Files
hermes-agent/gateway/run_plugin_rewire.py
Teknium 15c394d9ca fix(gateway): reload-plugins connects the enabled plugin's MCP servers now
`hermes plugins install` (answering "y" or `--enable`), `plugins enable` and the
Desktop all ask the running gateway to `reload-plugins`. The forced re-discovery
registered the plugin's portable MCP server configs but connected nothing: the
servers came up only on the next housekeeping reconcile tick, up to 60 s later,
while the CLI had already printed "deferred: MCP servers (next session)". Every
session started in that window came up without the plugin's MCP tools.

Live (own `gateway run --force` in a scratch home, local stdio MCP plugin
installed from a file:// repo in a tmux PTY, "y" at Enable-now, loopback model):
before this change the session started 3 s after "y" had no
`mcp__ypingprobe__yprobe_ping` in its request (it appeared only after the
gateway's reconcile tick at +44 s); with it, that session has the tool in its
tool_search catalog.

The verb now runs the same `reconcile_mcp_servers_with_config()` the
housekeeping chore runs, under the same profile scope, on a thread it waits on
for up to 10 s so a slow server cannot push the answer past the caller's 30 s
socket timeout (it keeps connecting in the background).

Invariant test: tests/gateway/test_reload_plugins_connects_mcp.py drives the
real verb against a real portable plugin dir and stdio MCP server; red on the
parent (tool absent from the registry when the verb answers), green here.
2026-10-24 18:55:33 -07:00

