fix(types): batch P1 ty hotfixes + run_agent.py annotation pass

15 P1 ship-stopper runtime bugs from the ty triage plus the cross-bucket
cleanup in run_agent.py. Net: -138 ty diagnostics (1953 -> 1815). Major
wins on not-subscriptable (-34), unresolved-attribute (-29),
invalid-argument-type (-26), invalid-type-form (-20),
unsupported-operator
(-18), invalid-key (-9).

Missing refs (structural):
- tools/rl_training_tool.py: RunState dataclass gains api_log_file,
  trainer_log_file, env_log_file fields; stop-run was closing undeclared
  handles.
- agent/credential_pool.py: remove_entry(entry_id) added, symmetric with
  add_entry; used by hermes_cli/web_server.py OAuth dashboard cleanup.
- hermes_cli/config.py: _CamofoxConfig TypedDict defined (was referenced
  by _BrowserConfig but never declared).
- hermes_cli/gateway.py: _setup_wecom_callback() added, mirroring
  _setup_wecom().
- tui_gateway/server.py: skills_hub imports corrected from
  hermes_cli.skills_hub -> tools.skills_hub.

Typo / deprecation:
- tools/transcription_tools.py: os.sys.modules -> sys.modules.
- gateway/platforms/bluebubbles.py: datetime.utcnow() ->
  datetime.now(timezone.utc).

None-guards:
- gateway/platforms/telegram.py:~2798 - msg.sticker None guard.
- gateway/platforms/discord.py:3602/3637 - interaction.data None +
  SelectMenu narrowing; :3009 - thread_id None before `in`; :1893 -
  guild.member_count None.
- gateway/platforms/matrix.py:2174/2185 - walrus-narrow
  re.search().group().
- agent/display.py:732 - start_time None before elapsed subtraction.
- gateway/run.py:10334 - assert _agent_timeout is not None before `//
  60`.

Platform override signature match:
- gateway/platforms/email.py: send_image accepts metadata kwarg;
  send_document accepts **kwargs (matches base class).

run_agent.py annotation pass:
- callable/any -> Callable/Any in annotation position (15 sites in
  run_agent.py + 5 in cli.py, toolset_distributions.py,
  tools/delegate_tool.py, hermes_cli/dingtalk_auth.py,
  tui_gateway/server.py).
- conversation_history param widened to list[dict[str, Any]] | None.
- OMIT_TEMPERATURE sentinel guarded from leaking into
  call_llm(temperature): kwargs-dict pattern at run_agent.py:7337 +
  scripts/trajectory_compressor.py:618/688.
- build_anthropic_client(timeout) widened to Optional[float].

Tests:
- tests/agent/test_credential_pool.py: remove_entry (id match,
  unknown-id, priority renumbering).
- tests/hermes_cli/test_config_shapes.py: _CamofoxConfig shape +
  nesting.
- tests/tools/test_rl_training_tool.py: RunState log_file fields.

run_agent.py

@@ -37,7 +37,7 @@ import time
 import threading
 from types import SimpleNamespace
 import uuid
-from typing import List, Dict, Any, Optional, TYPE_CHECKING
+from typing import Callable, List, Dict, Any, Optional, TYPE_CHECKING
 
 if TYPE_CHECKING:
     from agent.rate_limit_tracker import RateLimitState
@@ -736,17 +736,17 @@ class AIAgent:
         provider_require_parameters: bool = False,
         provider_data_collection: str = None,
         session_id: str = None,
-        tool_progress_callback: callable = None,
-        tool_start_callback: callable = None,
-        tool_complete_callback: callable = None,
-        thinking_callback: callable = None,
-        reasoning_callback: callable = None,
-        clarify_callback: callable = None,
-        step_callback: callable = None,
-        stream_delta_callback: callable = None,
-        interim_assistant_callback: callable = None,
-        tool_gen_callback: callable = None,
-        status_callback: callable = None,
+        tool_progress_callback: Callable[..., Any] = None,
+        tool_start_callback: Callable[..., Any] = None,
+        tool_complete_callback: Callable[..., Any] = None,
+        thinking_callback: Callable[..., Any] = None,
+        reasoning_callback: Callable[..., Any] = None,
+        clarify_callback: Callable[..., Any] = None,
+        step_callback: Callable[..., Any] = None,
+        stream_delta_callback: Callable[..., Any] = None,
+        interim_assistant_callback: Callable[..., Any] = None,
+        tool_gen_callback: Callable[..., Any] = None,
+        status_callback: Callable[..., Any] = None,
         max_tokens: int = None,
         reasoning_config: Dict[str, Any] = None,
         service_tier: str = None,
@@ -4688,7 +4688,7 @@ class AIAgent:
     def _close_request_openai_client(self, client: Any, *, reason: str) -> None:
         self._close_openai_client(client, reason=reason, shared=False)
 
-    def _run_codex_stream(self, api_kwargs: dict, client: Any = None, on_first_delta: callable = None):
+    def _run_codex_stream(self, api_kwargs: dict, client: Any = None, on_first_delta: Callable[..., Any] = None):
         """Execute one streaming Responses API request and return the final response."""
         import httpx as _httpx
 
@@ -5384,7 +5384,7 @@ class AIAgent:
         )
 
     def _interruptible_streaming_api_call(
-        self, api_kwargs: dict, *, on_first_delta: callable = None
+        self, api_kwargs: dict, *, on_first_delta: Callable[..., Any] = None
     ):
         """Streaming variant of _interruptible_api_call for real-time token delivery.
 
@@ -7334,12 +7334,15 @@ class AIAgent:
                 _flush_temperature = _fixed_temp
             else:
                 _flush_temperature = 0.3
+            _flush_llm_kwargs: dict = {}
+            if _flush_temperature is not None:
+                _flush_llm_kwargs["temperature"] = _flush_temperature
             try:
                 response = _call_llm(
                     task="flush_memories",
                     messages=api_messages,
                     tools=[memory_tool_def],
-                    temperature=_flush_temperature,
+                    **_flush_llm_kwargs,
                     max_tokens=5120,
                     # timeout resolved from auxiliary.flush_memories.timeout config
                 )
@@ -8531,9 +8534,9 @@ class AIAgent:
         self,
         user_message: str,
         system_message: str = None,
-        conversation_history: List[Dict[str, Any]] = None,
+        conversation_history: List[Dict[str, Any]] | None = None,
         task_id: str = None,
-        stream_callback: Optional[callable] = None,
+        stream_callback: Optional[Callable[..., Any]] = None,
         persist_user_message: Optional[str] = None,
     ) -> Dict[str, Any]:
         """
@@ -11783,7 +11786,7 @@ class AIAgent:
 
         return result
 
-    def chat(self, message: str, stream_callback: Optional[callable] = None) -> str:
+    def chat(self, message: str, stream_callback: Optional[Callable[..., Any]] = None) -> str:
         """
         Simple chat interface that returns just the final response.