon.py

@@ -1196,7 +1196,7 @@ except (ImportError, ModuleNotFoundError):
 # DAEMON_VERSION is the protocol/logic version of the daemon code.
 # Kept in sync with pyproject.toml version via bump-version.sh.
 # CLIENT_TYPE identifies which packaging/distribution this daemon runs in.
-DAEMON_VERSION = "1.62.2"
+DAEMON_VERSION = "1.62.3"
 
 
 def _detect_client_type() -> str:
@@ -5919,6 +5919,8 @@ class ProcessManager:
         lower = (text or "").lower()
         if any(p in lower for p in ProcessManager.RATE_LIMIT_PATTERNS):
             return "Agent hit rate/usage limit or quota exhaustion"
+        if agent_core.is_runtime_resource_exhaustion_error(lower):
+            return "Agent host OS resources exhausted"
         if any(p in lower for p in ProcessManager.AGENT_UNAVAILABLE_PATTERNS):
             return "Agent backend unavailable or unreachable"
         return None
@@ -6015,11 +6017,13 @@ class ProcessManager:
             "all_agents_rate_limited", "all_agents_unavailable", "acceptance_criteria_uncovered",
         }:
             return None
-        if result.failure_code == "agent_platform_unsupported":
+        if result.failure_code in {"agent_platform_unsupported", "runtime_resource_exhausted"}:
             return "agent_unavailable"
         error_text = ProcessManager._failure_pattern_channels(result).lower()
         if any(p in error_text for p in ProcessManager.RATE_LIMIT_PATTERNS):
             return "rate_limit"
+        if agent_core.is_runtime_resource_exhaustion_error(error_text):
+            return "agent_unavailable"
         if any(p in error_text for p in ProcessManager.AGENT_UNAVAILABLE_PATTERNS):
             return "agent_unavailable"
         return None
@@ -6192,9 +6196,8 @@ class ProcessManager:
                 "blocking": True,
                 "failure_class": "resource_exhausted",
                 "detail": resource_reason,
-                "remediation": (
-                    "Stop orphaned agent processes or raise the user/cgroup PID limit, "
-                    "then retry"
+                "remediation": agent_core.runtime_resource_exhaustion_remediation(
+                    resource_metrics,
                 ),
             }
             logger.error(
@@ -8741,9 +8744,8 @@ class RuntimePreflightMonitor:
                 "blocking": True,
                 "failure_class": "resource_exhausted",
                 "detail": reason,
-                "remediation": (
-                    "Stop orphaned agent processes or raise the user/cgroup PID limit, "
-                    "then restart the daemon"
+                "remediation": agent_core.runtime_resource_exhaustion_remediation(
+                    metrics,
                 ),
             }
         if metrics:

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/forgexa_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: forgexa-cli
-Version: 1.62.2
+Version: 1.62.3
 Summary: Forgexa CLI — command-line client and AI agent runtime for the Forgexa platform
 Author-email: Jason Sun <dev.winds@gmail.com>
 License-Expression: MIT

{forgexa_cli-1.62.2 → forgexa_cli-1.62.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "forgexa-cli"
-version = "1.62.2"
+version = "1.62.3"
 description = "Forgexa CLI — command-line client and AI agent runtime for the Forgexa platform"
 requires-python = ">=3.9"
 license = "MIT"