feat: instrument agent run startup for hang diagnosis

packages/uipath/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "uipath"
-version = "2.13.13"
+version = "2.13.14"
 description = "Python SDK and CLI for UiPath Platform, enabling programmatic interaction with automation services, process management, and deployment tools."
 readme = { file = "README.md", content-type = "text/markdown" }
 requires-python = ">=3.11"

packages/uipath/src/uipath/_cli/__init__.py

@@ -1,3 +1,4 @@
+import faulthandler
 import importlib.metadata
 import os
 import sys
@@ -9,6 +10,30 @@
 from uipath._utils._logs import setup_logging
 from uipath.platform.constants import DOTENV_FILE
 
+
+def _arm_startup_traceback_watchdog() -> None:
+    """Dump all thread stacks to stderr if startup blocks past a threshold.
+
+    Armed at import time (before the lazy command modules load) so a hang
+    in module import or runtime construction is captured in the job logs,
+    where a stalled process is otherwise silent. Cancelled by the run
+    command once the runtime is ready, so a healthy run never dumps.
+    Opt-in via UIPATH_STARTUP_TRACEBACK_SECONDS; auto-enabled in the
+    serverless runtime with a conservative default.
+    """
+    raw = os.environ.get("UIPATH_STARTUP_TRACEBACK_SECONDS")
+    if raw is None and os.environ.get("UIPATH_ENVIRONMENT_TAG") == "Serverless":
+        raw = "120"
+    try:
+        seconds = int(raw) if raw else 0
+    except ValueError:
+        seconds = 0
+    if seconds > 0:
+        faulthandler.dump_traceback_later(seconds, repeat=True, file=sys.stderr)
+
+
+_arm_startup_traceback_watchdog()
+
 # Windows console uses codepages (e.g. cp1252) that can't encode Unicode
 # characters used by Rich spinners (Braille) and emoji output.
 # When piped (not a TTY), keep the system encoding so the parent process

packages/uipath/src/uipath/_cli/cli_run.py

@@ -1,4 +1,6 @@
 import asyncio
+import faulthandler
+import logging
 from typing import Any
 
 import click
@@ -41,6 +43,7 @@
 from .middlewares import Middlewares
 
 console = ConsoleLogger()
+logger = logging.getLogger(__name__)
 
 
 class _RunDiscoveryError(EntrypointDiscoveryException):
@@ -222,16 +225,19 @@ async def execute() -> None:
                         factory: UiPathRuntimeFactoryProtocol | None = None
                         governance_bootstrap: GovernanceBootstrap | None = None
                         try:
+                            logger.info("[startup] resolving runtime factory")
                             factory = UiPathRuntimeFactoryRegistry.get(context=ctx)
 
                             resolved_entrypoint = entrypoint
                             if not resolved_entrypoint:
+                                logger.info("[startup] discovering entrypoint")
                                 available = factory.discover_entrypoints()
                                 if len(available) == 1:
                                     resolved_entrypoint = available[0]
                                 else:
                                     raise _RunDiscoveryError(available)
 
+                            logger.info("[startup] loading factory settings")
                             factory_settings = await factory.get_settings()
                             trace_settings = (
                                 factory_settings.trace_settings
@@ -248,6 +254,7 @@ async def execute() -> None:
                                 if factory_settings
                                 else None
                             )
+                            logger.info("[startup] resolving governance")
                             governance_bootstrap = await resolve_governance(
                                 agent_framework=agent_framework,
                                 agent_type=agent_type,
@@ -262,11 +269,18 @@ async def execute() -> None:
                                     governance_bootstrap.evaluator
                                 )
 
+                            logger.info(
+                                "[startup] building runtime and compiling graph "
+                                "(entrypoint=%s)",
+                                resolved_entrypoint,
+                            )
                             base_runtime = await factory.new_runtime(
                                 resolved_entrypoint,
                                 governance_runtime_id,
                                 **new_runtime_kwargs,
                             )
+                            logger.info("[startup] runtime ready, starting execution")
+                            faulthandler.cancel_dump_traceback_later()
 
                             if governance_bootstrap is not None:
                                 base_runtime = governance_bootstrap.wrap_runtime(