🐛 Bugfix: use agent model for text file analysis tool

backend/agents/create_agent_info.py

@@ -10,7 +10,7 @@
 from nexent.core.agents.agent_context import ContextManagerConfig
 from nexent.memory.memory_service import search_memory_in_levels
 
-from services.file_management_service import get_llm_model, validate_urls_access
+from services.file_management_service import get_llm_model, build_llm_model, validate_urls_access
 from services.vectordatabase_service import (
     ElasticSearchService,
     get_vector_db_core,
@@ -412,6 +412,7 @@
         user_id,
         version_no=version_no,
         tool_params=normalized_tool_params,
+        override_model_id=override_model_id,
     )
 
     # Build system prompt: prioritize segmented fields, fallback to original prompt field if not available
@@ -644,6 +645,7 @@
     user_id,
     version_no: int = 0,
     tool_params: Optional[ToolParamsRequest | Dict[str, Any]] = None,
+    override_model_id: int | None = None,
 ):
     tool_config_list = []
     langchain_tools = await discover_langchain_tools()
@@ -657,6 +659,14 @@
     # but we include it in error messages so callers can identify which agent/tool caused a failure.
     agent_info = search_agent_info_by_agent_id(agent_id=agent_id, tenant_id=tenant_id, version_no=version_no)
     agent_name = agent_info.get("name") if agent_info else None
+
+    # Resolve the model the agent itself runs with, so tools that rely on an LLM
+    # (e.g. AnalyzeTextFileTool) use the agent-configured model rather than the
+    # tenant-wide default from the model management page. Mirrors the main model
+    # resolution in create_agent_config (override takes precedence over the
+    # persisted agent model_id).
+    agent_model_id = override_model_id if override_model_id else (
+        agent_info.get("model_id") if agent_info else None)
     agent_tool_overrides = _get_agent_tool_overrides(normalized_tool_params, agent_name)
 
     tool_keys_seen = set()
@@ -759,8 +769,18 @@
                 "rerank_model": rerank_model,
             }
         elif tool_config.class_name == "AnalyzeTextFileTool":
+            # Prefer the agent-configured LLM; fall back to the tenant default
+            # when the agent has no model configured.
+            agent_model_config = (
+                get_model_by_model_id(agent_model_id, tenant_id=tenant_id)
+                if agent_model_id else None
+            )
+            llm_model = (
+                build_llm_model(agent_model_config)
+                if agent_model_config else get_llm_model(tenant_id=tenant_id)
+            )
             tool_config.metadata = {
-                "llm_model": get_llm_model(tenant_id=tenant_id),
+                "llm_model": llm_model,
                 "storage_client": minio_client,
                 "data_process_service_url": DATA_PROCESS_SERVICE,
                 "validate_url_access": lambda urls: validate_urls_access(urls, user_id)

backend/services/file_management_service.py

@@ -448,24 +448,29 @@ async def list_files_impl(prefix: str, limit: Optional[int] = None):
     return files
 
 
-def get_llm_model(tenant_id: str):
-    # Get the tenant config
-    main_model_config = tenant_config_manager.get_model_config(
-        key=MODEL_CONFIG_MAPPING["llm"], tenant_id=tenant_id)
-    timeout_seconds = main_model_config.get(
-        "timeout_seconds") if main_model_config else None
+def build_llm_model(model_config):
+    """Build an OpenAILongContextModel from a model configuration dict."""
+    timeout_seconds = model_config.get(
+        "timeout_seconds") if model_config else None
     long_text_to_text_model = OpenAILongContextModel(
         observer=MessageObserver(),
-        model_id=get_model_name_from_config(main_model_config),
-        api_base=main_model_config.get("base_url"),
-        api_key=main_model_config.get("api_key"),
-        max_context_tokens=main_model_config.get("max_tokens"),
-        ssl_verify=main_model_config.get("ssl_verify", True),
+        model_id=get_model_name_from_config(model_config),
+        api_base=model_config.get("base_url"),
+        api_key=model_config.get("api_key"),
+        max_context_tokens=model_config.get("max_tokens"),
+        ssl_verify=model_config.get("ssl_verify", True),
         timeout_seconds=timeout_seconds,
     )
     return long_text_to_text_model
 
