Merge remote-tracking branch 'origin/main' into feat/model-plugins-implementing

- Implemented tests to verify the display of the compatible-only helper text based on the presence of scope features.
- Updated the Popup component to conditionally render a banner when scope features are applied.
- Added localization for the new helper text in English, Japanese, and Simplified Chinese.

.github/workflows/api-tests.yml

@@ -27,7 +27,7 @@ jobs:
           persist-credentials: false
 
       - name: Setup UV and Python
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: true
           python-version: ${{ matrix.python-version }}

.github/workflows/autofix.yml

@@ -39,7 +39,7 @@ jobs:
         with:
           python-version: "3.11"
 
-      - uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+      - uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
 
       - name: Generate Docker Compose
         if: steps.docker-compose-changes.outputs.any_changed == 'true'

.github/workflows/build-push.yml

@@ -113,7 +113,7 @@ jobs:
             context: "web"
     steps:
       - name: Download digests
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3 # v8.0.0
         with:
           path: /tmp/digests
           pattern: digests-${{ matrix.context }}-*

.github/workflows/db-migration-test.yml

@@ -19,7 +19,7 @@ jobs:
           persist-credentials: false
 
       - name: Setup UV and Python
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: true
           python-version: "3.12"
@@ -69,7 +69,7 @@ jobs:
           persist-credentials: false
 
       - name: Setup UV and Python
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: true
           python-version: "3.12"

.github/workflows/main-ci.yml

@@ -28,7 +28,7 @@ jobs:
       migration-changed: ${{ steps.changes.outputs.migration }}
     steps:
       - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
-      - uses: dorny/paths-filter@fbd0ab8f3e69293af611ebaee6363fc25e6d187d # v4.0.1
+      - uses: dorny/paths-filter@de90cc6fb38fc0963ad72b210f1f284cd68cea36 # v3.0.2
         id: changes
         with:
           filters: |

.github/workflows/pyrefly-diff.yml

@@ -22,7 +22,7 @@ jobs:
           fetch-depth: 0
 
       - name: Setup Python & UV
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: true
 

.github/workflows/style.yml

@@ -33,7 +33,7 @@ jobs:
 
       - name: Setup UV and Python
         if: steps.changed-files.outputs.any_changed == 'true'
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: false
           python-version: "3.12"

.github/workflows/translate-i18n-claude.yml

@@ -120,7 +120,7 @@ jobs:
 
       - name: Run Claude Code for Translation Sync
         if: steps.detect_changes.outputs.CHANGED_FILES != ''
-        uses: anthropics/claude-code-action@cd77b50d2b0808657f8e6774085c8bf54484351c # v1.0.72
+        uses: anthropics/claude-code-action@26ec041249acb0a944c0a47b6c0c13f05dbc5b44 # v1.0.70
         with:
           anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
           github_token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/vdb-tests.yml

@@ -31,7 +31,7 @@ jobs:
           remove_tool_cache: true
 
       - name: Setup UV and Python
-        uses: astral-sh/setup-uv@e06108dd0aef18192324c70427afc47652e63a82 # v7.5.0
+        uses: astral-sh/setup-uv@6ee6290f1cbc4156c0bdd66691b2c144ef8df19a # v7.4.0
         with:
           enable-cache: true
           python-version: ${{ matrix.python-version }}

.github/workflows/web-tests.yml

@@ -26,8 +26,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        shardIndex: [1, 2, 3, 4]
-        shardTotal: [4]
+        shardIndex: [1, 2, 3, 4, 5, 6]
+        shardTotal: [6]
     defaults:
       run:
         shell: bash
@@ -77,7 +77,7 @@ jobs:
         uses: ./.github/actions/setup-web
 
       - name: Download blob reports
-        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3 # v8.0.0
         with:
           path: web/.vitest-reports
           pattern: blob-report-*

.gitignore

@@ -237,6 +237,3 @@ scripts/stress-test/reports/
 # settings
 *.local.json
 *.local.md
-
-# Code Agent Folder
-.qoder/*

api/.env.example

@@ -22,10 +22,10 @@ APP_WEB_URL=http://localhost:3000
 # Files URL
 FILES_URL=http://localhost:5001
 
-# INTERNAL_FILES_URL is used by services running in Docker to reach the API file endpoints.
-# For Docker Desktop (Mac/Windows), use http://host.docker.internal:5001 when the API runs on the host.
-# For Docker Compose on Linux, use http://api:5001 when the API runs inside the Docker network.
-INTERNAL_FILES_URL=http://host.docker.internal:5001
+# INTERNAL_FILES_URL is used for plugin daemon communication within Docker network.
+# Set this to the internal Docker service URL for proper plugin file access.
+# Example: INTERNAL_FILES_URL=http://api:5001
+INTERNAL_FILES_URL=http://127.0.0.1:5001
 
 # TRIGGER URL
 TRIGGER_URL=http://localhost:5001
@@ -180,7 +180,7 @@ CONSOLE_CORS_ALLOW_ORIGINS=http://localhost:3000,*
 COOKIE_DOMAIN=
 
 # Vector database configuration
-# Supported values are `weaviate`, `oceanbase`, `qdrant`, `milvus`, `myscale`, `relyt`, `pgvector`, `pgvecto-rs`, `chroma`, `opensearch`, `oracle`, `tencent`, `elasticsearch`, `elasticsearch-ja`, `analyticdb`, `couchbase`, `vikingdb`,  `opengauss`, `tablestore`,`vastbase`,`tidb`,`tidb_on_qdrant`,`baidu`,`lindorm`,`huawei_cloud`,`upstash`, `matrixone`, `hologres`.
+# Supported values are `weaviate`, `oceanbase`, `qdrant`, `milvus`, `myscale`, `relyt`, `pgvector`, `pgvecto-rs`, `chroma`, `opensearch`, `oracle`, `tencent`, `elasticsearch`, `elasticsearch-ja`, `analyticdb`, `couchbase`, `vikingdb`,  `opengauss`, `tablestore`,`vastbase`,`tidb`,`tidb_on_qdrant`,`baidu`,`lindorm`,`huawei_cloud`,`upstash`, `matrixone`.
 VECTOR_STORE=weaviate
 # Prefix used to create collection name in vector database
 VECTOR_INDEX_NAME_PREFIX=Vector_index
@@ -217,20 +217,6 @@ COUCHBASE_PASSWORD=password
 COUCHBASE_BUCKET_NAME=Embeddings
 COUCHBASE_SCOPE_NAME=_default
 
-# Hologres configuration
-# access_key_id is used as the PG username, access_key_secret is used as the PG password
-HOLOGRES_HOST=
-HOLOGRES_PORT=80
-HOLOGRES_DATABASE=
-HOLOGRES_ACCESS_KEY_ID=
-HOLOGRES_ACCESS_KEY_SECRET=
-HOLOGRES_SCHEMA=public
-HOLOGRES_TOKENIZER=jieba
-HOLOGRES_DISTANCE_METHOD=Cosine
-HOLOGRES_BASE_QUANTIZATION_TYPE=rabitq
-HOLOGRES_MAX_DEGREE=64
-HOLOGRES_EF_CONSTRUCTION=400
-
 # Milvus configuration
 MILVUS_URI=http://127.0.0.1:19530
 MILVUS_TOKEN=

api/.importlinter

@@ -96,6 +96,7 @@ ignore_imports =
     dify_graph.nodes.tool.tool_node -> core.callback_handler.workflow_tool_callback_handler
     dify_graph.nodes.tool.tool_node -> core.tools.tool_engine
     dify_graph.nodes.tool.tool_node -> core.tools.tool_manager
+    dify_graph.nodes.knowledge_retrieval.knowledge_retrieval_node -> core.app.app_config.entities
     dify_graph.nodes.parameter_extractor.parameter_extractor_node -> core.prompt.advanced_prompt_transform
     dify_graph.nodes.parameter_extractor.parameter_extractor_node -> core.prompt.simple_prompt_transform
     dify_graph.nodes.parameter_extractor.parameter_extractor_node -> dify_graph.model_runtime.model_providers.__base.large_language_model
@@ -103,6 +104,7 @@ ignore_imports =
     dify_graph.nodes.parameter_extractor.parameter_extractor_node -> core.model_manager
     dify_graph.nodes.question_classifier.question_classifier_node -> core.model_manager
     dify_graph.nodes.tool.tool_node -> core.tools.utils.message_transformer
+    dify_graph.nodes.llm.node -> core.helper.code_executor
     dify_graph.nodes.llm.node -> core.llm_generator.output_parser.errors
     dify_graph.nodes.llm.node -> core.llm_generator.output_parser.structured_output
     dify_graph.nodes.llm.node -> core.model_manager
@@ -114,6 +116,7 @@ ignore_imports =
     dify_graph.nodes.parameter_extractor.parameter_extractor_node -> core.prompt.utils.prompt_message_util
     dify_graph.nodes.question_classifier.entities -> core.prompt.entities.advanced_prompt_entities
     dify_graph.nodes.question_classifier.question_classifier_node -> core.prompt.utils.prompt_message_util
+    dify_graph.nodes.knowledge_index.entities -> core.rag.retrieval.retrieval_methods
     dify_graph.nodes.llm.node -> models.dataset
     dify_graph.nodes.llm.file_saver -> core.tools.signature
     dify_graph.nodes.llm.file_saver -> core.tools.tool_file_manager

api/Dockerfile

@@ -97,7 +97,7 @@ ENV PATH="${VIRTUAL_ENV}/bin:${PATH}"
 
 # Download nltk data
 RUN mkdir -p /usr/local/share/nltk_data \
-    && NLTK_DATA=/usr/local/share/nltk_data python -c "import nltk; nltk.download('punkt'); nltk.download('averaged_perceptron_tagger'); nltk.download('stopwords')" \
+    && NLTK_DATA=/usr/local/share/nltk_data python -c "import nltk; from unstructured.nlp.tokenize import download_nltk_packages; nltk.download('punkt'); nltk.download('averaged_perceptron_tagger'); nltk.download('stopwords'); download_nltk_packages()" \
     && chmod -R 755 /usr/local/share/nltk_data
 
 ENV TIKTOKEN_CACHE_DIR=/app/api/.tiktoken_cache

api/app_factory.py

@@ -1,45 +1,16 @@
 import logging
 import time
 
-from flask import request
 from opentelemetry.trace import get_current_span
 from opentelemetry.trace.span import INVALID_SPAN_ID, INVALID_TRACE_ID
 
 from configs import dify_config
 from contexts.wrapper import RecyclableContextVar
-from controllers.console.error import UnauthorizedAndForceLogout
 from core.logging.context import init_request_context
 from dify_app import DifyApp
-from services.enterprise.enterprise_service import EnterpriseService
-from services.feature_service import LicenseStatus
 
 logger = logging.getLogger(__name__)
 
-# Console bootstrap APIs exempt from license check.
-# Defined at module level to avoid per-request tuple construction.
-# - system-features: license status for expiry UI (GlobalPublicStoreProvider)
-# - setup: install/setup status check (AppInitializer)
-# - init: init password validation for fresh install (InitPasswordPopup)
-# - login: auto-login after setup completion (InstallForm)
-# - features: billing/plan features (ProviderContextProvider)
-# - account/profile: login check + user profile (AppContextProvider, useIsLogin)
-# - workspaces/current: workspace + model providers (AppContextProvider)
-# - version: version check (AppContextProvider)
-# - activate/check: invitation link validation (signin page)
-# Without these exemptions, the signin page triggers location.reload()
-# on unauthorized_and_force_logout, causing an infinite loop.
-_CONSOLE_EXEMPT_PREFIXES = (
-    "/console/api/system-features",
-    "/console/api/setup",
-    "/console/api/init",
-    "/console/api/login",
-    "/console/api/features",
-    "/console/api/account/profile",
-    "/console/api/workspaces/current",
-    "/console/api/version",
-    "/console/api/activate/check",
-)
-
 
 # ----------------------------
 # Application Factory Function
@@ -60,39 +31,6 @@ def create_flask_app_with_configs() -> DifyApp:
         init_request_context()
         RecyclableContextVar.increment_thread_recycles()
 
-        # Enterprise license validation for API endpoints (both console and webapp)
-        # When license expires, block all API access except bootstrap endpoints needed
-        # for the frontend to load the license expiration page without infinite reloads.
-        if dify_config.ENTERPRISE_ENABLED:
-            is_console_api = request.path.startswith("/console/api/")
-            is_webapp_api = request.path.startswith("/api/")
-
-            if is_console_api or is_webapp_api:
-                if is_console_api:
-                    is_exempt = any(request.path.startswith(p) for p in _CONSOLE_EXEMPT_PREFIXES)
-                else:  # webapp API
-                    is_exempt = request.path.startswith("/api/system-features")
-
-                if not is_exempt:
-                    try:
-                        # Check license status (cached — see EnterpriseService for TTL details)
-                        license_status = EnterpriseService.get_cached_license_status()
-                        if license_status in (LicenseStatus.INACTIVE, LicenseStatus.EXPIRED, LicenseStatus.LOST):
-                            raise UnauthorizedAndForceLogout(
-                                f"Enterprise license is {license_status}. Please contact your administrator."
-                            )
-                        if license_status is None:
-                            raise UnauthorizedAndForceLogout(
-                                "Unable to verify enterprise license. Please contact your administrator."
-                            )
-                    except UnauthorizedAndForceLogout:
-                        raise
-                    except Exception:
-                        logger.exception("Failed to check enterprise license status")
-                        raise UnauthorizedAndForceLogout(
-                            "Unable to verify enterprise license. Please contact your administrator."
-                        )
-
     # add after request hook for injecting trace headers from OpenTelemetry span context
     # Only adds headers when OTEL is enabled and has valid context
     @dify_app.after_request

api/commands/vector.py

@@ -160,7 +160,6 @@ def migrate_knowledge_vector_database():
     }
     lower_collection_vector_types = {
         VectorType.ANALYTICDB,
-        VectorType.HOLOGRES,
         VectorType.CHROMA,
         VectorType.MYSCALE,
         VectorType.PGVECTO_RS,

api/configs/middleware/__init__.py

@@ -26,7 +26,6 @@ from .vdb.chroma_config import ChromaConfig
 from .vdb.clickzetta_config import ClickzettaConfig
 from .vdb.couchbase_config import CouchbaseConfig
 from .vdb.elasticsearch_config import ElasticsearchConfig
-from .vdb.hologres_config import HologresConfig
 from .vdb.huawei_cloud_config import HuaweiCloudConfig
 from .vdb.iris_config import IrisVectorConfig
 from .vdb.lindorm_config import LindormConfig
@@ -348,7 +347,6 @@ class MiddlewareConfig(
     AnalyticdbConfig,
     ChromaConfig,
     ClickzettaConfig,
-    HologresConfig,
     HuaweiCloudConfig,
     IrisVectorConfig,
     MilvusConfig,

api/configs/middleware/vdb/hologres_config.py

@@ -1,68 +0,0 @@
-from holo_search_sdk.types import BaseQuantizationType, DistanceType, TokenizerType
-from pydantic import Field
-from pydantic_settings import BaseSettings
-
-
-class HologresConfig(BaseSettings):
-    """
-    Configuration settings for Hologres vector database.
-
-    Hologres is compatible with PostgreSQL protocol.
-    access_key_id is used as the PostgreSQL username,
-    and access_key_secret is used as the PostgreSQL password.
-    """
-
-    HOLOGRES_HOST: str | None = Field(
-        description="Hostname or IP address of the Hologres instance.",
-        default=None,
-    )
-
-    HOLOGRES_PORT: int = Field(
-        description="Port number for connecting to the Hologres instance.",
-        default=80,
-    )
-
-    HOLOGRES_DATABASE: str | None = Field(
-        description="Name of the Hologres database to connect to.",
-        default=None,
-    )
-
-    HOLOGRES_ACCESS_KEY_ID: str | None = Field(
-        description="Alibaba Cloud AccessKey ID, also used as the PostgreSQL username.",
-        default=None,
-    )
-
-    HOLOGRES_ACCESS_KEY_SECRET: str | None = Field(
-        description="Alibaba Cloud AccessKey Secret, also used as the PostgreSQL password.",
-        default=None,
-    )
-
-    HOLOGRES_SCHEMA: str = Field(
-        description="Schema name in the Hologres database.",
-        default="public",
-    )
-
-    HOLOGRES_TOKENIZER: TokenizerType = Field(
-        description="Tokenizer for full-text search index (e.g., 'jieba', 'ik', 'standard', 'simple').",
-        default="jieba",
-    )
-
-    HOLOGRES_DISTANCE_METHOD: DistanceType = Field(
-        description="Distance method for vector index (e.g., 'Cosine', 'Euclidean', 'InnerProduct').",
-        default="Cosine",
-    )
-
-    HOLOGRES_BASE_QUANTIZATION_TYPE: BaseQuantizationType = Field(
-        description="Base quantization type for vector index (e.g., 'rabitq', 'sq8', 'fp16', 'fp32').",
-        default="rabitq",
-    )
-
-    HOLOGRES_MAX_DEGREE: int = Field(
-        description="Max degree (M) parameter for HNSW vector index.",
-        default=64,
-    )
-
-    HOLOGRES_EF_CONSTRUCTION: int = Field(
-        description="ef_construction parameter for HNSW vector index.",
-        default=400,
-    )

