refactor(agents): move mac-only modules out of the cross-agent shared kernel

app/agents/shared/ is a sibling of anonymous_chat/podcaster/multi_agent_chat/
video_presentation, so it should only hold code shared across 2+ of those
agents. In practice podcaster and video_presentation import nothing from it,
and anonymous_chat needs only context + compaction + retry_after + web_search.
Everything else was multi_agent_chat-only (the boundary just passes through).

Move the multi_agent_chat-only cluster into multi_agent_chat/shared/ (files
moved verbatim via git rename; ~116 import sites rewritten):

  errors, feature_flags, filesystem_selection, path_resolver, prompt_caching,
  sandbox, llm_config, mention_resolver
  middleware/busy_mutex, middleware/kb_persistence

busy_mutex/llm_config/mention_resolver are boundary-only but import the moved
modules, so they were folded in to avoid a backwards shared -> multi_agent_chat
dependency. main_agent builders now import the impls directly; the shared
middleware barrel keeps only the genuinely-shared compaction + retry_after.

Also delete the dead leftover shared/plugins and shared/skills dirs (live
copies already live under main_agent/).

Remaining in app/agents/shared/: context, system_prompt(+prompts), checkpointer,
middleware/{compaction,retry_after,dedup_tool_calls}, tools/. checkpointer and
system_prompt are boundary-only infra pending a dedicated home decision.

surfsense_backend/app/services/llm_service.py

@@ -204,7 +204,7 @@ async def validate_llm_config(
         if litellm_params:
             litellm_kwargs.update(litellm_params)
 
-        from app.agents.shared.llm_config import SanitizedChatLiteLLM
+        from app.agents.multi_agent_chat.shared.llm_config import SanitizedChatLiteLLM
 
         llm = SanitizedChatLiteLLM(**litellm_kwargs)
 
@@ -379,7 +379,9 @@ async def get_search_space_llm_instance(
             if disable_streaming:
                 litellm_kwargs["disable_streaming"] = True
 
-            from app.agents.shared.llm_config import SanitizedChatLiteLLM
+            from app.agents.multi_agent_chat.shared.llm_config import (
+                SanitizedChatLiteLLM,
+            )
 
             return SanitizedChatLiteLLM(**litellm_kwargs)
 
@@ -458,7 +460,7 @@ async def get_search_space_llm_instance(
         if disable_streaming:
             litellm_kwargs["disable_streaming"] = True
 
-        from app.agents.shared.llm_config import SanitizedChatLiteLLM
+        from app.agents.multi_agent_chat.shared.llm_config import SanitizedChatLiteLLM
 
         return SanitizedChatLiteLLM(**litellm_kwargs)
 
@@ -580,7 +582,9 @@ async def get_vision_llm(
             if global_cfg.get("litellm_params"):
                 litellm_kwargs.update(global_cfg["litellm_params"])
 
-            from app.agents.shared.llm_config import SanitizedChatLiteLLM
+            from app.agents.multi_agent_chat.shared.llm_config import (
+                SanitizedChatLiteLLM,
+            )
 
             inner_llm = SanitizedChatLiteLLM(**litellm_kwargs)
 
@@ -634,7 +638,7 @@ async def get_vision_llm(
         if vision_cfg.litellm_params:
             litellm_kwargs.update(vision_cfg.litellm_params)
 
-        from app.agents.shared.llm_config import SanitizedChatLiteLLM
+        from app.agents.multi_agent_chat.shared.llm_config import SanitizedChatLiteLLM
 
         return SanitizedChatLiteLLM(**litellm_kwargs)
 
@@ -679,7 +683,9 @@ def get_planner_llm() -> ChatLiteLLM | None:
     Callers MUST fall back to their chat LLM when this returns ``None`` so
     deployments without a planner config keep working unchanged.
     """
-    from app.agents.shared.llm_config import create_chat_litellm_from_config
+    from app.agents.multi_agent_chat.shared.llm_config import (
+        create_chat_litellm_from_config,
+    )
 
     planner_cfg = next(
         (cfg for cfg in config.GLOBAL_LLM_CONFIGS if cfg.get("is_planner") is True),

surfsense_backend/app/services/provider_capabilities.py

@@ -53,7 +53,7 @@ logger = logging.getLogger(__name__)
 #
 # Owned here because ``app.services.provider_capabilities`` is the
 # only edge that's safe to call from ``app.config``'s YAML loader at
-# class-body init time. ``app.agents.shared.llm_config`` re-exports
+# class-body init time. ``app.agents.multi_agent_chat.shared.llm_config`` re-exports
 # this constant under the historical ``PROVIDER_MAP`` name; placing the
 # map there directly would re-introduce the
 # ``app.config -> ... -> deliverables/tools/generate_image ->

surfsense_backend/app/services/revert_service.py

@@ -38,7 +38,7 @@ from typing import Any, Literal
 from sqlalchemy import delete, select
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.agents.shared.path_resolver import (
+from app.agents.multi_agent_chat.shared.path_resolver import (
     DOCUMENTS_ROOT,
     safe_filename,
     safe_folder_segment,