 
+def get_llm_model(tenant_id: str):
+    # Get the tenant config
+    main_model_config = tenant_config_manager.get_model_config(
+        key=MODEL_CONFIG_MAPPING["llm"], tenant_id=tenant_id)
+    return build_llm_model(main_model_config)
+
+
 async def resolve_preview_file(object_name: str) -> Tuple[str, str, int]:
     """
     Resolve the actual object name, content type, and total size for preview.

test/backend/agents/test_create_agent_info.py

@@ -233,6 +233,7 @@
 sys.modules['services.file_management_service'] = _create_stub_module(
     "services.file_management_service",
     get_llm_model=MagicMock(return_value="stub_llm_model"),
+    build_llm_model=MagicMock(return_value="stub_llm_model_from_config"),
     validate_urls_access=MagicMock(),
 )
 sys.modules['services.tool_configuration_service'] = _create_stub_module(
@@ -887,16 +888,19 @@
             assert callable(mock_tool_instance.metadata["validate_url_access"])
 
     @pytest.mark.asyncio
-    async def test_create_tool_config_list_with_analyze_text_file_tool(self):
-        """Ensure AnalyzeTextFileTool receives text-specific metadata."""
+    async def test_create_tool_config_list_analyze_text_file_uses_agent_model(self):
+        """AnalyzeTextFileTool should use the agent-configured LLM, not the tenant default."""
         mock_tool_instance = MagicMock()
         mock_tool_instance.class_name = "AnalyzeTextFileTool"
         mock_tool_config.return_value = mock_tool_instance
 
         with patch('backend.agents.create_agent_info.discover_langchain_tools', return_value=[]), \
                 patch('backend.agents.create_agent_info.search_tools_for_sub_agent') as mock_search_tools, \
+                patch('backend.agents.create_agent_info.search_agent_info_by_agent_id') as mock_search_agent, \
+                patch('backend.agents.create_agent_info.get_model_by_model_id') as mock_get_model_by_id, \
+                patch('backend.agents.create_agent_info.build_llm_model') as mock_build_llm_model, \
                 patch('backend.agents.create_agent_info.get_llm_model') as mock_get_llm_model, \
-                patch('backend.agents.create_agent_info.minio_client', new_callable=MagicMock) as mock_minio_client:
+                patch('backend.agents.create_agent_info.minio_client', new_callable=MagicMock):
 
             mock_search_tools.return_value = [
                 {
@@ -910,20 +914,120 @@
                     "usage": None
                 }
             ]
-            mock_get_llm_model.return_value = "mock_llm_model"
+            mock_search_agent.return_value = {"name": "agent", "model_id": 42}
+            agent_model_config = {
+                "model_repo": "openai",
+                "model_name": "agent-llm",
+                "base_url": "https://agent.example.com/v1",
+                "api_key": "agent-key",
+                "max_tokens": 8192,
+                "ssl_verify": True,
+                "timeout_seconds": 30,
+            }
+            mock_get_model_by_id.return_value = agent_model_config
+            mock_build_llm_model.return_value = "agent_llm_model"
 
             result = await create_tool_config_list("agent_1", "tenant_1", "user_1")
 
             assert len(result) == 1
             assert result[0] is mock_tool_instance
-            mock_get_llm_model.assert_called_once_with(tenant_id="tenant_1")
-            # Verify metadata includes validate_url_access lambda
-            assert "llm_model" in mock_tool_instance.metadata
+            # Agent model is resolved and built; tenant default is not used.
+            mock_get_model_by_id.assert_called_once_with(42, tenant_id="tenant_1")
+            mock_build_llm_model.assert_called_once_with(agent_model_config)
+            mock_get_llm_model.assert_not_called()
+            assert mock_tool_instance.metadata["llm_model"] == "agent_llm_model"
             assert "storage_client" in mock_tool_instance.metadata
             assert "data_process_service_url" in mock_tool_instance.metadata
-            assert "validate_url_access" in mock_tool_instance.metadata
             assert callable(mock_tool_instance.metadata["validate_url_access"])
 