api/controllers/console/app/app.py

@@ -25,8 +25,7 @@ from controllers.console.wraps import (
 )
 from core.ops.ops_trace_manager import OpsTraceManager
 from core.rag.retrieval.retrieval_methods import RetrievalMethod
-from core.trigger.constants import TRIGGER_NODE_TYPES
-from dify_graph.enums import WorkflowExecutionStatus
+from dify_graph.enums import NodeType, WorkflowExecutionStatus
 from dify_graph.file import helpers as file_helpers
 from extensions.ext_database import db
 from libs.login import current_account_with_tenant, login_required
@@ -509,7 +508,11 @@ class AppListApi(Resource):
                 .scalars()
                 .all()
             )
-            trigger_node_types = TRIGGER_NODE_TYPES
+            trigger_node_types = {
+                NodeType.TRIGGER_WEBHOOK,
+                NodeType.TRIGGER_SCHEDULE,
+                NodeType.TRIGGER_PLUGIN,
+            }
             for workflow in draft_workflows:
                 node_id = None
                 try:

api/controllers/console/app/workflow.py

@@ -22,7 +22,6 @@ from core.app.apps.workflow.app_generator import SKIP_PREPARE_USER_INPUTS_KEY
 from core.app.entities.app_invoke_entities import InvokeFrom
 from core.helper.trace_id_helper import get_external_trace_id
 from core.plugin.impl.exc import PluginInvokeError
-from core.trigger.constants import TRIGGER_SCHEDULE_NODE_TYPE
 from core.trigger.debug.event_selectors import (
     TriggerDebugEvent,
     TriggerDebugEventPoller,
@@ -1210,7 +1209,7 @@ class DraftWorkflowTriggerNodeApi(Resource):
         node_type: NodeType = draft_workflow.get_node_type_from_node_config(node_config)
         event: TriggerDebugEvent | None = None
         # for schedule trigger, when run single node, just execute directly
-        if node_type == TRIGGER_SCHEDULE_NODE_TYPE:
+        if node_type == NodeType.TRIGGER_SCHEDULE:
             event = TriggerDebugEvent(
                 workflow_args={},
                 node_id=node_id,

api/controllers/console/datasets/datasets.py

@@ -263,7 +263,6 @@ def _get_retrieval_methods_by_vector_type(vector_type: str | None, is_mock: bool
         VectorType.BAIDU,
         VectorType.ALIBABACLOUD_MYSQL,
         VectorType.IRIS,
-        VectorType.HOLOGRES,
     }
 
     semantic_methods = {"retrieval_method": [RetrievalMethod.SEMANTIC_SEARCH.value]}

api/controllers/service_api/wraps.py

@@ -3,7 +3,7 @@ import time
 from collections.abc import Callable
 from enum import StrEnum, auto
 from functools import wraps
-from typing import Concatenate, ParamSpec, TypeVar, cast, overload
+from typing import Concatenate, ParamSpec, TypeVar, cast
 
 from flask import current_app, request
 from flask_login import user_logged_in
@@ -44,22 +44,10 @@ class FetchUserArg(BaseModel):
     required: bool = False
 
 
-@overload
-def validate_app_token(view: Callable[P, R]) -> Callable[P, R]: ...
-
-
-@overload
-def validate_app_token(
-    view: None = None, *, fetch_user_arg: FetchUserArg | None = None
-) -> Callable[[Callable[P, R]], Callable[P, R]]: ...
-
-
-def validate_app_token(
-    view: Callable[P, R] | None = None, *, fetch_user_arg: FetchUserArg | None = None
-) -> Callable[P, R] | Callable[[Callable[P, R]], Callable[P, R]]:
-    def decorator(view_func: Callable[P, R]) -> Callable[P, R]:
+def validate_app_token(view: Callable[P, R] | None = None, *, fetch_user_arg: FetchUserArg | None = None):
+    def decorator(view_func: Callable[P, R]):
         @wraps(view_func)
-        def decorated_view(*args: P.args, **kwargs: P.kwargs) -> R:
+        def decorated_view(*args: P.args, **kwargs: P.kwargs):
             api_token = validate_and_get_api_token("app")
 
             app_model = db.session.query(App).where(App.id == api_token.app_id).first()
@@ -225,20 +213,10 @@ def cloud_edition_billing_rate_limit_check(resource: str, api_token_type: str):
     return interceptor
 
 
-@overload
-def validate_dataset_token(view: Callable[Concatenate[T, P], R]) -> Callable[P, R]: ...
-
-
-@overload
-def validate_dataset_token(view: None = None) -> Callable[[Callable[Concatenate[T, P], R]], Callable[P, R]]: ...
-
-
-def validate_dataset_token(
-    view: Callable[Concatenate[T, P], R] | None = None,
-) -> Callable[P, R] | Callable[[Callable[Concatenate[T, P], R]], Callable[P, R]]:
-    def decorator(view_func: Callable[Concatenate[T, P], R]) -> Callable[P, R]:
-        @wraps(view_func)
-        def decorated(*args: P.args, **kwargs: P.kwargs) -> R:
+def validate_dataset_token(view: Callable[Concatenate[T, P], R] | None = None):
+    def decorator(view: Callable[Concatenate[T, P], R]):
+        @wraps(view)
+        def decorated(*args: P.args, **kwargs: P.kwargs):
             api_token = validate_and_get_api_token("dataset")
 
             # get url path dataset_id from positional args or kwargs
@@ -309,7 +287,7 @@ def validate_dataset_token(
                     raise Unauthorized("Tenant owner account does not exist.")
             else:
                 raise Unauthorized("Tenant does not exist.")
-            return view_func(api_token.tenant_id, *args, **kwargs)  # type: ignore[arg-type]
+            return view(api_token.tenant_id, *args, **kwargs)
 
         return decorated
 

api/core/app/apps/advanced_chat/generate_task_pipeline.py

@@ -69,7 +69,7 @@ from dify_graph.entities.pause_reason import HumanInputRequired
 from dify_graph.enums import WorkflowExecutionStatus
 from dify_graph.model_runtime.entities.llm_entities import LLMUsage
 from dify_graph.model_runtime.utils.encoders import jsonable_encoder
-from dify_graph.nodes import BuiltinNodeTypes
+from dify_graph.nodes import NodeType
 from dify_graph.repositories.draft_variable_repository import DraftVariableSaverFactory
 from dify_graph.runtime import GraphRuntimeState
 from dify_graph.system_variable import SystemVariable
@@ -357,7 +357,7 @@ class AdvancedChatAppGenerateTaskPipeline(GraphRuntimeStateSupport):
     ) -> Generator[StreamResponse, None, None]:
         """Handle node succeeded events."""
         # Record files if it's an answer node or end node
-        if event.node_type in [BuiltinNodeTypes.ANSWER, BuiltinNodeTypes.END, BuiltinNodeTypes.LLM]:
+        if event.node_type in [NodeType.ANSWER, NodeType.END, NodeType.LLM]:
             self._recorded_files.extend(
                 self._workflow_response_converter.fetch_files_from_node_outputs(event.outputs or {})
             )

api/core/app/apps/common/workflow_response_converter.py

@@ -48,13 +48,12 @@ from core.app.entities.task_entities import (
 from core.plugin.impl.datasource import PluginDatasourceManager
 from core.tools.entities.tool_entities import ToolProviderType
 from core.tools.tool_manager import ToolManager
-from core.trigger.constants import TRIGGER_PLUGIN_NODE_TYPE
 from core.trigger.trigger_manager import TriggerManager
 from core.workflow.workflow_entry import WorkflowEntry
 from dify_graph.entities.pause_reason import HumanInputRequired
 from dify_graph.entities.workflow_start_reason import WorkflowStartReason
 from dify_graph.enums import (
-    BuiltinNodeTypes,
+    NodeType,
     SystemVariableKey,
     WorkflowExecutionStatus,
     WorkflowNodeExecutionMetadataKey,
@@ -443,7 +442,7 @@ class WorkflowResponseConverter:
         event: QueueNodeStartedEvent,
         task_id: str,
     ) -> NodeStartStreamResponse | None:
-        if event.node_type in {BuiltinNodeTypes.ITERATION, BuiltinNodeTypes.LOOP}:
+        if event.node_type in {NodeType.ITERATION, NodeType.LOOP}:
             return None
         run_id = self._ensure_workflow_run_id()
         snapshot = self._store_snapshot(event)
@@ -465,13 +464,13 @@ class WorkflowResponseConverter:
         )
 
         try:
-            if event.node_type == BuiltinNodeTypes.TOOL:
+            if event.node_type == NodeType.TOOL:
                 response.data.extras["icon"] = ToolManager.get_tool_icon(
                     tenant_id=self._application_generate_entity.app_config.tenant_id,
                     provider_type=ToolProviderType(event.provider_type),
                     provider_id=event.provider_id,
                 )
-            elif event.node_type == BuiltinNodeTypes.DATASOURCE:
+            elif event.node_type == NodeType.DATASOURCE:
                 manager = PluginDatasourceManager()
                 provider_entity = manager.fetch_datasource_provider(
                     self._application_generate_entity.app_config.tenant_id,
@@ -480,7 +479,7 @@ class WorkflowResponseConverter:
                 response.data.extras["icon"] = provider_entity.declaration.identity.generate_datasource_icon_url(
                     self._application_generate_entity.app_config.tenant_id
                 )
-            elif event.node_type == TRIGGER_PLUGIN_NODE_TYPE:
+            elif event.node_type == NodeType.TRIGGER_PLUGIN:
                 response.data.extras["icon"] = TriggerManager.get_trigger_plugin_icon(
                     self._application_generate_entity.app_config.tenant_id,
                     event.provider_id,
@@ -497,7 +496,7 @@ class WorkflowResponseConverter:
         event: QueueNodeSucceededEvent | QueueNodeFailedEvent | QueueNodeExceptionEvent,
         task_id: str,
     ) -> NodeFinishStreamResponse | None:
-        if event.node_type in {BuiltinNodeTypes.ITERATION, BuiltinNodeTypes.LOOP}:
+        if event.node_type in {NodeType.ITERATION, NodeType.LOOP}:
             return None
         run_id = self._ensure_workflow_run_id()
         snapshot = self._pop_snapshot(event.node_execution_id)
@@ -555,7 +554,7 @@ class WorkflowResponseConverter:
         event: QueueNodeRetryEvent,
         task_id: str,
     ) -> NodeRetryStreamResponse | None:
-        if event.node_type in {BuiltinNodeTypes.ITERATION, BuiltinNodeTypes.LOOP}:
+        if event.node_type in {NodeType.ITERATION, NodeType.LOOP}:
             return None
         run_id = self._ensure_workflow_run_id()
 
@@ -613,7 +612,7 @@ class WorkflowResponseConverter:
             data=IterationNodeStartStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 created_at=int(time.time()),
                 extras={},
@@ -636,7 +635,7 @@ class WorkflowResponseConverter:
             data=IterationNodeNextStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 index=event.index,
                 created_at=int(time.time()),
@@ -663,7 +662,7 @@ class WorkflowResponseConverter:
             data=IterationNodeCompletedStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 outputs=new_outputs,
                 outputs_truncated=outputs_truncated,
@@ -693,7 +692,7 @@ class WorkflowResponseConverter:
             data=LoopNodeStartStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 created_at=int(time.time()),
                 extras={},
@@ -716,7 +715,7 @@ class WorkflowResponseConverter:
             data=LoopNodeNextStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 index=event.index,
                 # The `pre_loop_output` field is not utilized by the frontend.
@@ -745,7 +744,7 @@ class WorkflowResponseConverter:
             data=LoopNodeCompletedStreamResponse.Data(
                 id=event.node_id,
                 node_id=event.node_id,
-                node_type=event.node_type,
+                node_type=event.node_type.value,
                 title=event.node_title,
                 outputs=new_outputs,
                 outputs_truncated=outputs_truncated,

api/core/app/apps/pipeline/pipeline_runner.py

@@ -12,7 +12,7 @@ from core.app.entities.app_invoke_entities import (
     build_dify_run_context,
 )
 from core.app.workflow.layers.persistence import PersistenceWorkflowInfo, WorkflowPersistenceLayer
-from core.workflow.node_factory import DifyNodeFactory, get_default_root_node_id
+from core.workflow.node_factory import DifyNodeFactory
 from core.workflow.workflow_entry import WorkflowEntry
 from dify_graph.entities.graph_init_params import GraphInitParams
 from dify_graph.enums import WorkflowType
@@ -274,8 +274,6 @@ class PipelineRunner(WorkflowBasedAppRunner):
             graph_init_params=graph_init_params,
             graph_runtime_state=graph_runtime_state,
         )
-        if start_node_id is None:
-            start_node_id = get_default_root_node_id(graph_config)
         graph = Graph.init(graph_config=graph_config, node_factory=node_factory, root_node_id=start_node_id)
 
         if not graph:

api/core/app/apps/workflow_app_runner.py

@@ -32,8 +32,8 @@ from core.app.entities.queue_entities import (
     QueueWorkflowStartedEvent,
     QueueWorkflowSucceededEvent,
 )
-from core.rag.entities.citation_metadata import RetrievalSourceMetadata
-from core.workflow.node_factory import DifyNodeFactory, get_default_root_node_id, resolve_workflow_node_class
+from core.workflow.node_factory import DifyNodeFactory
+from core.workflow.node_resolution import resolve_workflow_node_class
 from core.workflow.workflow_entry import WorkflowEntry
 from dify_graph.entities import GraphInitParams
 from dify_graph.entities.graph_config import NodeConfigDictAdapter
@@ -140,9 +140,6 @@ class WorkflowBasedAppRunner:
             graph_runtime_state=graph_runtime_state,
         )
 
-        if root_node_id is None:
-            root_node_id = get_default_root_node_id(graph_config)
-
         # init graph
         graph = Graph.init(graph_config=graph_config, node_factory=node_factory, root_node_id=root_node_id)
 
@@ -508,9 +505,7 @@ class WorkflowBasedAppRunner:
         elif isinstance(event, NodeRunRetrieverResourceEvent):
             self._publish_event(
                 QueueRetrieverResourcesEvent(
-                    retriever_resources=[
-                        RetrievalSourceMetadata.model_validate(resource) for resource in event.retriever_resources
-                    ],
+                    retriever_resources=event.retriever_resources,
                     in_iteration_id=event.in_iteration_id,
                     in_loop_id=event.in_loop_id,
                 )

api/core/app/entities/queue_entities.py

@@ -9,8 +9,9 @@ from core.app.entities.agent_strategy import AgentStrategyInfo
 from core.rag.entities.citation_metadata import RetrievalSourceMetadata
 from dify_graph.entities.pause_reason import PauseReason
 from dify_graph.entities.workflow_start_reason import WorkflowStartReason
-from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import WorkflowNodeExecutionMetadataKey
 from dify_graph.model_runtime.entities.llm_entities import LLMResult, LLMResultChunk
+from dify_graph.nodes import NodeType
 
 
 class QueueEvent(StrEnum):

api/core/app/layers/conversation_variable_persist_layer.py

@@ -2,7 +2,7 @@ import logging
 
 from dify_graph.constants import CONVERSATION_VARIABLE_NODE_ID
 from dify_graph.conversation_variable_updater import ConversationVariableUpdater
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from dify_graph.graph_engine.layers.base import GraphEngineLayer
 from dify_graph.graph_events import GraphEngineEvent, NodeRunSucceededEvent
 from dify_graph.nodes.variable_assigner.common import helpers as common_helpers
@@ -22,7 +22,7 @@ class ConversationVariablePersistenceLayer(GraphEngineLayer):
     def on_event(self, event: GraphEngineEvent) -> None:
         if not isinstance(event, NodeRunSucceededEvent):
             return
-        if event.node_type != BuiltinNodeTypes.VARIABLE_ASSIGNER:
+        if event.node_type != NodeType.VARIABLE_ASSIGNER:
             return
         if self.graph_runtime_state is None:
             return

api/core/app/workflow/layers/llm_quota.py

@@ -12,7 +12,7 @@ from typing_extensions import override
 from core.app.llm import deduct_llm_quota, ensure_llm_quota_available
 from core.errors.error import QuotaExceededError
 from core.model_manager import ModelInstance
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from dify_graph.graph_engine.entities.commands import AbortCommand, CommandType
 from dify_graph.graph_engine.layers.base import GraphEngineLayer
 from dify_graph.graph_events import GraphEngineEvent, GraphNodeEventBase
@@ -113,11 +113,11 @@ class LLMQuotaLayer(GraphEngineLayer):
     def _extract_model_instance(node: Node) -> ModelInstance | None:
         try:
             match node.node_type:
-                case BuiltinNodeTypes.LLM:
+                case NodeType.LLM:
                     return cast("LLMNode", node).model_instance
-                case BuiltinNodeTypes.PARAMETER_EXTRACTOR:
+                case NodeType.PARAMETER_EXTRACTOR:
                     return cast("ParameterExtractorNode", node).model_instance
-                case BuiltinNodeTypes.QUESTION_CLASSIFIER:
+                case NodeType.QUESTION_CLASSIFIER:
                     return cast("QuestionClassifierNode", node).model_instance
                 case _:
                     return None

api/core/app/workflow/layers/observability.py

@@ -16,7 +16,7 @@ from opentelemetry.trace import Span, SpanKind, Tracer, get_tracer, set_span_in_
 from typing_extensions import override
 
 from configs import dify_config
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.graph_engine.layers.base import GraphEngineLayer
 from dify_graph.graph_events import GraphNodeEventBase
 from dify_graph.nodes.base.node import Node
@@ -74,13 +74,16 @@ class ObservabilityLayer(GraphEngineLayer):
     def _build_parser_registry(self) -> None:
         """Initialize parser registry for node types."""
         self._parsers = {
-            BuiltinNodeTypes.TOOL: ToolNodeOTelParser(),
-            BuiltinNodeTypes.LLM: LLMNodeOTelParser(),
-            BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL: RetrievalNodeOTelParser(),
+            NodeType.TOOL: ToolNodeOTelParser(),
+            NodeType.LLM: LLMNodeOTelParser(),
+            NodeType.KNOWLEDGE_RETRIEVAL: RetrievalNodeOTelParser(),
         }
 
     def _get_parser(self, node: Node) -> NodeOTelParser:
-        return self._parsers.get(node.node_type, self._default_parser)
+        node_type = getattr(node, "node_type", None)
+        if isinstance(node_type, NodeType):
+            return self._parsers.get(node_type, self._default_parser)
+        return self._default_parser
 