167 lines
8.7 KiB
Python

"""Late plugin loads re-wire the running adapters (#87770).
A plugin that finishes loading after an adapter connected — a tool-triggered force re-discovery, or
``hermes plugins install/enable`` / the Desktop / ``plugins.manage`` nudging the gateway over the control
socket — used to leave its platform handlers (slash commands, button callbacks, inbound transforms)
silently unwired until a restart. The runner subscribes to ``PluginManager.on_plugin_loaded`` per served
profile and calls every live adapter's idempotent ``rewire_plugin_handlers()`` on the event loop.
Scope is handlers only: a mid-run-loaded plugin's tools and system-prompt sections stay deferred to the
next session (prompt-cache invariant; same as ``/skills install``).
"""
from __future__ import annotations
import asyncio
import contextlib
import logging
import threading
from pathlib import Path
from typing import Any, Callable, Dict, Optional
logger = logging.getLogger(__name__)
class GatewayPluginRewireMixin:
"""Subscribe once per plugin manager; re-wire that profile's adapters on every loaded event."""
_plugin_rewire_unsubscribe: Optional[dict[str, Callable[[], None]]] = None
def _subscribe_plugin_rewire(self, manager: Any, profile_name: Optional[str] = None,
profile_home: Optional[Path] = None) -> None:
"""Idempotent per manager scope: a served-profile rescan re-enters ``_load_secondary_profile_config``
and must not stack a second listener (two listeners = two re-wire passes, still deduped, but noise)."""
subs = self._plugin_rewire_unsubscribe
if subs is None:
subs = self._plugin_rewire_unsubscribe = {}
scope = str(getattr(manager, "scope_key", "") or id(manager))
if scope in subs:
return
try:
loop = asyncio.get_running_loop()
except RuntimeError:
loop = None
def _on_loaded(_names) -> None:
# Fires on the discovering thread (tool worker, control-socket executor, or the loop itself):
# adapters own PTB/Bolt handler tables, so always hop onto the gateway loop.
if loop is None or loop.is_closed():
return
loop.call_soon_threadsafe(self._rewire_plugin_handlers, profile_name, profile_home)
subs[scope] = manager.on_plugin_loaded(_on_loaded)
def _rewire_plugin_handlers(self, profile_name: Optional[str] = None,
profile_home: Optional[Path] = None) -> int:
"""Call ``rewire_plugin_handlers()`` on every live adapter of one profile (``None`` = the launch
profile's ``self.adapters``); a secondary's adapters read their own manager, so bind its scope.
Returns the number of adapters re-wired."""
if profile_name is None:
adapters = dict(getattr(self, "adapters", None) or {})
else:
adapters = dict((getattr(self, "_profile_adapters", None) or {}).get(profile_name) or {})
if not adapters:
return 0
from gateway.run import _profile_runtime_scope
scope = (_profile_runtime_scope(profile_home, hydrate_secrets=False)
if profile_home is not None else contextlib.nullcontext())
count = 0
with scope:
for platform, adapter in adapters.items():
try:
adapter.rewire_plugin_handlers()
count += 1
except Exception:
logger.warning("[%s] plugin handler re-wire failed", getattr(platform, "value", platform),
exc_info=True)
logger.info("Re-wired plugin handlers on %d adapter(s)%s", count,
f" for profile '{profile_name}'" if profile_name else "")
return count
def reload_plugins_verb(runner: Any, loop: asyncio.AbstractEventLoop) -> Callable[..., dict]:
"""Control-socket ``reload-plugins``: force re-discovery under the requested home's scope so a plugin
installed/enabled by another process (CLI, Desktop, ``plugins.manage``) loads now; the loaded event
then re-wires that profile's adapters. Only the gateway home and served profile homes are accepted.
Answer: ``{"reloaded", "home", "plugins", "activations", "adapters_rewired"}``. Runs on the socket executor thread
(discovery is blocking); the count is read on the loop AFTER the re-wire callback (FIFO), so a
truthful "active now" reaches the caller."""
def _handler(params: Optional[dict] = None) -> dict:
from hermes_constants import get_hermes_home, hermes_home_key
from hermes_cli.plugins import discover_plugins, get_plugin_manager
from hermes_cli.plugins_activation import activation_summaries, dependency_blocked_summaries
from gateway.run import _profile_runtime_scope
params = params or {}
gateway_home = Path(get_hermes_home())
requested = Path(str(params.get("home") or gateway_home)).expanduser()
req_key = hermes_home_key(requested)
profile_name: Optional[str] = None
if req_key != hermes_home_key(gateway_home):
served = (getattr(runner, "_served_profile_homes", None) or {}).items()
profile_name = next((str(n) for n, h in served if hermes_home_key(h) == req_key), None)
if profile_name is None:
return {"reloaded": False, "error": "home is not served by this gateway", "home": str(requested)}
if profile_name == (getattr(runner, "_primary_profile_name", None) or "default"):
profile_name = None
scope = (_profile_runtime_scope(requested, hydrate_secrets=False) if profile_name
else contextlib.nullcontext())
with scope:
discover_plugins(force=True)
manager = get_plugin_manager()
# A plugin that failed on a dependency this gateway booted without rides along, so the
# enabling caller asks for the restart that loads it.
names = sorted(manager._plugins)
activations = activation_summaries(manager) + dependency_blocked_summaries(manager)
_reconcile_plugin_mcp_servers(runner, requested)
try:
rewired = asyncio.run_coroutine_threadsafe(_count_adapters(runner, profile_name), loop).result(timeout=5.0)
except Exception:
rewired = None
return {"reloaded": True, "home": str(requested), "plugins": names, "activations": activations,
"adapters_rewired": rewired}
return _handler
# Long enough for a local stdio server to connect, short enough to stay well inside the caller's
# 30 s socket timeout: a slow server keeps connecting in the background after the answer goes out.
_MCP_CONNECT_WAIT_S = 10.0
def _reconcile_plugin_mcp_servers(runner: Any, home: Path) -> None:
"""Connect the MCP servers a just-enabled plugin declares (and drop a disabled one's) now, so the
next session's tool list has them. The forced re-discovery only registers their configs; until
this, the housekeeping reconcile connected them up to a minute later, and every session started in
that window came up without the plugin's tools. Runs under the same profile scope that chore binds,
on its own thread so a slow server delays the answer by at most ``_MCP_CONNECT_WAIT_S``."""
from gateway.run import _profile_runtime_scope
from gateway.run_turn import GatewayTurnMixin
multiplex = bool(getattr(getattr(runner, "config", None), "multiplex_profiles", False))
def _reconcile() -> None:
from tools.mcp_oauth import suppress_interactive_oauth
from tools.mcp_tool_discovery import reconcile_mcp_servers_with_config
scope = _profile_runtime_scope(home) if multiplex else GatewayTurnMixin._standalone_launch_scope()
try:
with scope, suppress_interactive_oauth():
result = reconcile_mcp_servers_with_config()
except Exception:
logger.warning("MCP reconcile after reload-plugins failed for %s", home, exc_info=True)
return
if result["removed"] or result["added"]:
logger.info("MCP servers reconciled after reload-plugins (%s): removed=%s added=%s",
home, result["removed"], result["added"])
thread = threading.Thread(target=_reconcile, name="reload-plugins-mcp", daemon=True)
thread.start()
thread.join(timeout=_MCP_CONNECT_WAIT_S)
async def _count_adapters(runner: Any, profile_name: Optional[str]) -> int:
"""Live adapters for the profile; scheduled after the loaded-event callback, so they are re-wired."""
if profile_name is None:
return len(getattr(runner, "adapters", None) or {})
return len((getattr(runner, "_profile_adapters", None) or {}).get(profile_name) or {})