+    @pytest.mark.asyncio
+    async def test_create_tool_config_list_analyze_text_file_falls_back_to_tenant_model(self):
+        """When the agent has no configured model, fall back to the tenant default LLM."""
+        mock_tool_instance = MagicMock()
+        mock_tool_instance.class_name = "AnalyzeTextFileTool"
+        mock_tool_config.return_value = mock_tool_instance
+
+        with patch('backend.agents.create_agent_info.discover_langchain_tools', return_value=[]), \
+                patch('backend.agents.create_agent_info.search_tools_for_sub_agent') as mock_search_tools, \
+                patch('backend.agents.create_agent_info.search_agent_info_by_agent_id') as mock_search_agent, \
+                patch('backend.agents.create_agent_info.get_model_by_model_id') as mock_get_model_by_id, \
+                patch('backend.agents.create_agent_info.build_llm_model') as mock_build_llm_model, \
+                patch('backend.agents.create_agent_info.get_llm_model') as mock_get_llm_model, \
+                patch('backend.agents.create_agent_info.minio_client', new_callable=MagicMock):
+
+            mock_search_tools.return_value = [
+                {
+                    "class_name": "AnalyzeTextFileTool",
+                    "name": "analyze_text_file",
+                    "description": "Analyze text file tool",
+                    "inputs": "string",
+                    "output_type": "array",
+                    "params": [{"name": "prompt", "default": "describe"}],
+                    "source": "local",
+                    "usage": None
+                }
+            ]
+            mock_search_agent.return_value = {"name": "agent", "model_id": None}
+            mock_get_llm_model.return_value = "tenant_llm_model"
+
+            result = await create_tool_config_list("agent_1", "tenant_1", "user_1")
+
+            assert len(result) == 1
+            mock_get_model_by_id.assert_not_called()
+            mock_build_llm_model.assert_not_called()
+            mock_get_llm_model.assert_called_once_with(tenant_id="tenant_1")
+            assert mock_tool_instance.metadata["llm_model"] == "tenant_llm_model"
+
+    @pytest.mark.asyncio
+    async def test_create_tool_config_list_analyze_text_file_override_model(self):
+        """An explicit override_model_id should win over the persisted agent model."""
+        mock_tool_instance = MagicMock()
+        mock_tool_instance.class_name = "AnalyzeTextFileTool"
+        mock_tool_config.return_value = mock_tool_instance
+
+        with patch('backend.agents.create_agent_info.discover_langchain_tools', return_value=[]), \
+                patch('backend.agents.create_agent_info.search_tools_for_sub_agent') as mock_search_tools, \
+                patch('backend.agents.create_agent_info.search_agent_info_by_agent_id') as mock_search_agent, \
+                patch('backend.agents.create_agent_info.get_model_by_model_id') as mock_get_model_by_id, \
+                patch('backend.agents.create_agent_info.build_llm_model') as mock_build_llm_model, \
+                patch('backend.agents.create_agent_info.get_llm_model') as mock_get_llm_model, \
+                patch('backend.agents.create_agent_info.minio_client', new_callable=MagicMock):
+
+            mock_search_tools.return_value = [
+                {
+                    "class_name": "AnalyzeTextFileTool",
+                    "name": "analyze_text_file",
+                    "description": "Analyze text file tool",
+                    "inputs": "string",
+                    "output_type": "array",
+                    "params": [{"name": "prompt", "default": "describe"}],
+                    "source": "local",
+                    "usage": None
+                }
+            ]
+            mock_search_agent.return_value = {"name": "agent", "model_id": 42}
+            override_model_config = {
+                "model_repo": "openai",
+                "model_name": "override-llm",
+                "base_url": "https://override.example.com/v1",
+                "api_key": "override-key",
+                "max_tokens": 16384,
+                "ssl_verify": False,
+                "timeout_seconds": 45,
+            }
+            mock_get_model_by_id.return_value = override_model_config
+            mock_build_llm_model.return_value = "override_llm_model"
+
+            result = await create_tool_config_list(
+                "agent_1", "tenant_1", "user_1", override_model_id=99)
+
+            assert len(result) == 1
+            # Override id (99) takes precedence over the persisted agent model_id (42).
+            mock_get_model_by_id.assert_called_once_with(99, tenant_id="tenant_1")
+            mock_build_llm_model.assert_called_once_with(override_model_config)
+            mock_get_llm_model.assert_not_called()
+            assert mock_tool_instance.metadata["llm_model"] == "override_llm_model"
+
     @pytest.mark.asyncio
     async def test_create_tool_config_list_with_knowledge_base_tool_metadata(self):
         """
@@ -1667,7 +1771,23 @@
                 "system_prompt": "populated_system_prompt"}
             mock_get_model_by_id.return_value = {"display_name": "test_model"}
 