     @override
     def on_graph_start(self) -> None:

api/core/datasource/datasource_manager.py

@@ -24,12 +24,12 @@ from core.datasource.utils.message_transformer import DatasourceFileMessageTrans
 from core.datasource.website_crawl.website_crawl_provider import WebsiteCrawlDatasourcePluginProviderController
 from core.db.session_factory import session_factory
 from core.plugin.impl.datasource import PluginDatasourceManager
-from core.workflow.nodes.datasource.entities import DatasourceParameter, OnlineDriveDownloadFileParam
 from dify_graph.entities.workflow_node_execution import WorkflowNodeExecutionStatus
 from dify_graph.enums import WorkflowNodeExecutionMetadataKey
 from dify_graph.file import File
 from dify_graph.file.enums import FileTransferMethod, FileType
 from dify_graph.node_events import NodeRunResult, StreamChunkEvent, StreamCompletedEvent
+from dify_graph.repositories.datasource_manager_protocol import DatasourceParameter, OnlineDriveDownloadFileParam
 from factories import file_factory
 from models.model import UploadFile
 from models.tools import ToolFile

api/core/ops/aliyun_trace/aliyun_trace.py

@@ -58,7 +58,7 @@ from core.ops.entities.trace_entity import (
 )
 from core.repositories import DifyCoreRepositoryFactory
 from dify_graph.entities import WorkflowNodeExecution
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey
 from extensions.ext_database import db
 from models import WorkflowNodeExecutionTriggeredFrom
 
@@ -302,11 +302,11 @@ class AliyunDataTrace(BaseTraceInstance):
         self, node_execution: WorkflowNodeExecution, trace_info: WorkflowTraceInfo, trace_metadata: TraceMetadata
     ):
         try:
-            if node_execution.node_type == BuiltinNodeTypes.LLM:
+            if node_execution.node_type == NodeType.LLM:
                 node_span = self.build_workflow_llm_span(trace_info, node_execution, trace_metadata)
-            elif node_execution.node_type == BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL:
+            elif node_execution.node_type == NodeType.KNOWLEDGE_RETRIEVAL:
                 node_span = self.build_workflow_retrieval_span(trace_info, node_execution, trace_metadata)
-            elif node_execution.node_type == BuiltinNodeTypes.TOOL:
+            elif node_execution.node_type == NodeType.TOOL:
                 node_span = self.build_workflow_tool_span(trace_info, node_execution, trace_metadata)
             else:
                 node_span = self.build_workflow_task_span(trace_info, node_execution, trace_metadata)

api/core/ops/arize_phoenix_trace/arize_phoenix_trace.py

@@ -155,8 +155,8 @@ def wrap_span_metadata(metadata, **kwargs):
     return metadata
 
 
-# Mapping from built-in node type strings to OpenInference span kinds.
-# Node types not listed here default to CHAIN.
+# Mapping from NodeType string values to OpenInference span kinds.
+# NodeType values not listed here default to CHAIN.
 _NODE_TYPE_TO_SPAN_KIND: dict[str, OpenInferenceSpanKindValues] = {
     "llm": OpenInferenceSpanKindValues.LLM,
     "knowledge-retrieval": OpenInferenceSpanKindValues.RETRIEVER,
@@ -168,7 +168,7 @@ _NODE_TYPE_TO_SPAN_KIND: dict[str, OpenInferenceSpanKindValues] = {
 def _get_node_span_kind(node_type: str) -> OpenInferenceSpanKindValues:
     """Return the OpenInference span kind for a given workflow node type.
 
-    Covers every built-in node type string. Nodes that do not have a
+    Covers every ``NodeType`` enum value.  Nodes that do not have a
     specialised span kind (e.g. ``start``, ``end``, ``if-else``,
     ``code``, ``loop``, ``iteration``, etc.) are mapped to ``CHAIN``.
     """

api/core/ops/langfuse_trace/langfuse_trace.py

@@ -28,7 +28,7 @@ from core.ops.langfuse_trace.entities.langfuse_trace_entity import (
 )
 from core.ops.utils import filter_none_values
 from core.repositories import DifyCoreRepositoryFactory
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from extensions.ext_database import db
 from models import EndUser, WorkflowNodeExecutionTriggeredFrom
 from models.enums import MessageStatus
@@ -141,7 +141,7 @@ class LangFuseDataTrace(BaseTraceInstance):
             node_name = node_execution.title
             node_type = node_execution.node_type
             status = node_execution.status
-            if node_type == BuiltinNodeTypes.LLM:
+            if node_type == NodeType.LLM:
                 inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
             else:
                 inputs = node_execution.inputs or {}

api/core/ops/langsmith_trace/langsmith_trace.py

@@ -28,7 +28,7 @@ from core.ops.langsmith_trace.entities.langsmith_trace_entity import (
 )
 from core.ops.utils import filter_none_values, generate_dotted_order
 from core.repositories import DifyCoreRepositoryFactory
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey
 from extensions.ext_database import db
 from models import EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
 
@@ -163,7 +163,7 @@ class LangSmithDataTrace(BaseTraceInstance):
             node_name = node_execution.title
             node_type = node_execution.node_type
             status = node_execution.status
-            if node_type == BuiltinNodeTypes.LLM:
+            if node_type == NodeType.LLM:
                 inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
             else:
                 inputs = node_execution.inputs or {}
@@ -197,7 +197,7 @@ class LangSmithDataTrace(BaseTraceInstance):
                         "ls_model_name": process_data.get("model_name", ""),
                     }
                 )
-            elif node_type == BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL:
+            elif node_type == NodeType.KNOWLEDGE_RETRIEVAL:
                 run_type = LangSmithRunType.retriever
             else:
                 run_type = LangSmithRunType.tool

api/core/ops/mlflow_trace/mlflow_trace.py

@@ -23,7 +23,7 @@ from core.ops.entities.trace_entity import (
     TraceTaskName,
     WorkflowTraceInfo,
 )
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from extensions.ext_database import db
 from models import EndUser
 from models.workflow import WorkflowNodeExecutionModel
@@ -145,10 +145,10 @@ class MLflowDataTrace(BaseTraceInstance):
                     "app_name": node.title,
                 }
 
-                if node.node_type in (BuiltinNodeTypes.LLM, BuiltinNodeTypes.QUESTION_CLASSIFIER):
+                if node.node_type in (NodeType.LLM, NodeType.QUESTION_CLASSIFIER):
                     inputs, llm_attributes = self._parse_llm_inputs_and_attributes(node)
                     attributes.update(llm_attributes)
-                elif node.node_type == BuiltinNodeTypes.HTTP_REQUEST:
+                elif node.node_type == NodeType.HTTP_REQUEST:
                     inputs = node.process_data  # contains request URL
 
                 if not inputs:
@@ -180,9 +180,9 @@ class MLflowDataTrace(BaseTraceInstance):
                 # End node span
                 finished_at = node.created_at + timedelta(seconds=node.elapsed_time)
                 outputs = json.loads(node.outputs) if node.outputs else {}
-                if node.node_type == BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL:
+                if node.node_type == NodeType.KNOWLEDGE_RETRIEVAL:
                     outputs = self._parse_knowledge_retrieval_outputs(outputs)
-                elif node.node_type == BuiltinNodeTypes.LLM:
+                elif node.node_type == NodeType.LLM:
                     outputs = outputs.get("text", outputs)
                 node_span.end(
                     outputs=outputs,
@@ -471,13 +471,13 @@ class MLflowDataTrace(BaseTraceInstance):
     def _get_node_span_type(self, node_type: str) -> str:
         """Map Dify node types to MLflow span types"""
         node_type_mapping = {
-            BuiltinNodeTypes.LLM: SpanType.LLM,
-            BuiltinNodeTypes.QUESTION_CLASSIFIER: SpanType.LLM,
-            BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL: SpanType.RETRIEVER,
-            BuiltinNodeTypes.TOOL: SpanType.TOOL,
-            BuiltinNodeTypes.CODE: SpanType.TOOL,
-            BuiltinNodeTypes.HTTP_REQUEST: SpanType.TOOL,
-            BuiltinNodeTypes.AGENT: SpanType.AGENT,
+            NodeType.LLM: SpanType.LLM,
+            NodeType.QUESTION_CLASSIFIER: SpanType.LLM,
+            NodeType.KNOWLEDGE_RETRIEVAL: SpanType.RETRIEVER,
+            NodeType.TOOL: SpanType.TOOL,
+            NodeType.CODE: SpanType.TOOL,
+            NodeType.HTTP_REQUEST: SpanType.TOOL,
+            NodeType.AGENT: SpanType.AGENT,
         }
         return node_type_mapping.get(node_type, "CHAIN")  # type: ignore[arg-type,call-overload]
 

api/core/ops/opik_trace/opik_trace.py

@@ -23,7 +23,7 @@ from core.ops.entities.trace_entity import (
     WorkflowTraceInfo,
 )
 from core.repositories import DifyCoreRepositoryFactory
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey
 from extensions.ext_database import db
 from models import EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
 
@@ -187,7 +187,7 @@ class OpikDataTrace(BaseTraceInstance):
             node_name = node_execution.title
             node_type = node_execution.node_type
             status = node_execution.status
-            if node_type == BuiltinNodeTypes.LLM:
+            if node_type == NodeType.LLM:
                 inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
             else:
                 inputs = node_execution.inputs or {}

api/core/ops/tencent_trace/tencent_trace.py

@@ -27,7 +27,7 @@ from core.repositories import SQLAlchemyWorkflowNodeExecutionRepository
 from dify_graph.entities.workflow_node_execution import (
     WorkflowNodeExecution,
 )
-from dify_graph.nodes import BuiltinNodeTypes
+from dify_graph.nodes import NodeType
 from extensions.ext_database import db
 from models import Account, App, TenantAccountJoin, WorkflowNodeExecutionTriggeredFrom
 
@@ -179,7 +179,7 @@ class TencentDataTrace(BaseTraceInstance):
                     if node_span:
                         self.trace_client.add_span(node_span)
 
-                        if node_execution.node_type == BuiltinNodeTypes.LLM:
+                        if node_execution.node_type == NodeType.LLM:
                             self._record_llm_metrics(node_execution)
                 except Exception:
                     logger.exception("[Tencent APM] Failed to process node execution: %s", node_execution.id)
@@ -192,15 +192,15 @@ class TencentDataTrace(BaseTraceInstance):
     ) -> SpanData | None:
         """Build span for different node types"""
         try:
-            if node_execution.node_type == BuiltinNodeTypes.LLM:
+            if node_execution.node_type == NodeType.LLM:
                 return TencentSpanBuilder.build_workflow_llm_span(
                     trace_id, workflow_span_id, trace_info, node_execution
                 )
-            elif node_execution.node_type == BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL:
+            elif node_execution.node_type == NodeType.KNOWLEDGE_RETRIEVAL:
                 return TencentSpanBuilder.build_workflow_retrieval_span(
                     trace_id, workflow_span_id, trace_info, node_execution
                 )
-            elif node_execution.node_type == BuiltinNodeTypes.TOOL:
+            elif node_execution.node_type == NodeType.TOOL:
                 return TencentSpanBuilder.build_workflow_tool_span(
                     trace_id, workflow_span_id, trace_info, node_execution
                 )

api/core/ops/weave_trace/weave_trace.py

@@ -31,7 +31,7 @@ from core.ops.entities.trace_entity import (
 )
 from core.ops.weave_trace.entities.weave_trace_entity import WeaveTraceModel
 from core.repositories import DifyCoreRepositoryFactory
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey
 from extensions.ext_database import db
 from models import EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
 
@@ -175,7 +175,7 @@ class WeaveDataTrace(BaseTraceInstance):
             node_name = node_execution.title
             node_type = node_execution.node_type
             status = node_execution.status
-            if node_type == BuiltinNodeTypes.LLM:
+            if node_type == NodeType.LLM:
                 inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
             else:
                 inputs = node_execution.inputs or {}

api/core/plugin/backwards_invocation/node.py

@@ -1,5 +1,5 @@
 from core.plugin.backwards_invocation.base import BaseBackwardsInvocation
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from dify_graph.nodes.parameter_extractor.entities import (
     ModelConfig as ParameterExtractorModelConfig,
 )
@@ -52,7 +52,7 @@ class PluginNodeBackwardsInvocation(BaseBackwardsInvocation):
             instruction=instruction,  # instruct with variables are not supported
         )
         node_data_dict = node_data.model_dump()
-        node_data_dict["type"] = BuiltinNodeTypes.PARAMETER_EXTRACTOR
+        node_data_dict["type"] = NodeType.PARAMETER_EXTRACTOR
         execution = workflow_service.run_free_workflow_node(
             node_data_dict,
             tenant_id=tenant_id,

api/core/provider_manager.py

@@ -305,7 +305,9 @@ class ProviderManager:
             available_models = provider_configurations.get_models(model_type=model_type, only_active=True)
 
             if available_models:
-                available_model = available_models[0]
+                available_model = next(
+                    (model for model in available_models if model.model == "gpt-4"), available_models[0]
+                )
 
