fix: remove background init timeout and allow skills reload retry

- Remove 5s thread.join() timeout so git clone/pull is not truncated
- Remove _init_complete flag, reset _init_in_progress on failure instead
- Let /reload/skills API retry initialization after network failure
- Raise GIT_HTTP_LOW_SPEED_LIMIT from 1 KB/s to 10 KB/s

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

gns3server/agent/__init__.py

@@ -37,26 +37,13 @@ try:
     from .gns3_copilot.project_agent_manager import ProjectAgentManager
     AI_COPILOT_AVAILABLE = True
 
-    # Start skills repository initialization in background with 5s timeout.
-    # This clones/pulls GNS3-Skills during server startup.
-    # If GitHub is unreachable (common in some regions), the timeout
-    # ensures the server starts without delay, using fallback defaults.
-    def _init_skills_background():
-        """Initialize skills repo - called from background thread."""
-        from gns3server.agent.gns3_copilot.skills.registry import _ensure_skills_manager
+    # Start skills repository initialization in background.
+    # Clones/pulls GNS3-Skills during server startup.
+    # Git network timeouts are configured in SkillsManager (_GIT_TIMEOUT_ENV).
+    # The daemon thread auto-terminates when the process exits.
+    from gns3server.agent.gns3_copilot.skills.registry import _ensure_skills_manager
 
-        t = threading.Thread(target=_ensure_skills_manager, daemon=True)
-        t.start()
-        t.join(5)
-        if t.is_alive():
-            log = logging.getLogger(__name__)
-            log.warning(
-                "Skills repository initialization timed out (5s). "
-                "Will use fallback defaults. "
-                "The background thread will complete when network is available."
-            )
-
-    threading.Thread(target=_init_skills_background, daemon=True).start()
+    threading.Thread(target=_ensure_skills_manager, daemon=True).start()
 
 except ImportError as e:
     # AI dependencies not installed, disable AI Copilot feature

gns3server/agent/gns3_copilot/skills/manager.py

@@ -51,7 +51,7 @@ logger = logging.getLogger(__name__)
 _GIT_TIMEOUT_ENV = {
     'GIT_HTTP_TIMEOUT': '10',           # Connection timeout (default: 120s)
     'GIT_HTTP_LOW_SPEED_TIME': '5',     # Slow speed threshold window
-    'GIT_HTTP_LOW_SPEED_LIMIT': '1000', # < 1 KB/s = slow → abort
+    'GIT_HTTP_LOW_SPEED_LIMIT': '10240', # < 10 KB/s = slow → abort
 }
 
 

gns3server/agent/gns3_copilot/skills/registry.py

@@ -57,7 +57,6 @@ INJECTION_SKILLS_REGISTRY: dict[str, dict[str, Any]] = {}
 # Global skills manager instance for hot reload
 _skills_manager = None
 _init_in_progress = False
-_init_complete = False
 
 
 def set_skills_manager(manager):
@@ -75,16 +74,21 @@ def set_skills_manager(manager):
 
 def _ensure_skills_manager():
     """
-    Initialize the SkillsManager (runs once, idempotent).
+    Initialize the SkillsManager (idempotent, retryable on failure).
 
     Reads config, creates SkillsManager, clones/pulls repo,
     and loads skills/prompts into memory. Safe to call from
     background threads - uses _init_in_progress to prevent
     concurrent initialization.
-    """
-    global _skills_manager, _init_in_progress, _init_complete
 
-    if _skills_manager is not None or _init_complete:
+    On failure, resets _init_in_progress so future calls
+    (e.g., /reload/skills API) can retry. On success, the
+    manager is stored in _skills_manager and subsequent
+    calls return immediately.
+    """
+    global _skills_manager, _init_in_progress
+
+    if _skills_manager is not None:
         return
 
     if _init_in_progress:
@@ -130,7 +134,7 @@ def _ensure_skills_manager():
     except Exception as e:
         logger.error(f"Error initializing skills manager: {e}", exc_info=True)
     finally:
-        _init_complete = True
+        _init_in_progress = False
 
 
 def get_skills_manager():