feat(assistant): add multi-dialog UX, task-aware llm settings, and i18n cleanup

backend/src/services/__tests__/test_llm_prompt_templates.py

@@ -7,8 +7,12 @@
 # @INVARIANT: All required prompt keys remain available after normalization.
 
 from src.services.llm_prompt_templates import (
+    DEFAULT_LLM_ASSISTANT_SETTINGS,
+    DEFAULT_LLM_PROVIDER_BINDINGS,
     DEFAULT_LLM_PROMPTS,
+    is_multimodal_model,
     normalize_llm_settings,
+    resolve_bound_provider_id,
     render_prompt,
 )
 
@@ -22,10 +26,15 @@ def test_normalize_llm_settings_adds_default_prompts():
     normalized = normalize_llm_settings({"default_provider": "x"})
 
     assert "prompts" in normalized
+    assert "provider_bindings" in normalized
     assert normalized["default_provider"] == "x"
     for key in DEFAULT_LLM_PROMPTS:
         assert key in normalized["prompts"]
         assert isinstance(normalized["prompts"][key], str)
+    for key in DEFAULT_LLM_PROVIDER_BINDINGS:
+        assert key in normalized["provider_bindings"]
+    for key in DEFAULT_LLM_ASSISTANT_SETTINGS:
+        assert key in normalized
 # [/DEF:test_normalize_llm_settings_adds_default_prompts:Function]
 
 
@@ -59,4 +68,42 @@ def test_render_prompt_replaces_known_placeholders():
 # [/DEF:test_render_prompt_replaces_known_placeholders:Function]
 
 
+# [DEF:test_is_multimodal_model_detects_known_vision_models:Function]
+# @TIER: STANDARD
+# @PURPOSE: Ensure multimodal model detection recognizes common vision-capable model names.
+def test_is_multimodal_model_detects_known_vision_models():
+    assert is_multimodal_model("gpt-4o") is True
+    assert is_multimodal_model("claude-3-5-sonnet") is True
+    assert is_multimodal_model("text-only-model") is False
+# [/DEF:test_is_multimodal_model_detects_known_vision_models:Function]
+
+
+# [DEF:test_resolve_bound_provider_id_prefers_binding_then_default:Function]
+# @TIER: STANDARD
+# @PURPOSE: Verify provider binding resolution priority.
+def test_resolve_bound_provider_id_prefers_binding_then_default():
+    settings = {
+        "default_provider": "default-1",
+        "provider_bindings": {"dashboard_validation": "vision-1"},
+    }
+    assert resolve_bound_provider_id(settings, "dashboard_validation") == "vision-1"
+    assert resolve_bound_provider_id(settings, "documentation") == "default-1"
+# [/DEF:test_resolve_bound_provider_id_prefers_binding_then_default:Function]
+
+
+# [DEF:test_normalize_llm_settings_keeps_assistant_planner_settings:Function]
+# @TIER: STANDARD
+# @PURPOSE: Ensure assistant planner provider/model fields are preserved and normalized.
+def test_normalize_llm_settings_keeps_assistant_planner_settings():
+    normalized = normalize_llm_settings(
+        {
+            "assistant_planner_provider": "provider-a",
+            "assistant_planner_model": "gpt-4.1-mini",
+        }
+    )
+    assert normalized["assistant_planner_provider"] == "provider-a"
+    assert normalized["assistant_planner_model"] == "gpt-4.1-mini"
+# [/DEF:test_normalize_llm_settings_keeps_assistant_planner_settings:Function]
+
+
 # [/DEF:backend.src.services.__tests__.test_llm_prompt_templates:Module]