                 default_model = TenantDefaultModel(
                     tenant_id=tenant_id,

api/core/rag/datasource/vdb/hologres/hologres_vector.py

@@ -1,361 +0,0 @@
-import json
-import logging
-import time
-from typing import Any
-
-import holo_search_sdk as holo  # type: ignore
-from holo_search_sdk.types import BaseQuantizationType, DistanceType, TokenizerType
-from psycopg import sql as psql
-from pydantic import BaseModel, model_validator
-
-from configs import dify_config
-from core.rag.datasource.vdb.vector_base import BaseVector
-from core.rag.datasource.vdb.vector_factory import AbstractVectorFactory
-from core.rag.datasource.vdb.vector_type import VectorType
-from core.rag.embedding.embedding_base import Embeddings
-from core.rag.models.document import Document
-from extensions.ext_redis import redis_client
-from models.dataset import Dataset
-
-logger = logging.getLogger(__name__)
-
-
-class HologresVectorConfig(BaseModel):
-    """
-    Configuration for Hologres vector database connection.
-
-    In Hologres, access_key_id is used as the PostgreSQL username,
-    and access_key_secret is used as the PostgreSQL password.
-    """
-
-    host: str
-    port: int = 80
-    database: str
-    access_key_id: str
-    access_key_secret: str
-    schema_name: str = "public"
-    tokenizer: TokenizerType = "jieba"
-    distance_method: DistanceType = "Cosine"
-    base_quantization_type: BaseQuantizationType = "rabitq"
-    max_degree: int = 64
-    ef_construction: int = 400
-
-    @model_validator(mode="before")
-    @classmethod
-    def validate_config(cls, values: dict):
-        if not values.get("host"):
-            raise ValueError("config HOLOGRES_HOST is required")
-        if not values.get("database"):
-            raise ValueError("config HOLOGRES_DATABASE is required")
-        if not values.get("access_key_id"):
-            raise ValueError("config HOLOGRES_ACCESS_KEY_ID is required")
-        if not values.get("access_key_secret"):
-            raise ValueError("config HOLOGRES_ACCESS_KEY_SECRET is required")
-        return values
-
-
-class HologresVector(BaseVector):
-    """
-    Hologres vector storage implementation using holo-search-sdk.
-
-    Supports semantic search (vector), full-text search, and hybrid search.
-    """
-
-    def __init__(self, collection_name: str, config: HologresVectorConfig):
-        super().__init__(collection_name)
-        self._config = config
-        self._client = self._init_client(config)
-        self.table_name = f"embedding_{collection_name}".lower()
-
-    def _init_client(self, config: HologresVectorConfig):
-        """Initialize and return a holo-search-sdk client."""
-        client = holo.connect(
-            host=config.host,
-            port=config.port,
-            database=config.database,
-            access_key_id=config.access_key_id,
-            access_key_secret=config.access_key_secret,
-            schema=config.schema_name,
-        )
-        client.connect()
-        return client
-
-    def get_type(self) -> str:
-        return VectorType.HOLOGRES
-
-    def create(self, texts: list[Document], embeddings: list[list[float]], **kwargs):
-        """Create collection table with vector and full-text indexes, then add texts."""
-        dimension = len(embeddings[0])
-        self._create_collection(dimension)
-        self.add_texts(texts, embeddings)
-
-    def add_texts(self, documents: list[Document], embeddings: list[list[float]], **kwargs):
-        """Add texts with embeddings to the collection using batch upsert."""
-        if not documents:
-            return []
-
-        pks: list[str] = []
-        batch_size = 100
-        for i in range(0, len(documents), batch_size):
-            batch_docs = documents[i : i + batch_size]
-            batch_embeddings = embeddings[i : i + batch_size]
-
-            values = []
-            column_names = ["id", "text", "meta", "embedding"]
-
-            for j, doc in enumerate(batch_docs):
-                doc_id = doc.metadata.get("doc_id", "") if doc.metadata else ""
-                pks.append(doc_id)
-                values.append(
-                    [
-                        doc_id,
-                        doc.page_content,
-                        json.dumps(doc.metadata or {}),
-                        batch_embeddings[j],
-                    ]
-                )
-
-            table = self._client.open_table(self.table_name)
-            table.upsert_multi(
-                index_column="id",
-                values=values,
-                column_names=column_names,
-                update=True,
-                update_columns=["text", "meta", "embedding"],
-            )
-
-        return pks
-
-    def text_exists(self, id: str) -> bool:
-        """Check if a text with the given doc_id exists in the collection."""
-        if not self._client.check_table_exist(self.table_name):
-            return False
-
-        result = self._client.execute(
-            psql.SQL("SELECT 1 FROM {} WHERE id = {} LIMIT 1").format(
-                psql.Identifier(self.table_name), psql.Literal(id)
-            ),
-            fetch_result=True,
-        )
-        return bool(result)
-
-    def get_ids_by_metadata_field(self, key: str, value: str) -> list[str] | None:
-        """Get document IDs by metadata field key and value."""
-        result = self._client.execute(
-            psql.SQL("SELECT id FROM {} WHERE meta->>{} = {}").format(
-                psql.Identifier(self.table_name), psql.Literal(key), psql.Literal(value)
-            ),
-            fetch_result=True,
-        )
-        if result:
-            return [row[0] for row in result]
-        return None
-
-    def delete_by_ids(self, ids: list[str]):
-        """Delete documents by their doc_id list."""
-        if not ids:
-            return
-        if not self._client.check_table_exist(self.table_name):
-            return
-
-        self._client.execute(
-            psql.SQL("DELETE FROM {} WHERE id IN ({})").format(
-                psql.Identifier(self.table_name),
-                psql.SQL(", ").join(psql.Literal(id) for id in ids),
-            )
-        )
-
-    def delete_by_metadata_field(self, key: str, value: str):
-        """Delete documents by metadata field key and value."""
-        if not self._client.check_table_exist(self.table_name):
-            return
-
-        self._client.execute(
-            psql.SQL("DELETE FROM {} WHERE meta->>{} = {}").format(
-                psql.Identifier(self.table_name), psql.Literal(key), psql.Literal(value)
-            )
-        )
-
-    def search_by_vector(self, query_vector: list[float], **kwargs: Any) -> list[Document]:
-        """Search for documents by vector similarity."""
-        if not self._client.check_table_exist(self.table_name):
-            return []
-
-        top_k = kwargs.get("top_k", 4)
-        score_threshold = float(kwargs.get("score_threshold") or 0.0)
-
-        table = self._client.open_table(self.table_name)
-        query = (
-            table.search_vector(
-                vector=query_vector,
-                column="embedding",
-                distance_method=self._config.distance_method,
-                output_name="distance",
-            )
-            .select(["id", "text", "meta"])
-            .limit(top_k)
-        )
-
-        # Apply document_ids_filter if provided
-        document_ids_filter = kwargs.get("document_ids_filter")
-        if document_ids_filter:
-            filter_sql = psql.SQL("meta->>'document_id' IN ({})").format(
-                psql.SQL(", ").join(psql.Literal(id) for id in document_ids_filter)
-            )
-            query = query.where(filter_sql)
-
-        results = query.fetchall()
-        return self._process_vector_results(results, score_threshold)
-
-    def _process_vector_results(self, results: list, score_threshold: float) -> list[Document]:
-        """Process vector search results into Document objects."""
-        docs = []
-        for row in results:
-            # row format: (distance, id, text, meta)
-            # distance is first because search_vector() adds the computed column before selected columns
-            distance = row[0]
-            text = row[2]
-            meta = row[3]
-
-            if isinstance(meta, str):
-                meta = json.loads(meta)
-
-            # Convert distance to similarity score (consistent with pgvector)
-            score = 1 - distance
-            meta["score"] = score
-
-            if score >= score_threshold:
-                docs.append(Document(page_content=text, metadata=meta))
-
-        return docs
-
-    def search_by_full_text(self, query: str, **kwargs: Any) -> list[Document]:
-        """Search for documents by full-text search."""
-        if not self._client.check_table_exist(self.table_name):
-            return []
-
-        top_k = kwargs.get("top_k", 4)
-
-        table = self._client.open_table(self.table_name)
-        search_query = table.search_text(
-            column="text",
-            expression=query,
-            return_score=True,
-            return_score_name="score",
-            return_all_columns=True,
-        ).limit(top_k)
-
-        # Apply document_ids_filter if provided
-        document_ids_filter = kwargs.get("document_ids_filter")
-        if document_ids_filter:
-            filter_sql = psql.SQL("meta->>'document_id' IN ({})").format(
-                psql.SQL(", ").join(psql.Literal(id) for id in document_ids_filter)
-            )
-            search_query = search_query.where(filter_sql)
-
-        results = search_query.fetchall()
-        return self._process_full_text_results(results)
-
-    def _process_full_text_results(self, results: list) -> list[Document]:
-        """Process full-text search results into Document objects."""
-        docs = []
-        for row in results:
-            # row format: (id, text, meta, embedding, score)
-            text = row[1]
-            meta = row[2]
-            score = row[-1]  # score is the last column from return_score
-
-            if isinstance(meta, str):
-                meta = json.loads(meta)
-
-            meta["score"] = score
-            docs.append(Document(page_content=text, metadata=meta))
-
-        return docs
-
-    def delete(self):
-        """Delete the entire collection table."""
-        if self._client.check_table_exist(self.table_name):
-            self._client.drop_table(self.table_name)
-
-    def _create_collection(self, dimension: int):
-        """Create the collection table with vector and full-text indexes."""
-        lock_name = f"vector_indexing_lock_{self._collection_name}"
-        with redis_client.lock(lock_name, timeout=20):
-            collection_exist_cache_key = f"vector_indexing_{self._collection_name}"
-            if redis_client.get(collection_exist_cache_key):
-                return
-
-            if not self._client.check_table_exist(self.table_name):
-                # Create table via SQL with CHECK constraint for vector dimension
-                create_table_sql = psql.SQL("""
-                    CREATE TABLE IF NOT EXISTS {} (
-                        id TEXT PRIMARY KEY,
-                        text TEXT NOT NULL,
-                        meta JSONB NOT NULL,
-                        embedding float4[] NOT NULL
-                            CHECK (array_ndims(embedding) = 1
-                                   AND array_length(embedding, 1) = {})
-                    );
-                """).format(psql.Identifier(self.table_name), psql.Literal(dimension))
-                self._client.execute(create_table_sql)
-
-                # Wait for table to be fully ready before creating indexes
-                max_wait_seconds = 30
-                poll_interval = 2
-                for _ in range(max_wait_seconds // poll_interval):
-                    if self._client.check_table_exist(self.table_name):
-                        break
-                    time.sleep(poll_interval)
-                else:
-                    raise RuntimeError(f"Table {self.table_name} was not ready after {max_wait_seconds}s")
-
-                # Open table and set vector index
-                table = self._client.open_table(self.table_name)
-                table.set_vector_index(
-                    column="embedding",
-                    distance_method=self._config.distance_method,
-                    base_quantization_type=self._config.base_quantization_type,
-                    max_degree=self._config.max_degree,
-                    ef_construction=self._config.ef_construction,
-                    use_reorder=self._config.base_quantization_type == "rabitq",
-                )
-
-                # Create full-text search index
-                table.create_text_index(
-                    index_name=f"ft_idx_{self._collection_name}",
-                    column="text",
-                    tokenizer=self._config.tokenizer,
-                )
-
-            redis_client.set(collection_exist_cache_key, 1, ex=3600)
-
-
-class HologresVectorFactory(AbstractVectorFactory):
-    """Factory class for creating HologresVector instances."""
-
-    def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> HologresVector:
-        if dataset.index_struct_dict:
-            class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
-            collection_name = class_prefix
-        else:
-            dataset_id = dataset.id
-            collection_name = Dataset.gen_collection_name_by_id(dataset_id)
-            dataset.index_struct = json.dumps(self.gen_index_struct_dict(VectorType.HOLOGRES, collection_name))
-
-        return HologresVector(
-            collection_name=collection_name,
-            config=HologresVectorConfig(
-                host=dify_config.HOLOGRES_HOST or "",
-                port=dify_config.HOLOGRES_PORT,
-                database=dify_config.HOLOGRES_DATABASE or "",
-                access_key_id=dify_config.HOLOGRES_ACCESS_KEY_ID or "",
-                access_key_secret=dify_config.HOLOGRES_ACCESS_KEY_SECRET or "",
-                schema_name=dify_config.HOLOGRES_SCHEMA,
-                tokenizer=dify_config.HOLOGRES_TOKENIZER,
-                distance_method=dify_config.HOLOGRES_DISTANCE_METHOD,
-                base_quantization_type=dify_config.HOLOGRES_BASE_QUANTIZATION_TYPE,
-                max_degree=dify_config.HOLOGRES_MAX_DEGREE,
-                ef_construction=dify_config.HOLOGRES_EF_CONSTRUCTION,
-            ),
-        )

api/core/rag/datasource/vdb/pgvecto_rs/pgvecto_rs.py

@@ -135,8 +135,8 @@ class PGVectoRS(BaseVector):
     def get_ids_by_metadata_field(self, key: str, value: str):
         result = None
         with Session(self._client) as session:
-            select_statement = sql_text(f"SELECT id FROM {self._collection_name} WHERE meta->>:key = :value")
-            result = session.execute(select_statement, {"key": key, "value": value}).fetchall()
+            select_statement = sql_text(f"SELECT id FROM {self._collection_name} WHERE meta->>'{key}' = '{value}'; ")
+            result = session.execute(select_statement).fetchall()
         if result:
             return [item[0] for item in result]
         else:
@@ -172,9 +172,9 @@ class PGVectoRS(BaseVector):
     def text_exists(self, id: str) -> bool:
         with Session(self._client) as session:
             select_statement = sql_text(
-                f"SELECT id FROM {self._collection_name} WHERE meta->>'doc_id' = :doc_id limit 1"
+                f"SELECT id FROM {self._collection_name} WHERE meta->>'doc_id' = '{id}' limit 1; "
             )
-            result = session.execute(select_statement, {"doc_id": id}).fetchall()
+            result = session.execute(select_statement).fetchall()
         return len(result) > 0
 
     def search_by_vector(self, query_vector: list[float], **kwargs: Any) -> list[Document]:

api/core/rag/datasource/vdb/relyt/relyt_vector.py

@@ -154,8 +154,10 @@ class RelytVector(BaseVector):
     def get_ids_by_metadata_field(self, key: str, value: str):
         result = None
         with Session(self.client) as session:
-            select_statement = sql_text(f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>:key = :value""")
-            result = session.execute(select_statement, {"key": key, "value": value}).fetchall()
+            select_statement = sql_text(
+                f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>'{key}' = '{value}'; """
+            )
+            result = session.execute(select_statement).fetchall()
         if result:
             return [item[0] for item in result]
         else:
@@ -199,10 +201,11 @@ class RelytVector(BaseVector):
 
     def delete_by_ids(self, ids: list[str]):
         with Session(self.client) as session:
+            ids_str = ",".join(f"'{doc_id}'" for doc_id in ids)
             select_statement = sql_text(
-                f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>'doc_id' = ANY(:doc_ids)"""
+                f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>'doc_id' in ({ids_str}); """
             )
-            result = session.execute(select_statement, {"doc_ids": ids}).fetchall()
+            result = session.execute(select_statement).fetchall()
         if result:
             ids = [item[0] for item in result]
             self.delete_by_uuids(ids)
@@ -215,9 +218,9 @@ class RelytVector(BaseVector):
     def text_exists(self, id: str) -> bool:
         with Session(self.client) as session:
             select_statement = sql_text(
-                f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>'doc_id' = :doc_id limit 1"""
+                f"""SELECT id FROM "{self._collection_name}" WHERE metadata->>'doc_id' = '{id}' limit 1; """
             )
-            result = session.execute(select_statement, {"doc_id": id}).fetchall()
+            result = session.execute(select_statement).fetchall()
         return len(result) > 0
 
     def search_by_vector(self, query_vector: list[float], **kwargs: Any) -> list[Document]:

api/core/rag/datasource/vdb/vector_factory.py

@@ -38,7 +38,7 @@ class AbstractVectorFactory(ABC):
 class Vector:
     def __init__(self, dataset: Dataset, attributes: list | None = None):
         if attributes is None:
-            attributes = ["doc_id", "dataset_id", "document_id", "doc_hash", "doc_type"]
+            attributes = ["doc_id", "dataset_id", "document_id", "doc_hash"]
         self._dataset = dataset
         self._embeddings = self._get_embeddings()
         self._attributes = attributes
@@ -191,10 +191,6 @@ class Vector:
                 from core.rag.datasource.vdb.iris.iris_vector import IrisVectorFactory
 
                 return IrisVectorFactory
-            case VectorType.HOLOGRES:
-                from core.rag.datasource.vdb.hologres.hologres_vector import HologresVectorFactory
-
-                return HologresVectorFactory
             case _:
                 raise ValueError(f"Vector store {vector_type} is not supported.")
 