-            result = await create_agent_config("agent_1", "tenant_1", "user_1", "zh", "test query")
+            result = await create_agent_config(
+                "agent_1",
+                "tenant_1",
+                "user_1",
+                "zh",
+                "test query",
+                override_model_id=456,
+            )
+
+            mock_create_tools.assert_called_once_with(
+                "agent_1",
+                "tenant_1",
+                "user_1",
+                version_no=0,
+                tool_params=ANY,
+                override_model_id=456,
+            )
 
             # Verify that AgentConfig was called correctly
             mock_agent_config.assert_called_once_with(

test/backend/services/test_file_management_service.py

@@ -1494,6 +1494,44 @@ def test_get_llm_model_with_different_tenant_ids(self, mock_tenant_config, mock_
         assert mock_tenant_config.get_model_config.call_args_list[1][1]["tenant_id"] == "tenant2"
 
 
+class TestBuildLlmModel:
+    """Test cases for build_llm_model function"""
+
+    @patch('backend.services.file_management_service.MessageObserver')
+    @patch('backend.services.file_management_service.OpenAILongContextModel')
+    @patch('backend.services.file_management_service.get_model_name_from_config')
+    def test_build_llm_model_from_config(self, mock_get_model_name, mock_openai_model, mock_message_observer):
+        """build_llm_model builds an OpenAILongContextModel from an explicit config dict."""
+        from backend.services.file_management_service import build_llm_model
+
+        mock_config = {
+            "base_url": "http://agent.example.com",
+            "api_key": "agent_api_key",
+            "max_tokens": 8192,
+            "ssl_verify": False,
+            "timeout_seconds": 30,
+        }
+        mock_get_model_name.return_value = "agent-model"
+        mock_observer_instance = Mock()
+        mock_message_observer.return_value = mock_observer_instance
+        mock_model_instance = Mock()
+        mock_openai_model.return_value = mock_model_instance
+
+        result = build_llm_model(mock_config)
+
+        assert result == mock_model_instance
+        mock_get_model_name.assert_called_once_with(mock_config)
+        mock_openai_model.assert_called_once_with(
+            observer=mock_observer_instance,
+            model_id="agent-model",
+            api_base="http://agent.example.com",
+            api_key="agent_api_key",
+            max_context_tokens=8192,
+            ssl_verify=False,
+            timeout_seconds=30,
+        )
+
+
 class TestResolvePreviewFile:
     """Test cases for resolve_preview_file function"""