refactor(agents): move llm_config + prompt_caching to app/agents/shared (slice 4b)

Relocate the mutually-dependent LLM config layer and the LiteLLM prompt-caching
helper to the shared kernel as one unit, rewiring their internal cross-reference
to the shared paths. Flip 21 non-frozen importers. Re-export shims remain at
new_chat/{llm_config,prompt_caching}.py for the frozen single-agent stack
(chat_deepagent); they will be removed when that stack is retired.

surfsense_backend/tests/unit/agents/new_chat/test_prompt_caching.py

@@ -1,5 +1,5 @@
 r"""Tests for ``apply_litellm_prompt_caching`` in
-:mod:`app.agents.new_chat.prompt_caching`.
+:mod:`app.agents.shared.prompt_caching`.
 
 The helper replaces the legacy ``AnthropicPromptCachingMiddleware`` (which
 never activated for our LiteLLM stack) with LiteLLM-native multi-provider
@@ -34,8 +34,8 @@ from typing import Any
 
 import pytest
 
-from app.agents.new_chat.llm_config import AgentConfig
-from app.agents.new_chat.prompt_caching import apply_litellm_prompt_caching
+from app.agents.shared.llm_config import AgentConfig
+from app.agents.shared.prompt_caching import apply_litellm_prompt_caching
 
 pytestmark = pytest.mark.unit
 

surfsense_backend/tests/unit/agents/new_chat/test_resolve_prompt_model_name.py

@@ -17,7 +17,7 @@ from __future__ import annotations
 import pytest
 
 from app.agents.new_chat.chat_deepagent import _resolve_prompt_model_name
-from app.agents.new_chat.llm_config import AgentConfig
+from app.agents.shared.llm_config import AgentConfig
 
 pytestmark = pytest.mark.unit
 

surfsense_backend/tests/unit/automations/services/test_model_policy.py

@@ -44,7 +44,7 @@ def patched_globals(monkeypatch: pytest.MonkeyPatch):
         -2: {"id": -2, "billing_tier": "free"},
     }
     monkeypatch.setattr(
-        "app.agents.new_chat.llm_config.load_global_llm_config_by_id",
+        "app.agents.shared.llm_config.load_global_llm_config_by_id",
         lambda cid: llm_configs.get(cid),
     )
 

surfsense_backend/tests/unit/services/test_supports_image_input.py

@@ -227,7 +227,7 @@ global_llm_configs:
 
 
 def test_agent_config_from_yaml_explicit_overrides_resolver():
-    from app.agents.new_chat.llm_config import AgentConfig
+    from app.agents.shared.llm_config import AgentConfig
 
     cfg_text_only = AgentConfig.from_yaml_config(
         {
@@ -256,7 +256,7 @@ def test_agent_config_from_yaml_explicit_overrides_resolver():
 def test_agent_config_from_yaml_unannotated_uses_resolver():
     """Without an explicit YAML key, AgentConfig defers to the catalog
     resolver — for ``gpt-4o`` LiteLLM's map says supports_vision=True."""
-    from app.agents.new_chat.llm_config import AgentConfig
+    from app.agents.shared.llm_config import AgentConfig
 
     cfg = AgentConfig.from_yaml_config(
         {
@@ -275,7 +275,7 @@ def test_agent_config_auto_mode_supports_image_input():
     so users can keep their selection on Auto with a vision-capable
     deployment somewhere in the pool. The router's own `allowed_fails`
     handles non-vision deployments via fallback."""
-    from app.agents.new_chat.llm_config import AgentConfig
+    from app.agents.shared.llm_config import AgentConfig
 
     auto = AgentConfig.from_auto_mode()
     assert auto.supports_image_input is True

surfsense_backend/tests/unit/services/test_vision_llm_api_base_defense.py

@@ -61,7 +61,7 @@ async def test_get_vision_llm_global_openrouter_sets_api_base():
             return_value=cfg,
         ),
         patch(
-            "app.agents.new_chat.llm_config.SanitizedChatLiteLLM",
+            "app.agents.shared.llm_config.SanitizedChatLiteLLM",
             new=FakeSanitized,
         ),
     ):