api/core/rag/datasource/vdb/vector_type.py

@@ -34,4 +34,3 @@ class VectorType(StrEnum):
     MATRIXONE = "matrixone"
     CLICKZETTA = "clickzetta"
     IRIS = "iris"
-    HOLOGRES = "hologres"

api/core/rag/datasource/vdb/weaviate/weaviate_vector.py

@@ -196,7 +196,6 @@ class WeaviateVector(BaseVector):
                             ),
                             wc.Property(name="document_id", data_type=wc.DataType.TEXT),
                             wc.Property(name="doc_id", data_type=wc.DataType.TEXT),
-                            wc.Property(name="doc_type", data_type=wc.DataType.TEXT),
                             wc.Property(name="chunk_index", data_type=wc.DataType.INT),
                         ],
                         vector_config=wc.Configure.Vectors.self_provided(),
@@ -226,8 +225,6 @@ class WeaviateVector(BaseVector):
             to_add.append(wc.Property(name="document_id", data_type=wc.DataType.TEXT))
         if "doc_id" not in existing:
             to_add.append(wc.Property(name="doc_id", data_type=wc.DataType.TEXT))
-        if "doc_type" not in existing:
-            to_add.append(wc.Property(name="doc_type", data_type=wc.DataType.TEXT))
         if "chunk_index" not in existing:
             to_add.append(wc.Property(name="chunk_index", data_type=wc.DataType.INT))
 

api/core/rag/index_processor/index_processor.py

@@ -9,8 +9,8 @@ from flask import current_app
 from sqlalchemy import delete, func, select
 
 from core.db.session_factory import session_factory
-from core.workflow.nodes.knowledge_index.exc import KnowledgeIndexNodeError
-from core.workflow.nodes.knowledge_index.protocols import Preview, PreviewItem, QaPreview
+from dify_graph.nodes.knowledge_index.exc import KnowledgeIndexNodeError
+from dify_graph.repositories.index_processor_protocol import Preview, PreviewItem, QaPreview
 from models.dataset import Dataset, Document, DocumentSegment
 
 from .index_processor_factory import IndexProcessorFactory

api/core/rag/index_processor/index_processor_base.py

@@ -294,7 +294,7 @@ class BaseIndexProcessor(ABC):
             logging.warning("Error downloading image from %s: %s", image_url, str(e))
             return None
         except Exception:
-            logging.warning("Unexpected error downloading image from %s", image_url, exc_info=True)
+            logging.exception("Unexpected error downloading image from %s", image_url)
             return None
 
     def _download_tool_file(self, tool_file_id: str, current_user: Account) -> str | None:

api/core/rag/retrieval/dataset_retrieval.py

@@ -56,18 +56,18 @@ from core.rag.retrieval.template_prompts import (
 )
 from core.tools.signature import sign_upload_file
 from core.tools.utils.dataset_retriever.dataset_retriever_base_tool import DatasetRetrieverBaseTool
-from core.workflow.nodes.knowledge_retrieval import exc
-from core.workflow.nodes.knowledge_retrieval.retrieval import (
-    KnowledgeRetrievalRequest,
-    Source,
-    SourceChildChunk,
-    SourceMetadata,
-)
 from dify_graph.file import File, FileTransferMethod, FileType
 from dify_graph.model_runtime.entities.llm_entities import LLMMode, LLMResult, LLMUsage
 from dify_graph.model_runtime.entities.message_entities import PromptMessage, PromptMessageRole, PromptMessageTool
 from dify_graph.model_runtime.entities.model_entities import ModelFeature, ModelType
 from dify_graph.model_runtime.model_providers.__base.large_language_model import LargeLanguageModel
+from dify_graph.nodes.knowledge_retrieval import exc
+from dify_graph.repositories.rag_retrieval_protocol import (
+    KnowledgeRetrievalRequest,
+    Source,
+    SourceChildChunk,
+    SourceMetadata,
+)
 from extensions.ext_database import db
 from extensions.ext_redis import redis_client
 from libs.json_in_md_parser import parse_and_check_json_markdown

api/core/repositories/sqlalchemy_workflow_node_execution_repository.py

@@ -18,7 +18,7 @@ from tenacity import before_sleep_log, retry, retry_if_exception, stop_after_att
 
 from configs import dify_config
 from dify_graph.entities import WorkflowNodeExecution
-from dify_graph.enums import WorkflowNodeExecutionMetadataKey, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey, WorkflowNodeExecutionStatus
 from dify_graph.model_runtime.utils.encoders import jsonable_encoder
 from dify_graph.repositories.workflow_node_execution_repository import OrderConfig, WorkflowNodeExecutionRepository
 from dify_graph.workflow_type_encoder import WorkflowRuntimeTypeConverter
@@ -146,7 +146,7 @@ class SQLAlchemyWorkflowNodeExecutionRepository(WorkflowNodeExecutionRepository)
             index=db_model.index,
             predecessor_node_id=db_model.predecessor_node_id,
             node_id=db_model.node_id,
-            node_type=db_model.node_type,
+            node_type=NodeType(db_model.node_type),
             title=db_model.title,
             inputs=inputs,
             process_data=process_data,

api/core/tools/utils/configuration.py

@@ -116,7 +116,6 @@ class ToolParameterConfigurationManager:
 
         return a deep copy of parameters with decrypted values
         """
-        parameters = self._deep_copy(parameters)
 
         cache = ToolParameterCache(
             tenant_id=self.tenant_id,

api/core/tools/utils/workflow_configuration_sync.py

@@ -3,7 +3,7 @@ from typing import Any
 
 from core.tools.entities.tool_entities import WorkflowToolParameterConfiguration
 from core.tools.errors import WorkflowToolHumanInputNotSupportedError
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 from dify_graph.nodes.base.entities import OutputVariableEntity
 from dify_graph.variables.input_entities import VariableEntity
 
@@ -51,7 +51,7 @@ class WorkflowToolConfigurationUtils:
     def ensure_no_human_input_nodes(cls, graph: Mapping[str, Any]) -> None:
         nodes = graph.get("nodes", [])
         for node in nodes:
-            if node.get("data", {}).get("type") == BuiltinNodeTypes.HUMAN_INPUT:
+            if node.get("data", {}).get("type") == NodeType.HUMAN_INPUT:
                 raise WorkflowToolHumanInputNotSupportedError()
 
     @classmethod

api/core/trigger/constants.py

@@ -1,18 +0,0 @@
-from typing import Final
-
-TRIGGER_WEBHOOK_NODE_TYPE: Final[str] = "trigger-webhook"
-TRIGGER_SCHEDULE_NODE_TYPE: Final[str] = "trigger-schedule"
-TRIGGER_PLUGIN_NODE_TYPE: Final[str] = "trigger-plugin"
-TRIGGER_INFO_METADATA_KEY: Final[str] = "trigger_info"
-
-TRIGGER_NODE_TYPES: Final[frozenset[str]] = frozenset(
-    {
-        TRIGGER_WEBHOOK_NODE_TYPE,
-        TRIGGER_SCHEDULE_NODE_TYPE,
-        TRIGGER_PLUGIN_NODE_TYPE,
-    }
-)
-
-
-def is_trigger_node_type(node_type: str) -> bool:
-    return node_type in TRIGGER_NODE_TYPES

api/core/trigger/debug/event_selectors.py

@@ -11,11 +11,6 @@ from typing import Any
 from pydantic import BaseModel
 
 from core.plugin.entities.request import TriggerInvokeEventResponse
-from core.trigger.constants import (
-    TRIGGER_PLUGIN_NODE_TYPE,
-    TRIGGER_SCHEDULE_NODE_TYPE,
-    TRIGGER_WEBHOOK_NODE_TYPE,
-)
 from core.trigger.debug.event_bus import TriggerDebugEventBus
 from core.trigger.debug.events import (
     PluginTriggerDebugEvent,
@@ -24,9 +19,10 @@ from core.trigger.debug.events import (
     build_plugin_pool_key,
     build_webhook_pool_key,
 )
-from core.workflow.nodes.trigger_plugin.entities import TriggerEventNodeData
-from core.workflow.nodes.trigger_schedule.entities import ScheduleConfig
 from dify_graph.entities.graph_config import NodeConfigDict
+from dify_graph.enums import NodeType
+from dify_graph.nodes.trigger_plugin.entities import TriggerEventNodeData
+from dify_graph.nodes.trigger_schedule.entities import ScheduleConfig
 from extensions.ext_redis import redis_client
 from libs.datetime_utils import ensure_naive_utc, naive_utc_now
 from libs.schedule_utils import calculate_next_run_at
@@ -210,19 +206,21 @@ def create_event_poller(
     if not node_config:
         raise ValueError("Node data not found for node %s", node_id)
     node_type = draft_workflow.get_node_type_from_node_config(node_config)
-    if node_type == TRIGGER_PLUGIN_NODE_TYPE:
-        return PluginTriggerDebugEventPoller(
-            tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
-        )
-    if node_type == TRIGGER_WEBHOOK_NODE_TYPE:
-        return WebhookTriggerDebugEventPoller(
-            tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
-        )
-    if node_type == TRIGGER_SCHEDULE_NODE_TYPE:
-        return ScheduleTriggerDebugEventPoller(
-            tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
-        )
-    raise ValueError("unable to create event poller for node type %s", node_type)
+    match node_type:
+        case NodeType.TRIGGER_PLUGIN:
+            return PluginTriggerDebugEventPoller(
+                tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
+            )
+        case NodeType.TRIGGER_WEBHOOK:
+            return WebhookTriggerDebugEventPoller(
+                tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
+            )
+        case NodeType.TRIGGER_SCHEDULE:
+            return ScheduleTriggerDebugEventPoller(
+                tenant_id=tenant_id, user_id=user_id, app_id=app_id, node_config=node_config, node_id=node_id
+            )
+        case _:
+            raise ValueError("unable to create event poller for node type %s", node_type)
 
 
 def select_trigger_debug_events(

api/core/workflow/__init__.py

@@ -1 +1,4 @@
-"""Core workflow package."""
+from .node_factory import DifyNodeFactory
+from .workflow_entry import WorkflowEntry
+
+__all__ = ["DifyNodeFactory", "WorkflowEntry"]

api/core/workflow/node_factory.py

@@ -1,7 +1,4 @@
-import importlib
-import pkgutil
-from collections.abc import Callable, Iterator, Mapping, MutableMapping
-from functools import lru_cache
+from collections.abc import Callable, Mapping
 from typing import TYPE_CHECKING, Any, TypeAlias, cast, final
 
 from sqlalchemy import select
@@ -11,6 +8,7 @@ from typing_extensions import override
 from configs import dify_config
 from core.app.entities.app_invoke_entities import DifyRunContext
 from core.app.llm.model_access import build_dify_model_access
+from core.datasource.datasource_manager import DatasourceManager
 from core.helper.code_executor.code_executor import (
     CodeExecutionError,
     CodeExecutor,
@@ -19,9 +17,12 @@ from core.helper.ssrf_proxy import ssrf_proxy
 from core.memory.token_buffer_memory import TokenBufferMemory
 from core.model_manager import ModelInstance
 from core.prompt.entities.advanced_prompt_entities import MemoryConfig
+from core.rag.index_processor.index_processor import IndexProcessor
+from core.rag.retrieval.dataset_retrieval import DatasetRetrieval
+from core.rag.summary_index.summary_index import SummaryIndex
 from core.repositories.human_input_repository import HumanInputFormRepositoryImpl
 from core.tools.tool_file_manager import ToolFileManager
-from core.trigger.constants import TRIGGER_NODE_TYPES
+from core.workflow.node_resolution import resolve_workflow_node_class
 from core.workflow.nodes.agent.message_transformer import AgentMessageTransformer
 from core.workflow.nodes.agent.plugin_strategy_adapter import (
     PluginAgentStrategyPresentationProvider,
@@ -31,7 +32,7 @@ from core.workflow.nodes.agent.runtime_support import AgentRuntimeSupport
 from dify_graph.entities.base_node_data import BaseNodeData
 from dify_graph.entities.graph_config import NodeConfigDict, NodeConfigDictAdapter
 from dify_graph.entities.graph_init_params import DIFY_RUN_CONTEXT_KEY
-from dify_graph.enums import BuiltinNodeTypes, NodeType, SystemVariableKey
+from dify_graph.enums import NodeType, SystemVariableKey
 from dify_graph.file.file_manager import file_manager
 from dify_graph.graph.graph import NodeFactory
 from dify_graph.model_runtime.entities.model_entities import ModelType
@@ -45,7 +46,6 @@ from dify_graph.nodes.document_extractor import UnstructuredApiConfig
 from dify_graph.nodes.http_request import build_http_request_config
 from dify_graph.nodes.llm.entities import LLMNodeData
 from dify_graph.nodes.llm.exc import LLMModeRequiredError, ModelNotExistError
-from dify_graph.nodes.llm.protocols import TemplateRenderer
 from dify_graph.nodes.parameter_extractor.entities import ParameterExtractorNodeData
 from dify_graph.nodes.question_classifier.entities import QuestionClassifierNodeData
 from dify_graph.nodes.template_transform.template_renderer import (
@@ -59,135 +59,6 @@ if TYPE_CHECKING:
     from dify_graph.entities import GraphInitParams
     from dify_graph.runtime import GraphRuntimeState
 
-LATEST_VERSION = "latest"
-_START_NODE_TYPES: frozenset[NodeType] = frozenset(
-    (BuiltinNodeTypes.START, BuiltinNodeTypes.DATASOURCE, *TRIGGER_NODE_TYPES)
-)
-
-
-def _import_node_package(package_name: str, *, excluded_modules: frozenset[str] = frozenset()) -> None:
-    package = importlib.import_module(package_name)
-    for _, module_name, _ in pkgutil.walk_packages(package.__path__, package.__name__ + "."):
-        if module_name in excluded_modules:
-            continue
-        importlib.import_module(module_name)
-
-
-@lru_cache(maxsize=1)
-def register_nodes() -> None:
-    """Import production node modules so they self-register with ``Node``."""
-    _import_node_package("dify_graph.nodes")
-    _import_node_package("core.workflow.nodes")
-
-
-def get_node_type_classes_mapping() -> Mapping[NodeType, Mapping[str, type[Node]]]:
-    """Return a read-only snapshot of the current production node registry.
-
-    The workflow layer owns node bootstrap because it must compose built-in
-    `dify_graph.nodes.*` implementations with workflow-local nodes under
-    `core.workflow.nodes.*`. Keeping this import side effect here avoids
-    reintroducing registry bootstrapping into lower-level graph primitives.
-    """
-    register_nodes()
-    return Node.get_node_type_classes_mapping()
-
-
-def resolve_workflow_node_class(*, node_type: NodeType, node_version: str) -> type[Node]:
-    node_mapping = get_node_type_classes_mapping().get(node_type)
-    if not node_mapping:
-        raise ValueError(f"No class mapping found for node type: {node_type}")
-
-    latest_node_class = node_mapping.get(LATEST_VERSION)
-    matched_node_class = node_mapping.get(node_version)
-    node_class = matched_node_class or latest_node_class
-    if not node_class:
-        raise ValueError(f"No latest version class found for node type: {node_type}")
-    return node_class
-
-
-def is_start_node_type(node_type: NodeType) -> bool:
-    """Return True when the node type can serve as a workflow entry point."""
-    return node_type in _START_NODE_TYPES
-
-
-def get_default_root_node_id(graph_config: Mapping[str, Any]) -> str:
-    """Resolve the default entry node for a persisted top-level workflow graph.
-
-    This workflow-layer helper depends on start-node semantics defined by
-    `is_start_node_type`, so it intentionally lives next to the node registry
-    instead of in the raw `dify_graph.entities.graph_config` schema module.
-    """
-    nodes = graph_config.get("nodes")
-    if not isinstance(nodes, list):
-        raise ValueError("nodes in workflow graph must be a list")
-
-    for node in nodes:
-        if not isinstance(node, Mapping):
-            continue
-
-        if node.get("type") == "custom-note":
-            continue
-
-        node_id = node.get("id")
-        data = node.get("data")
-        if not isinstance(node_id, str) or not isinstance(data, Mapping):
-            continue
-
-        node_type = data.get("type")
-        if isinstance(node_type, str) and is_start_node_type(node_type):
-            return node_id
-
-    raise ValueError("Unable to determine default root node ID from workflow graph")
-
-
-class _LazyNodeTypeClassesMapping(MutableMapping[NodeType, Mapping[str, type[Node]]]):
-    """Mutable dict-like view over the current node registry."""
-
-    def __init__(self) -> None:
-        self._cached_snapshot: dict[NodeType, Mapping[str, type[Node]]] = {}
-        self._cached_version = -1
-        self._deleted: set[NodeType] = set()
-        self._overrides: dict[NodeType, Mapping[str, type[Node]]] = {}
-
-    def _snapshot(self) -> dict[NodeType, Mapping[str, type[Node]]]:
-        current_version = Node.get_registry_version()
-        if self._cached_version != current_version:
-            self._cached_snapshot = dict(get_node_type_classes_mapping())
-            self._cached_version = current_version
-        if not self._deleted and not self._overrides:
-            return self._cached_snapshot
-
-        snapshot = {key: value for key, value in self._cached_snapshot.items() if key not in self._deleted}
-        snapshot.update(self._overrides)
-        return snapshot
-
-    def __getitem__(self, key: NodeType) -> Mapping[str, type[Node]]:
-        return self._snapshot()[key]
-
-    def __setitem__(self, key: NodeType, value: Mapping[str, type[Node]]) -> None:
-        self._deleted.discard(key)
-        self._overrides[key] = value
-
-    def __delitem__(self, key: NodeType) -> None:
-        if key in self._overrides:
-            del self._overrides[key]
-            return
-        if key in self._cached_snapshot:
-            self._deleted.add(key)
-            return
-        raise KeyError(key)
-
-    def __iter__(self) -> Iterator[NodeType]:
-        return iter(self._snapshot())
-
-    def __len__(self) -> int:
-        return len(self._snapshot())
-
-
-# Keep the canonical node-class mapping in the workflow layer that also bootstraps
-# legacy `core.workflow.nodes.*` registrations.
-NODE_TYPE_CLASSES_MAPPING: MutableMapping[NodeType, Mapping[str, type[Node]]] = _LazyNodeTypeClassesMapping()
-
 
