fix: filter agent files by type-specific model feature instead of vision alone

api/core/agent/base_agent_runner.py

@@ -31,6 +31,7 @@
 from extensions.ext_database import db
 from factories import file_factory
 from graphon.file import file_manager
+from graphon.file.enums import FileType
 from graphon.model_runtime.entities import (
     AssistantPromptMessage,
     LLMUsage,
@@ -53,6 +54,16 @@
 logger = logging.getLogger(__name__)
 _file_access_controller = DatabaseFileAccessController()
 
+# Maps each file type to the model feature required to pass it to the LLM.
+# Mirrors _REQUIRED_MODEL_FEATURE_BY_CONTENT_TYPE in graphon so that non-image
+# files (e.g. documents) are not dropped just because the model lacks vision.
+_REQUIRED_MODEL_FEATURE_BY_FILE_TYPE: dict[FileType, ModelFeature] = {
+    FileType.IMAGE: ModelFeature.VISION,
+    FileType.DOCUMENT: ModelFeature.DOCUMENT,
+    FileType.AUDIO: ModelFeature.AUDIO,
+    FileType.VIDEO: ModelFeature.VIDEO,
+}
+
 
 class BaseAgentRunner(AppRunner):
     def __init__(
@@ -131,7 +142,15 @@ def __init__(
         features = model_schema.features if model_schema and model_schema.features else []
         self.stream_tool_call = ModelFeature.STREAM_TOOL_CALL in features
         self.vision_enabled = ModelFeature.VISION in features
-        self.files = application_generate_entity.files if self.vision_enabled else []
+        # Filter files by the model feature each file type requires, instead of
+        # gating all files on vision alone. Without this, document/audio/video
+        # files are silently dropped when the model does not support vision.
+        file_type_feature = _REQUIRED_MODEL_FEATURE_BY_FILE_TYPE.get
+        self.files = [
+            f
+            for f in (application_generate_entity.files or [])
+            if not hasattr(f, "type") or file_type_feature(f.type, ModelFeature.VISION) in features
+        ]
         self.query: str = ""
         self._current_thoughts: list[PromptMessage] = []
 

api/tests/unit_tests/core/agent/test_base_agent_runner.py

@@ -22,6 +22,7 @@
 from core.app.entities.app_invoke_entities import (
     AgentChatAppGenerateEntity,
     InvokeFrom,
+    ModelConfigWithCredentialsEntity,
 )
 from core.model_manager import ModelInstance
 from core.tools.__base.tool import Tool
@@ -767,3 +768,66 @@ def test_init_uses_real_session_for_count_and_dependencies(
     assert initialized.agent_thought_count == 2
     assert initialized.history_prompt_messages == []
     assert get_dataset_tools.call_args.kwargs["session"] is sqlite_session
+
+
+def test_init_filters_files_by_type_not_vision(sqlite_session: Session, mocker: MockerFixture) -> None:
+    """Regression test for https://github.com/langgenius/dify/issues/39431
+
+    Documents should be kept when the model supports DOCUMENT, even if it
+    does not support VISION. The old code gated all files on VISION alone.
+    """
+    from graphon.file.enums import FileTransferMethod, FileType
+    from graphon.file.models import File
+
+    caller_session = mocker.MagicMock()
+    caller_session.scalar.return_value = 0
+    global_session = mocker.MagicMock()
+    mocker.patch.object(module.db, "session", global_session)
+    mocker.patch.object(BaseAgentRunner, "organize_agent_history", return_value=[])
+    mocker.patch.object(module.DatasetRetrieverTool, "get_dataset_tools", return_value=[])
+
+    llm = mocker.MagicMock()
+    # Model supports DOCUMENT but NOT VISION
+    llm.get_model_schema.return_value = mocker.MagicMock(features=[module.ModelFeature.DOCUMENT])
+    model_instance = mocker.MagicMock(model_type_instance=llm, model="m", credentials="c")
+
+    app_config = mocker.MagicMock()
+    app_config.app_id = "app1"
+    app_config.agent = None
+    app_config.dataset = mocker.MagicMock(dataset_ids=[], retrieve_config={"k": "v"})
+    app_config.additional_features = mocker.MagicMock(show_retrieve_source=False)
+
+    doc_file = File(
+        file_type=FileType.DOCUMENT,
+        transfer_method=FileTransferMethod.LOCAL_FILE,
+        upload_file_id="doc1",
+        url="http://x",
+    )
+    image_file = File(
+        file_type=FileType.IMAGE,
+        transfer_method=FileTransferMethod.LOCAL_FILE,
+        upload_file_id="img1",
+        url="http://y",
+    )
+
+    app_generate = mocker.MagicMock(invoke_from="test", inputs={}, files=[doc_file, image_file])
+
+    runner = BaseAgentRunner(
+        session=sqlite_session,
+        tenant_id="tenant",
+        application_generate_entity=app_generate,
+        conversation=mocker.MagicMock(),
+        app_config=app_config,
+        model_config=ModelConfigWithCredentialsEntity.model_construct(),
+        config=_agent(),
+        queue_manager=mocker.Mock(spec=module.AppQueueManager),
+        message=_message(message_id="msg1"),
+        user_id="user",
+        model_instance=model_instance,
+        execution_driver=WorkflowRunAgg.run,
+    )
+
+    # Document file should be kept; image file should be filtered out
+    assert doc_file in runner.files
+    assert image_file not in runner.files
+    assert len(runner.files) == 1

api/tests/unit_tests/core/agent/test_fc_agent_runner.py

@@ -340,6 +340,9 @@ def test_with_none_query(self, runner: FunctionCallAgentRunner):
         assert len(result) == 1
 
     def test_with_files_uses_image_detail_config(self, runner: FunctionCallAgentRunner, mocker: MockerFixture):
+        from graphon.file.enums import FileTransferMethod, FileType
+        from graphon.file.models import File
+
         file_content = TextPromptMessageContent(data="file-content")
         mock_to_prompt = mocker.patch(
             "core.agent.fc_agent_runner.file_manager.to_prompt_message_content",
@@ -348,13 +351,19 @@ def test_with_files_uses_image_detail_config(self, runner: FunctionCallAgentRunn
 
         image_config = MagicMock(detail=ImagePromptMessageContent.DETAIL.HIGH)
         runner.application_generate_entity.file_upload_config = MagicMock(image_config=image_config)
-        runner.files = ["file1"]
+        test_file = File(
+            file_type=FileType.IMAGE,
+            transfer_method=FileTransferMethod.REMOTE_URL,
+            remote_url="https://example.com/file1.png",
+            filename="file1.png",
+        )
+        runner.files = [test_file]
 
         result = runner._organize_user_query("query", [])
 
         assert len(result) == 1
         assert isinstance(result[0].content, list)
-        mock_to_prompt.assert_called_once_with("file1", image_detail_config=ImagePromptMessageContent.DETAIL.HIGH)
+        mock_to_prompt.assert_called_once_with(test_file, image_detail_config=ImagePromptMessageContent.DETAIL.HIGH)
 
 
 # ==============================