 LLMCompatibleNodeData: TypeAlias = LLMNodeData | QuestionClassifierNodeData | ParameterExtractorNodeData
 
@@ -229,16 +100,6 @@ class DefaultWorkflowCodeExecutor:
         return isinstance(error, CodeExecutionError)
 
 
-class DefaultLLMTemplateRenderer(TemplateRenderer):
-    def render_jinja2(self, *, template: str, inputs: Mapping[str, Any]) -> str:
-        result = CodeExecutor.execute_workflow_code_template(
-            language=CodeLanguage.JINJA2,
-            code=template,
-            inputs=inputs,
-        )
-        return str(result.get("result", ""))
-
-
 @final
 class DifyNodeFactory(NodeFactory):
     """
@@ -265,11 +126,11 @@ class DifyNodeFactory(NodeFactory):
             max_object_array_length=dify_config.CODE_MAX_OBJECT_ARRAY_LENGTH,
         )
         self._template_renderer = CodeExecutorJinja2TemplateRenderer(code_executor=self._code_executor)
-        self._llm_template_renderer: TemplateRenderer = DefaultLLMTemplateRenderer()
         self._template_transform_max_output_length = dify_config.TEMPLATE_TRANSFORM_MAX_LENGTH
         self._http_request_http_client = ssrf_proxy
         self._http_request_tool_file_manager_factory = ToolFileManager
         self._http_request_file_manager = file_manager
+        self._rag_retrieval = DatasetRetrieval()
         self._document_extractor_unstructured_api_config = UnstructuredApiConfig(
             api_url=dify_config.UNSTRUCTURED_API_URL,
             api_key=dify_config.UNSTRUCTURED_API_KEY or "",
@@ -316,46 +177,56 @@ class DifyNodeFactory(NodeFactory):
         node_class = self._resolve_node_class(node_type=node_data.type, node_version=str(node_data.version))
         node_type = node_data.type
         node_init_kwargs_factories: Mapping[NodeType, Callable[[], dict[str, object]]] = {
-            BuiltinNodeTypes.CODE: lambda: {
+            NodeType.CODE: lambda: {
                 "code_executor": self._code_executor,
                 "code_limits": self._code_limits,
             },
-            BuiltinNodeTypes.TEMPLATE_TRANSFORM: lambda: {
+            NodeType.TEMPLATE_TRANSFORM: lambda: {
                 "template_renderer": self._template_renderer,
                 "max_output_length": self._template_transform_max_output_length,
             },
-            BuiltinNodeTypes.HTTP_REQUEST: lambda: {
+            NodeType.HTTP_REQUEST: lambda: {
                 "http_request_config": self._http_request_config,
                 "http_client": self._http_request_http_client,
                 "tool_file_manager_factory": self._http_request_tool_file_manager_factory,
                 "file_manager": self._http_request_file_manager,
             },
-            BuiltinNodeTypes.HUMAN_INPUT: lambda: {
+            NodeType.HUMAN_INPUT: lambda: {
                 "form_repository": HumanInputFormRepositoryImpl(tenant_id=self._dify_context.tenant_id),
             },
-            BuiltinNodeTypes.LLM: lambda: self._build_llm_compatible_node_init_kwargs(
+            NodeType.KNOWLEDGE_INDEX: lambda: {
+                "index_processor": IndexProcessor(),
+                "summary_index_service": SummaryIndex(),
+            },
+            NodeType.LLM: lambda: self._build_llm_compatible_node_init_kwargs(
                 node_class=node_class,
                 node_data=node_data,
                 include_http_client=True,
             ),
-            BuiltinNodeTypes.DOCUMENT_EXTRACTOR: lambda: {
+            NodeType.DATASOURCE: lambda: {
+                "datasource_manager": DatasourceManager,
+            },
+            NodeType.KNOWLEDGE_RETRIEVAL: lambda: {
+                "rag_retrieval": self._rag_retrieval,
+            },
+            NodeType.DOCUMENT_EXTRACTOR: lambda: {
                 "unstructured_api_config": self._document_extractor_unstructured_api_config,
                 "http_client": self._http_request_http_client,
             },
-            BuiltinNodeTypes.QUESTION_CLASSIFIER: lambda: self._build_llm_compatible_node_init_kwargs(
+            NodeType.QUESTION_CLASSIFIER: lambda: self._build_llm_compatible_node_init_kwargs(
                 node_class=node_class,
                 node_data=node_data,
                 include_http_client=True,
             ),
-            BuiltinNodeTypes.PARAMETER_EXTRACTOR: lambda: self._build_llm_compatible_node_init_kwargs(
+            NodeType.PARAMETER_EXTRACTOR: lambda: self._build_llm_compatible_node_init_kwargs(
                 node_class=node_class,
                 node_data=node_data,
                 include_http_client=False,
             ),
-            BuiltinNodeTypes.TOOL: lambda: {
+            NodeType.TOOL: lambda: {
                 "tool_file_manager_factory": self._http_request_tool_file_manager_factory(),
             },
-            BuiltinNodeTypes.AGENT: lambda: {
+            NodeType.AGENT: lambda: {
                 "strategy_resolver": self._agent_strategy_resolver,
                 "presentation_provider": self._agent_strategy_presentation_provider,
                 "runtime_support": self._agent_runtime_support,
@@ -403,8 +274,6 @@ class DifyNodeFactory(NodeFactory):
                 model_instance=model_instance,
             ),
         }
-        if validated_node_data.type in {BuiltinNodeTypes.LLM, BuiltinNodeTypes.QUESTION_CLASSIFIER}:
-            node_init_kwargs["template_renderer"] = self._llm_template_renderer
         if include_http_client:
             node_init_kwargs["http_client"] = self._http_request_http_client
         return node_init_kwargs

api/core/workflow/node_resolution.py

@@ -0,0 +1,42 @@
+from __future__ import annotations
+
+from collections.abc import Mapping
+from importlib import import_module
+
+from dify_graph.enums import NodeType
+from dify_graph.nodes.base.node import Node
+from dify_graph.nodes.node_mapping import LATEST_VERSION, get_node_type_classes_mapping
+
+_WORKFLOW_NODE_MODULES = ("core.workflow.nodes.agent",)
+_workflow_nodes_registered = False
+
+
+def ensure_workflow_nodes_registered() -> None:
+    """Import workflow-local node modules so they can register with `Node.__init_subclass__`."""
+    global _workflow_nodes_registered
+
+    if _workflow_nodes_registered:
+        return
+
+    for module_name in _WORKFLOW_NODE_MODULES:
+        import_module(module_name)
+
+    _workflow_nodes_registered = True
+
+
+def get_workflow_node_type_classes_mapping() -> Mapping[NodeType, Mapping[str, type[Node]]]:
+    ensure_workflow_nodes_registered()
+    return get_node_type_classes_mapping()
+
+
+def resolve_workflow_node_class(*, node_type: NodeType, node_version: str) -> type[Node]:
+    node_mapping = get_workflow_node_type_classes_mapping().get(node_type)
+    if not node_mapping:
+        raise ValueError(f"No class mapping found for node type: {node_type}")
+
+    latest_node_class = node_mapping.get(LATEST_VERSION)
+    matched_node_class = node_mapping.get(node_version)
+    node_class = matched_node_class or latest_node_class
+    if not node_class:
+        raise ValueError(f"No latest version class found for node type: {node_type}")
+    return node_class

api/core/workflow/nodes/__init__.py

@@ -1 +0,0 @@
-"""Workflow node implementations that remain under the legacy core.workflow namespace."""

api/core/workflow/nodes/agent/agent_node.py

@@ -4,7 +4,7 @@ from collections.abc import Generator, Mapping, Sequence
 from typing import TYPE_CHECKING, Any
 
 from dify_graph.entities.graph_config import NodeConfigDict
-from dify_graph.enums import BuiltinNodeTypes, SystemVariableKey, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, SystemVariableKey, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeEventBase, NodeRunResult, StreamCompletedEvent
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.base.variable_template_parser import VariableTemplateParser
@@ -24,7 +24,7 @@ if TYPE_CHECKING:
 
 
 class AgentNode(Node[AgentNodeData]):
-    node_type = BuiltinNodeTypes.AGENT
+    node_type = NodeType.AGENT
 
     _strategy_resolver: AgentStrategyResolver
     _presentation_provider: AgentStrategyPresentationProvider

api/core/workflow/nodes/agent/entities.py

@@ -6,11 +6,11 @@ from pydantic import BaseModel
 from core.prompt.entities.advanced_prompt_entities import MemoryConfig
 from core.tools.entities.tool_entities import ToolSelector
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 
 
 class AgentNodeData(BaseNodeData):
-    type: NodeType = BuiltinNodeTypes.AGENT
+    type: NodeType = NodeType.AGENT
     agent_strategy_provider_name: str
     agent_strategy_name: str
     agent_strategy_label: str

api/core/workflow/nodes/agent/message_transformer.py

@@ -8,7 +8,7 @@ from sqlalchemy.orm import Session
 
 from core.tools.entities.tool_entities import ToolInvokeMessage
 from core.tools.utils.message_transformer import ToolFileMessageTransformer
-from dify_graph.enums import BuiltinNodeTypes, NodeType, WorkflowNodeExecutionMetadataKey, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionMetadataKey, WorkflowNodeExecutionStatus
 from dify_graph.file import File, FileTransferMethod
 from dify_graph.model_runtime.entities.llm_entities import LLMUsage, LLMUsageMetadata
 from dify_graph.model_runtime.utils.encoders import jsonable_encoder
@@ -123,7 +123,7 @@ class AgentMessageTransformer:
                 )
             elif message.type == ToolInvokeMessage.MessageType.JSON:
                 assert isinstance(message.message, ToolInvokeMessage.JsonMessage)
-                if node_type == BuiltinNodeTypes.AGENT:
+                if node_type == NodeType.AGENT:
                     if isinstance(message.message.json_object, dict):
                         msg_metadata: dict[str, Any] = message.message.json_object.pop("execution_metadata", {})
                         llm_usage = LLMUsage.from_metadata(cast(LLMUsageMetadata, msg_metadata))

api/core/workflow/nodes/datasource/__init__.py

@@ -1 +0,0 @@
-"""Datasource workflow node package."""

api/core/workflow/nodes/knowledge_index/__init__.py

@@ -1,5 +0,0 @@
-"""Knowledge index workflow node package."""
-
-KNOWLEDGE_INDEX_NODE_TYPE = "knowledge-index"
-
-__all__ = ["KNOWLEDGE_INDEX_NODE_TYPE"]

api/core/workflow/nodes/knowledge_retrieval/__init__.py

@@ -1 +0,0 @@
-"""Knowledge retrieval workflow node package."""

api/core/workflow/nodes/trigger_schedule/__init__.py

@@ -1,3 +0,0 @@
-from .trigger_schedule_node import TriggerScheduleNode
-
-__all__ = ["TriggerScheduleNode"]

api/core/workflow/workflow_entry.py

@@ -8,7 +8,8 @@ from core.app.apps.exc import GenerateTaskStoppedError
 from core.app.entities.app_invoke_entities import InvokeFrom, UserFrom, build_dify_run_context
 from core.app.workflow.layers.llm_quota import LLMQuotaLayer
 from core.app.workflow.layers.observability import ObservabilityLayer
-from core.workflow.node_factory import DifyNodeFactory, resolve_workflow_node_class
+from core.workflow.node_factory import DifyNodeFactory
+from core.workflow.node_resolution import resolve_workflow_node_class
 from dify_graph.constants import ENVIRONMENT_VARIABLE_NODE_ID
 from dify_graph.entities import GraphInitParams
 from dify_graph.entities.graph_config import NodeConfigDictAdapter
@@ -21,7 +22,7 @@ from dify_graph.graph_engine.layers import DebugLoggingLayer, ExecutionLimitsLay
 from dify_graph.graph_engine.layers.base import GraphEngineLayer
 from dify_graph.graph_engine.protocols.command_channel import CommandChannel
 from dify_graph.graph_events import GraphEngineEvent, GraphNodeEventBase, GraphRunFailedEvent
-from dify_graph.nodes import BuiltinNodeTypes
+from dify_graph.nodes import NodeType
 from dify_graph.nodes.base.node import Node
 from dify_graph.runtime import ChildGraphNotFoundError, GraphRuntimeState, VariablePool
 from dify_graph.system_variable import SystemVariable
@@ -252,7 +253,7 @@ class WorkflowEntry:
             variable_mapping=variable_mapping,
             user_inputs=user_inputs,
         )
-        if node_type != BuiltinNodeTypes.DATASOURCE:
+        if node_type != NodeType.DATASOURCE:
             cls.mapping_user_inputs_to_variable_pool(
                 variable_mapping=variable_mapping,
                 user_inputs=user_inputs,
@@ -302,7 +303,7 @@ class WorkflowEntry:
             "height": node_height,
             "type": "custom",
             "data": {
-                "type": BuiltinNodeTypes.START,
+                "type": NodeType.START,
                 "title": "Start",
                 "desc": "Start",
             },
@@ -338,8 +339,8 @@ class WorkflowEntry:
         # Create a minimal graph for single node execution
         graph_dict = cls._create_single_node_graph(node_id, node_data)
 
-        node_type = node_data.get("type", "")
-        if node_type not in {BuiltinNodeTypes.PARAMETER_EXTRACTOR, BuiltinNodeTypes.QUESTION_CLASSIFIER}:
+        node_type = NodeType(node_data.get("type", ""))
+        if node_type not in {NodeType.PARAMETER_EXTRACTOR, NodeType.QUESTION_CLASSIFIER}:
             raise ValueError(f"Node type {node_type} not supported")
 
         node_cls = resolve_workflow_node_class(node_type=node_type, node_version="1")

api/dify_graph/README.md

@@ -113,7 +113,7 @@ The codebase enforces strict layering via import-linter:
 1. Create node class in `nodes/<node_type>/`
 1. Inherit from `BaseNode` or appropriate base class
 1. Implement `_run()` method
-1. Ensure the node module is importable under `nodes/<node_type>/`
+1. Register in `nodes/node_mapping.py`
 1. Add tests in `tests/unit_tests/dify_graph/nodes/`
 
 ### Implementing a Custom Layer

api/dify_graph/entities/base_node_data.py

@@ -121,8 +121,6 @@ class DefaultValue(BaseModel):
 class BaseNodeData(ABC, BaseModel):
     # Raw graph payloads are first validated through `NodeConfigDictAdapter`, where
     # `node["data"]` is typed as `BaseNodeData` before the concrete node class is known.
-    # `type` therefore accepts downstream string node kinds; unknown node implementations
-    # are rejected later when the node factory resolves the node registry.
     # At that boundary, node-specific fields are still "extra" relative to this shared DTO,
     # and persisted templates/workflows also carry undeclared compatibility keys such as
     # `selected`, `params`, `paramSchemas`, and `datasource_label`. Keep extras permissive

api/dify_graph/entities/workflow_node_execution.py

@@ -48,7 +48,7 @@ class WorkflowNodeExecution(BaseModel):
     index: int  # Sequence number for ordering in trace visualization
     predecessor_node_id: str | None = None  # ID of the node that executed before this one
     node_id: str  # ID of the node being executed
-    node_type: NodeType  # Type of node (e.g., start, llm, downstream response node)
+    node_type: NodeType  # Type of node (e.g., start, llm, knowledge)
     title: str  # Display title of the node
 
     # Execution data

api/dify_graph/enums.py

@@ -1,5 +1,4 @@
 from enum import StrEnum
-from typing import ClassVar, TypeAlias
 
 
 class NodeState(StrEnum):
@@ -34,71 +33,56 @@ class SystemVariableKey(StrEnum):
     INVOKE_FROM = "invoke_from"
 
 
-NodeType: TypeAlias = str
+class NodeType(StrEnum):
+    START = "start"
+    END = "end"
+    ANSWER = "answer"
+    LLM = "llm"
+    KNOWLEDGE_RETRIEVAL = "knowledge-retrieval"
+    KNOWLEDGE_INDEX = "knowledge-index"
+    IF_ELSE = "if-else"
+    CODE = "code"
+    TEMPLATE_TRANSFORM = "template-transform"
+    QUESTION_CLASSIFIER = "question-classifier"
+    HTTP_REQUEST = "http-request"
+    TOOL = "tool"
+    DATASOURCE = "datasource"
+    VARIABLE_AGGREGATOR = "variable-aggregator"
+    LEGACY_VARIABLE_AGGREGATOR = "variable-assigner"  # TODO: Merge this into VARIABLE_AGGREGATOR in the database.
+    LOOP = "loop"
+    LOOP_START = "loop-start"
+    LOOP_END = "loop-end"
+    ITERATION = "iteration"
+    ITERATION_START = "iteration-start"  # Fake start node for iteration.
+    PARAMETER_EXTRACTOR = "parameter-extractor"
+    VARIABLE_ASSIGNER = "assigner"
+    DOCUMENT_EXTRACTOR = "document-extractor"
+    LIST_OPERATOR = "list-operator"
+    AGENT = "agent"
+    TRIGGER_WEBHOOK = "trigger-webhook"
+    TRIGGER_SCHEDULE = "trigger-schedule"
+    TRIGGER_PLUGIN = "trigger-plugin"
+    HUMAN_INPUT = "human-input"
 
+    @property
+    def is_trigger_node(self) -> bool:
+        """Check if this node type is a trigger node."""
+        return self in [
+            NodeType.TRIGGER_WEBHOOK,
+            NodeType.TRIGGER_SCHEDULE,
+            NodeType.TRIGGER_PLUGIN,
+        ]
 
-class BuiltinNodeTypes:
-    """Built-in node type string constants.
-
-    `node_type` values are plain strings throughout the graph runtime. This namespace
-    only exposes the built-in values shipped by `dify_graph`; downstream packages can
-    use additional strings without extending this class.
-    """
-
-    START: ClassVar[NodeType] = "start"
-    END: ClassVar[NodeType] = "end"
-    ANSWER: ClassVar[NodeType] = "answer"
-    LLM: ClassVar[NodeType] = "llm"
-    KNOWLEDGE_RETRIEVAL: ClassVar[NodeType] = "knowledge-retrieval"
-    IF_ELSE: ClassVar[NodeType] = "if-else"
-    CODE: ClassVar[NodeType] = "code"
-    TEMPLATE_TRANSFORM: ClassVar[NodeType] = "template-transform"
-    QUESTION_CLASSIFIER: ClassVar[NodeType] = "question-classifier"
-    HTTP_REQUEST: ClassVar[NodeType] = "http-request"
-    TOOL: ClassVar[NodeType] = "tool"
-    DATASOURCE: ClassVar[NodeType] = "datasource"
-    VARIABLE_AGGREGATOR: ClassVar[NodeType] = "variable-aggregator"
-    LEGACY_VARIABLE_AGGREGATOR: ClassVar[NodeType] = "variable-assigner"
-    LOOP: ClassVar[NodeType] = "loop"
-    LOOP_START: ClassVar[NodeType] = "loop-start"
-    LOOP_END: ClassVar[NodeType] = "loop-end"
-    ITERATION: ClassVar[NodeType] = "iteration"
-    ITERATION_START: ClassVar[NodeType] = "iteration-start"
-    PARAMETER_EXTRACTOR: ClassVar[NodeType] = "parameter-extractor"
-    VARIABLE_ASSIGNER: ClassVar[NodeType] = "assigner"
-    DOCUMENT_EXTRACTOR: ClassVar[NodeType] = "document-extractor"
-    LIST_OPERATOR: ClassVar[NodeType] = "list-operator"
-    AGENT: ClassVar[NodeType] = "agent"
-    HUMAN_INPUT: ClassVar[NodeType] = "human-input"
-
-
-BUILT_IN_NODE_TYPES: tuple[NodeType, ...] = (
-    BuiltinNodeTypes.START,
-    BuiltinNodeTypes.END,
-    BuiltinNodeTypes.ANSWER,
-    BuiltinNodeTypes.LLM,
-    BuiltinNodeTypes.KNOWLEDGE_RETRIEVAL,
-    BuiltinNodeTypes.IF_ELSE,
-    BuiltinNodeTypes.CODE,
-    BuiltinNodeTypes.TEMPLATE_TRANSFORM,
-    BuiltinNodeTypes.QUESTION_CLASSIFIER,
-    BuiltinNodeTypes.HTTP_REQUEST,
-    BuiltinNodeTypes.TOOL,
-    BuiltinNodeTypes.DATASOURCE,
-    BuiltinNodeTypes.VARIABLE_AGGREGATOR,
-    BuiltinNodeTypes.LEGACY_VARIABLE_AGGREGATOR,
-    BuiltinNodeTypes.LOOP,
-    BuiltinNodeTypes.LOOP_START,
-    BuiltinNodeTypes.LOOP_END,
-    BuiltinNodeTypes.ITERATION,
-    BuiltinNodeTypes.ITERATION_START,
-    BuiltinNodeTypes.PARAMETER_EXTRACTOR,
-    BuiltinNodeTypes.VARIABLE_ASSIGNER,
-    BuiltinNodeTypes.DOCUMENT_EXTRACTOR,
-    BuiltinNodeTypes.LIST_OPERATOR,
-    BuiltinNodeTypes.AGENT,
-    BuiltinNodeTypes.HUMAN_INPUT,
-)
+    @property
+    def is_start_node(self) -> bool:
+        """Check if this node type can serve as a workflow entry point."""
+        return self in [
+            NodeType.START,
+            NodeType.DATASOURCE,
+            NodeType.TRIGGER_WEBHOOK,
+            NodeType.TRIGGER_SCHEDULE,
+            NodeType.TRIGGER_PLUGIN,
+        ]
 
 
 class NodeExecutionType(StrEnum):
@@ -252,6 +236,7 @@ class WorkflowNodeExecutionMetadataKey(StrEnum):
     CURRENCY = "currency"
     TOOL_INFO = "tool_info"
     AGENT_LOG = "agent_log"
+    TRIGGER_INFO = "trigger_info"
     ITERATION_ID = "iteration_id"
     ITERATION_INDEX = "iteration_index"
     LOOP_ID = "loop_id"

api/dify_graph/graph/graph.py

@@ -83,6 +83,50 @@ class Graph:
 
         return node_configs_map
 
+    @classmethod
+    def _find_root_node_id(
+        cls,
+        node_configs_map: Mapping[str, NodeConfigDict],
+        edge_configs: Sequence[Mapping[str, object]],
+        root_node_id: str | None = None,
+    ) -> str:
+        """
+        Find the root node ID if not specified.
+
+        :param node_configs_map: mapping of node ID to node config
+        :param edge_configs: list of edge configurations
+        :param root_node_id: explicitly specified root node ID
+        :return: determined root node ID
+        """
+        if root_node_id:
+            if root_node_id not in node_configs_map:
+                raise ValueError(f"Root node id {root_node_id} not found in the graph")
+            return root_node_id
+
+        # Find nodes with no incoming edges
+        nodes_with_incoming: set[str] = set()
+        for edge_config in edge_configs:
+            target = edge_config.get("target")
+            if isinstance(target, str):
+                nodes_with_incoming.add(target)
+
+        root_candidates = [nid for nid in node_configs_map if nid not in nodes_with_incoming]
+
+        # Prefer START node if available
+        start_node_id = None
+        for nid in root_candidates:
+            node_data = node_configs_map[nid]["data"]
+            if node_data.type.is_start_node:
+                start_node_id = nid
+                break
+
+        root_node_id = start_node_id or (root_candidates[0] if root_candidates else None)
+
+        if not root_node_id:
+            raise ValueError("Unable to determine root node ID")
+
+        return root_node_id
+
     @classmethod
     def _build_edges(
         cls, edge_configs: list[dict[str, object]]
@@ -257,15 +301,15 @@ class Graph:
         *,
         graph_config: Mapping[str, object],
         node_factory: NodeFactory,
-        root_node_id: str,
+        root_node_id: str | None = None,
         skip_validation: bool = False,
     ) -> Graph:
         """
-        Initialize a graph with an explicit execution entry point.
+        Initialize graph
 
         :param graph_config: graph config containing nodes and edges
         :param node_factory: factory for creating node instances from config data
-        :param root_node_id: active root node id
+        :param root_node_id: root node id
         :return: graph instance
         """
         # Parse configs
@@ -283,8 +327,8 @@ class Graph:
         # Parse node configurations
         node_configs_map = cls._parse_node_configs(node_configs)
 
-        if root_node_id not in node_configs_map:
-            raise ValueError(f"Root node id {root_node_id} not found in the graph")
+        # Find root node
+        root_node_id = cls._find_root_node_id(node_configs_map, edge_configs, root_node_id)
 
         # Build edges
         edges, in_edges, out_edges = cls._build_edges(edge_configs)

api/dify_graph/graph/validation.py

@@ -4,7 +4,7 @@ from collections.abc import Sequence
 from dataclasses import dataclass
 from typing import TYPE_CHECKING, Protocol
 
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, NodeType
+from dify_graph.enums import NodeExecutionType, NodeType
 
 if TYPE_CHECKING:
     from .graph import Graph
@@ -71,7 +71,7 @@ class _RootNodeValidator:
     """Validates root node invariants."""
 
     invalid_root_code: str = "INVALID_ROOT"
-    container_entry_types: tuple[NodeType, ...] = (BuiltinNodeTypes.ITERATION_START, BuiltinNodeTypes.LOOP_START)
+    container_entry_types: tuple[NodeType, ...] = (NodeType.ITERATION_START, NodeType.LOOP_START)
 
     def validate(self, graph: Graph) -> Sequence[GraphValidationIssue]:
         root_node = graph.root_node
@@ -86,7 +86,7 @@ class _RootNodeValidator:
             )
             return issues
 
-        node_type = root_node.node_type
+        node_type = getattr(root_node, "node_type", None)
         if root_node.execution_type != NodeExecutionType.ROOT and node_type not in self.container_entry_types:
             issues.append(
                 GraphValidationIssue(
@@ -114,9 +114,45 @@ class GraphValidator:
             raise GraphValidationError(issues)
 
 
+@dataclass(frozen=True, slots=True)
+class _TriggerStartExclusivityValidator:
+    """Ensures trigger nodes do not coexist with UserInput (start) nodes."""
+
+    conflict_code: str = "TRIGGER_START_NODE_CONFLICT"
+
+    def validate(self, graph: Graph) -> Sequence[GraphValidationIssue]:
+        start_node_id: str | None = None
+        trigger_node_ids: list[str] = []
+
+        for node in graph.nodes.values():
+            node_type = getattr(node, "node_type", None)
+            if not isinstance(node_type, NodeType):
+                continue
+
+            if node_type == NodeType.START:
+                start_node_id = node.id
+            elif node_type.is_trigger_node:
+                trigger_node_ids.append(node.id)
+
+        if start_node_id and trigger_node_ids:
+            trigger_list = ", ".join(trigger_node_ids)
+            return [
+                GraphValidationIssue(
+                    code=self.conflict_code,
+                    message=(
+                        f"UserInput (start) node '{start_node_id}' cannot coexist with trigger nodes: {trigger_list}."
+                    ),
+                    node_id=start_node_id,
+                )
+            ]
+
+        return []
+
+
 _DEFAULT_RULES: tuple[GraphValidationRule, ...] = (
     _EdgeEndpointValidator(),
     _RootNodeValidator(),
+    _TriggerStartExclusivityValidator(),
 )
 
 

api/dify_graph/graph_engine/response_coordinator/__init__.py

@@ -6,6 +6,5 @@ of responses based on upstream node outputs and constants.
 """
 
 from .coordinator import ResponseStreamCoordinator
-from .session import RESPONSE_SESSION_NODE_TYPES
 
-__all__ = ["RESPONSE_SESSION_NODE_TYPES", "ResponseStreamCoordinator"]
+__all__ = ["ResponseStreamCoordinator"]

api/dify_graph/graph_engine/response_coordinator/session.py

@@ -3,34 +3,19 @@ Internal response session management for response coordinator.
 
 This module contains the private ResponseSession class used internally
 by ResponseStreamCoordinator to manage streaming sessions.
-
-`RESPONSE_SESSION_NODE_TYPES` is intentionally mutable so downstream applications
-can opt additional response-capable node types into session creation without
-patching the coordinator.
 """
 
 from __future__ import annotations
 
 from dataclasses import dataclass
-from typing import Protocol, cast
 
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.nodes.answer.answer_node import AnswerNode
 from dify_graph.nodes.base.template import Template
+from dify_graph.nodes.end.end_node import EndNode
+from dify_graph.nodes.knowledge_index import KnowledgeIndexNode
 from dify_graph.runtime.graph_runtime_state import NodeProtocol
 
 
-class _ResponseSessionNodeProtocol(NodeProtocol, Protocol):
-    """Structural contract required from nodes that can open a response session."""
-
-    def get_streaming_template(self) -> Template: ...
-
-
-RESPONSE_SESSION_NODE_TYPES: list[NodeType] = [
-    BuiltinNodeTypes.ANSWER,
-    BuiltinNodeTypes.END,
-]
-
-
 @dataclass
 class ResponseSession:
     """
@@ -48,9 +33,10 @@ class ResponseSession:
         """
         Create a ResponseSession from a response-capable node.
 
-        The parameter is typed as `NodeProtocol` because the graph is exposed behind a protocol at the runtime layer.
-        At runtime this must be a node whose `node_type` is listed in `RESPONSE_SESSION_NODE_TYPES`
-        and which implements `get_streaming_template()`.
+        The parameter is typed as `NodeProtocol` because the graph is exposed behind a protocol at the runtime layer,
+        but at runtime this must be an `AnswerNode`, `EndNode`, or `KnowledgeIndexNode` that provides:
+        - `id: str`
+        - `get_streaming_template() -> Template`
 
         Args:
             node: Node from the materialized workflow graph.
@@ -61,22 +47,11 @@ class ResponseSession:
         Raises:
             TypeError: If node is not a supported response node type.
         """
-        if node.node_type not in RESPONSE_SESSION_NODE_TYPES:
-            supported_node_types = ", ".join(RESPONSE_SESSION_NODE_TYPES)
-            raise TypeError(
-                "ResponseSession.from_node only supports node types in "
-                f"RESPONSE_SESSION_NODE_TYPES: {supported_node_types}"
-            )
-
-        response_node = cast(_ResponseSessionNodeProtocol, node)
-        try:
-            template = response_node.get_streaming_template()
-        except AttributeError as exc:
-            raise TypeError("ResponseSession.from_node requires get_streaming_template() on response nodes") from exc
-
+        if not isinstance(node, AnswerNode | EndNode | KnowledgeIndexNode):
+            raise TypeError("ResponseSession.from_node only supports AnswerNode, EndNode, or KnowledgeIndexNode")
         return cls(
             node_id=node.id,
-            template=template,
+            template=node.get_streaming_template(),
         )
 
     def is_complete(self) -> bool:

api/dify_graph/node_events/node.py

@@ -1,9 +1,9 @@
-from collections.abc import Mapping, Sequence
+from collections.abc import Sequence
 from datetime import datetime
-from typing import Any
 
 from pydantic import Field
 
+from core.rag.entities.citation_metadata import RetrievalSourceMetadata
 from dify_graph.entities.pause_reason import PauseReason
 from dify_graph.file import File
 from dify_graph.model_runtime.entities.llm_entities import LLMUsage
@@ -13,7 +13,7 @@ from .base import NodeEventBase
 
 
 class RunRetrieverResourceEvent(NodeEventBase):
-    retriever_resources: Sequence[Mapping[str, Any]] = Field(..., description="retriever resources")
+    retriever_resources: Sequence[RetrievalSourceMetadata] = Field(..., description="retriever resources")
     context: str = Field(..., description="context")
     context_files: list[File] | None = Field(default=None, description="context files")
 

api/dify_graph/nodes/__init__.py

@@ -1,3 +1,3 @@
-from dify_graph.enums import BuiltinNodeTypes
+from dify_graph.enums import NodeType
 
-__all__ = ["BuiltinNodeTypes"]
+__all__ = ["NodeType"]

api/dify_graph/nodes/answer/answer_node.py

@@ -1,7 +1,7 @@
 from collections.abc import Mapping, Sequence
 from typing import Any
 
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeExecutionType, NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.answer.entities import AnswerNodeData
 from dify_graph.nodes.base.node import Node
@@ -11,7 +11,7 @@ from dify_graph.variables import ArrayFileSegment, FileSegment, Segment
 
 
 class AnswerNode(Node[AnswerNodeData]):
-    node_type = BuiltinNodeTypes.ANSWER
+    node_type = NodeType.ANSWER
     execution_type = NodeExecutionType.RESPONSE
 
     @classmethod

api/dify_graph/nodes/answer/entities.py

@@ -4,7 +4,7 @@ from enum import StrEnum, auto
 from pydantic import BaseModel, Field
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 
 
 class AnswerNodeData(BaseNodeData):
@@ -12,7 +12,7 @@ class AnswerNodeData(BaseNodeData):
     Answer Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.ANSWER
+    type: NodeType = NodeType.ANSWER
     answer: str = Field(..., description="answer template string")
 
 

api/dify_graph/nodes/base/node.py

@@ -1,7 +1,9 @@
 from __future__ import annotations
 
+import importlib
 import logging
 import operator
+import pkgutil
 from abc import abstractmethod
 from collections.abc import Generator, Mapping, Sequence
 from functools import singledispatchmethod
@@ -159,7 +161,7 @@ class Node(Generic[NodeDataT]):
 
         Example:
             class CodeNode(Node[CodeNodeData]):  # CodeNodeData is auto-extracted
-                node_type = BuiltinNodeTypes.CODE
+                node_type = NodeType.CODE
                 # No need to implement _get_title, _get_error_strategy, etc.
         """
         super().__init_subclass__(**kwargs)
@@ -177,8 +179,7 @@ class Node(Generic[NodeDataT]):
         # Skip base class itself
         if cls is Node:
             return
-        # Only register production node implementations defined under the
-        # canonical workflow namespaces.
+        # Only register production node implementations defined under dify_graph.nodes.*
         # This prevents test helper subclasses from polluting the global registry and
         # accidentally overriding real node types (e.g., a test Answer node).
         module_name = getattr(cls, "__module__", "")
@@ -186,7 +187,7 @@ class Node(Generic[NodeDataT]):
         node_type = cls.node_type
         version = cls.version()
         bucket = Node._registry.setdefault(node_type, {})
-        if module_name.startswith(("dify_graph.nodes.", "core.workflow.nodes.")):
+        if module_name.startswith("dify_graph.nodes."):
             # Production node definitions take precedence and may override
             bucket[version] = cls  # type: ignore[index]
         else:
@@ -202,7 +203,6 @@ class Node(Generic[NodeDataT]):
         else:
             latest_key = max(version_keys) if version_keys else version
         bucket["latest"] = bucket[latest_key]
-        Node._registry_version += 1
 
     @classmethod
     def _extract_node_data_type_from_generic(cls) -> type[BaseNodeData] | None:
@@ -237,11 +237,6 @@ class Node(Generic[NodeDataT]):
 
     # Global registry populated via __init_subclass__
     _registry: ClassVar[dict[NodeType, dict[str, type[Node]]]] = {}
-    _registry_version: ClassVar[int] = 0
-
-    @classmethod
-    def get_registry_version(cls) -> int:
-        return cls._registry_version
 
     def __init__(
         self,
@@ -274,10 +269,6 @@ class Node(Generic[NodeDataT]):
         """Validate shared graph node payloads against the subclass-declared NodeData model."""
         return cast(NodeDataT, cls._node_data_type.model_validate(node_data, from_attributes=True))
 
-    def init_node_data(self, data: BaseNodeData | Mapping[str, Any]) -> None:
-        """Hydrate `_node_data` for legacy callers that bypass `__init__`."""
-        self._node_data = self.validate_node_data(cast(BaseNodeData, data))
-
     def post_init(self) -> None:
         """Optional hook for subclasses requiring extra initialization."""
         return
@@ -498,19 +489,29 @@ class Node(Generic[NodeDataT]):
     def version(cls) -> str:
         """`node_version` returns the version of current node type."""
         # NOTE(QuantumGhost): Node versions must remain unique per `NodeType` so
-        # registry lookups can resolve numeric versions and `latest`.
+        # `Node.get_node_type_classes_mapping()` can resolve numeric versions and `latest`.
         raise NotImplementedError("subclasses of BaseNode must implement `version` method.")
 
     @classmethod
     def get_node_type_classes_mapping(cls) -> Mapping[NodeType, Mapping[str, type[Node]]]:
-        """Return a read-only view of the currently registered node classes.
+        """Return mapping of NodeType -> {version -> Node subclass} using __init_subclass__ registry.
 
-        This accessor intentionally performs no imports. The embedding layer that
-        owns bootstrap (for example `core.workflow.node_factory`) must import any
-        extension node packages before calling it so their subclasses register via
-        `__init_subclass__`.
+        Import all modules under dify_graph.nodes so subclasses register themselves on import.
+        Callers that rely on workflow-local nodes defined outside `dify_graph.nodes` must import
+        those modules before invoking this method so they can register through `__init_subclass__`.
+        We then return a readonly view of the registry to avoid accidental mutation.
         """
-        return {node_type: MappingProxyType(version_map) for node_type, version_map in cls._registry.items()}
+        # Import all node modules to ensure they are loaded (thus registered)
+        import dify_graph.nodes as _nodes_pkg
+
+        for _, _modname, _ in pkgutil.walk_packages(_nodes_pkg.__path__, _nodes_pkg.__name__ + "."):
+            # Avoid importing modules that depend on the registry to prevent circular imports.
+            if _modname == "dify_graph.nodes.node_mapping":
+                continue
+            importlib.import_module(_modname)
+
+        # Return a readonly view so callers can't mutate the registry by accident
+        return {nt: MappingProxyType(ver_map) for nt, ver_map in cls._registry.items()}
 
     @property
     def retry(self) -> bool:
@@ -785,16 +786,11 @@ class Node(Generic[NodeDataT]):
 
     @_dispatch.register
     def _(self, event: RunRetrieverResourceEvent) -> NodeRunRetrieverResourceEvent:
-        from core.rag.entities.citation_metadata import RetrievalSourceMetadata
-
-        retriever_resources = [
-            RetrievalSourceMetadata.model_validate(resource) for resource in event.retriever_resources
-        ]
         return NodeRunRetrieverResourceEvent(
             id=self.execution_id,
             node_id=self._node_id,
             node_type=self.node_type,
-            retriever_resources=retriever_resources,
+            retriever_resources=event.retriever_resources,
             context=event.context,
             node_version=self.version(),
         )

api/dify_graph/nodes/code/code_node.py

@@ -4,7 +4,7 @@ from textwrap import dedent
 from typing import TYPE_CHECKING, Any, Protocol, cast
 
 from dify_graph.entities.graph_config import NodeConfigDict
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.code.entities import CodeLanguage, CodeNodeData
@@ -72,7 +72,7 @@ _DEFAULT_CODE_BY_LANGUAGE: Mapping[CodeLanguage, str] = {
 
 
 class CodeNode(Node[CodeNodeData]):
-    node_type = BuiltinNodeTypes.CODE
+    node_type = NodeType.CODE
     _limits: CodeNodeLimits
 
     def __init__(

api/dify_graph/nodes/code/entities.py

@@ -4,7 +4,7 @@ from typing import Annotated, Literal
 from pydantic import AfterValidator, BaseModel
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.nodes.base.entities import VariableSelector
 from dify_graph.variables.types import SegmentType
 
@@ -40,7 +40,7 @@ class CodeNodeData(BaseNodeData):
     Code Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.CODE
+    type: NodeType = NodeType.CODE
 
     class Output(BaseModel):
         type: Annotated[SegmentType, AfterValidator(_validate_type)]

api/dify_graph/nodes/datasource/__init__.py

@@ -0,0 +1,3 @@
+from .datasource_node import DatasourceNode
+
+__all__ = ["DatasourceNode"]

api/dify_graph/nodes/datasource/datasource_node.py

@@ -1,17 +1,22 @@
 from collections.abc import Generator, Mapping, Sequence
 from typing import TYPE_CHECKING, Any
 
-from core.datasource.datasource_manager import DatasourceManager
 from core.datasource.entities.datasource_entities import DatasourceProviderType
 from core.plugin.impl.exc import PluginDaemonClientSideError
 from dify_graph.entities.graph_config import NodeConfigDict
 from dify_graph.entities.workflow_node_execution import WorkflowNodeExecutionStatus
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, SystemVariableKey, WorkflowNodeExecutionMetadataKey
+from dify_graph.enums import NodeExecutionType, NodeType, SystemVariableKey
 from dify_graph.node_events import NodeRunResult, StreamCompletedEvent
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.base.variable_template_parser import VariableTemplateParser
+from dify_graph.repositories.datasource_manager_protocol import (
+    DatasourceManagerProtocol,
+    DatasourceParameter,
+    OnlineDriveDownloadFileParam,
+)
 
-from .entities import DatasourceNodeData, DatasourceParameter, OnlineDriveDownloadFileParam
+from ...entities.workflow_node_execution import WorkflowNodeExecutionMetadataKey
+from .entities import DatasourceNodeData
 from .exc import DatasourceNodeError
 
 if TYPE_CHECKING:
@@ -24,7 +29,7 @@ class DatasourceNode(Node[DatasourceNodeData]):
     Datasource Node
     """
 
-    node_type = BuiltinNodeTypes.DATASOURCE
+    node_type = NodeType.DATASOURCE
     execution_type = NodeExecutionType.ROOT
 
     def __init__(
@@ -33,6 +38,7 @@ class DatasourceNode(Node[DatasourceNodeData]):
         config: NodeConfigDict,
         graph_init_params: "GraphInitParams",
         graph_runtime_state: "GraphRuntimeState",
+        datasource_manager: DatasourceManagerProtocol,
     ):
         super().__init__(
             id=id,
@@ -40,7 +46,7 @@ class DatasourceNode(Node[DatasourceNodeData]):
             graph_init_params=graph_init_params,
             graph_runtime_state=graph_runtime_state,
         )
-        self.datasource_manager = DatasourceManager
+        self.datasource_manager = datasource_manager
 
     def populate_start_event(self, event) -> None:
         event.provider_id = f"{self.node_data.plugin_id}/{self.node_data.provider_name}"

api/dify_graph/nodes/datasource/entities.py

@@ -4,7 +4,7 @@ from pydantic import BaseModel, field_validator
 from pydantic_core.core_schema import ValidationInfo
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 
 
 class DatasourceEntity(BaseModel):
@@ -17,7 +17,7 @@ class DatasourceEntity(BaseModel):
 
 
 class DatasourceNodeData(BaseNodeData, DatasourceEntity):
-    type: NodeType = BuiltinNodeTypes.DATASOURCE
+    type: NodeType = NodeType.DATASOURCE
 
     class DatasourceInput(BaseModel):
         # TODO: check this type
@@ -42,14 +42,3 @@ class DatasourceNodeData(BaseNodeData, DatasourceEntity):
             return typ
 
     datasource_parameters: dict[str, DatasourceInput] | None = None
-
-
-class DatasourceParameter(BaseModel):
-    workspace_id: str
-    page_id: str
-    type: str
-
-
-class OnlineDriveDownloadFileParam(BaseModel):
-    id: str
-    bucket: str

api/dify_graph/nodes/document_extractor/entities.py

@@ -2,11 +2,11 @@ from collections.abc import Sequence
 from dataclasses import dataclass
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 
 
 class DocumentExtractorNodeData(BaseNodeData):
-    type: NodeType = BuiltinNodeTypes.DOCUMENT_EXTRACTOR
+    type: NodeType = NodeType.DOCUMENT_EXTRACTOR
     variable_selector: Sequence[str]
 
 

api/dify_graph/nodes/document_extractor/node.py

@@ -22,7 +22,7 @@ from docx.table import Table
 from docx.text.paragraph import Paragraph
 
 from dify_graph.entities.graph_config import NodeConfigDict
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionStatus
 from dify_graph.file import File, FileTransferMethod, file_manager
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base.node import Node
@@ -46,7 +46,7 @@ class DocumentExtractorNode(Node[DocumentExtractorNodeData]):
     Supports plain text, PDF, and DOC/DOCX files.
     """
 
-    node_type = BuiltinNodeTypes.DOCUMENT_EXTRACTOR
+    node_type = NodeType.DOCUMENT_EXTRACTOR
 
     @classmethod
     def version(cls) -> str:

api/dify_graph/nodes/end/end_node.py

@@ -1,4 +1,4 @@
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeExecutionType, NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.base.template import Template
@@ -6,7 +6,7 @@ from dify_graph.nodes.end.entities import EndNodeData
 
 
 class EndNode(Node[EndNodeData]):
-    node_type = BuiltinNodeTypes.END
+    node_type = NodeType.END
     execution_type = NodeExecutionType.RESPONSE
 
     @classmethod

api/dify_graph/nodes/end/entities.py

@@ -1,7 +1,7 @@
 from pydantic import BaseModel, Field
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.nodes.base.entities import OutputVariableEntity
 
 
@@ -10,7 +10,7 @@ class EndNodeData(BaseNodeData):
     END Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.END
+    type: NodeType = NodeType.END
     outputs: list[OutputVariableEntity]
 
 

api/dify_graph/nodes/http_request/entities.py

@@ -9,7 +9,7 @@ import httpx
 from pydantic import BaseModel, Field, ValidationInfo, field_validator
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 
 HTTP_REQUEST_CONFIG_FILTER_KEY = "http_request_config"
 
@@ -90,7 +90,7 @@ class HttpRequestNodeData(BaseNodeData):
     Code Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.HTTP_REQUEST
+    type: NodeType = NodeType.HTTP_REQUEST
     method: Literal[
         "get",
         "post",

api/dify_graph/nodes/http_request/node.py

@@ -4,7 +4,7 @@ from collections.abc import Callable, Mapping, Sequence
 from typing import TYPE_CHECKING, Any
 
 from dify_graph.entities.graph_config import NodeConfigDict
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionStatus
 from dify_graph.file import File, FileTransferMethod
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base import variable_template_parser
@@ -33,7 +33,7 @@ if TYPE_CHECKING:
 
 
 class HttpRequestNode(Node[HttpRequestNodeData]):
-    node_type = BuiltinNodeTypes.HTTP_REQUEST
+    node_type = NodeType.HTTP_REQUEST
 
     def __init__(
         self,

api/dify_graph/nodes/human_input/entities.py

@@ -11,7 +11,7 @@ from typing import Annotated, Any, ClassVar, Literal, Self
 from pydantic import BaseModel, Field, field_validator, model_validator
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.nodes.base.variable_template_parser import VariableTemplateParser
 from dify_graph.runtime import VariablePool
 from dify_graph.variables.consts import SELECTORS_LENGTH
@@ -215,7 +215,7 @@ class UserAction(BaseModel):
 class HumanInputNodeData(BaseNodeData):
     """Human Input node data."""
 
-    type: NodeType = BuiltinNodeTypes.HUMAN_INPUT
+    type: NodeType = NodeType.HUMAN_INPUT
     delivery_methods: list[DeliveryChannelConfig] = Field(default_factory=list)
     form_content: str = ""
     inputs: list[FormInput] = Field(default_factory=list)

api/dify_graph/nodes/human_input/human_input_node.py

@@ -5,7 +5,7 @@ from typing import TYPE_CHECKING, Any
 
 from dify_graph.entities.graph_config import NodeConfigDict
 from dify_graph.entities.pause_reason import HumanInputRequired
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeExecutionType, NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import (
     HumanInputFormFilledEvent,
     HumanInputFormTimeoutEvent,
@@ -40,7 +40,7 @@ logger = logging.getLogger(__name__)
 
 
 class HumanInputNode(Node[HumanInputNodeData]):
-    node_type = BuiltinNodeTypes.HUMAN_INPUT
+    node_type = NodeType.HUMAN_INPUT
     execution_type = NodeExecutionType.BRANCH
 
     _BRANCH_SELECTION_KEYS: tuple[str, ...] = (

api/dify_graph/nodes/if_else/entities.py

@@ -3,7 +3,7 @@ from typing import Literal
 from pydantic import BaseModel, Field
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.utils.condition.entities import Condition
 
 
@@ -12,7 +12,7 @@ class IfElseNodeData(BaseNodeData):
     If Else Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.IF_ELSE
+    type: NodeType = NodeType.IF_ELSE
 
     class Case(BaseModel):
         """

api/dify_graph/nodes/if_else/if_else_node.py

@@ -3,7 +3,7 @@ from typing import Any, Literal
 
 from typing_extensions import deprecated
 
-from dify_graph.enums import BuiltinNodeTypes, NodeExecutionType, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeExecutionType, NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.if_else.entities import IfElseNodeData
@@ -13,7 +13,7 @@ from dify_graph.utils.condition.processor import ConditionProcessor
 
 
 class IfElseNode(Node[IfElseNodeData]):
-    node_type = BuiltinNodeTypes.IF_ELSE
+    node_type = NodeType.IF_ELSE
     execution_type = NodeExecutionType.BRANCH
 
     @classmethod

api/dify_graph/nodes/iteration/entities.py

@@ -4,7 +4,7 @@ from typing import Any
 from pydantic import Field
 
 from dify_graph.entities.base_node_data import BaseNodeData
-from dify_graph.enums import BuiltinNodeTypes, NodeType
+from dify_graph.enums import NodeType
 from dify_graph.nodes.base import BaseIterationNodeData, BaseIterationState
 
 
@@ -19,7 +19,7 @@ class IterationNodeData(BaseIterationNodeData):
     Iteration Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.ITERATION
+    type: NodeType = NodeType.ITERATION
     parent_loop_id: str | None = None  # redundant field, not used currently
     iterator_selector: list[str]  # variable selector
     output_selector: list[str]  # output selector
@@ -34,7 +34,7 @@ class IterationStartNodeData(BaseNodeData):
     Iteration Start Node Data.
     """
 
-    type: NodeType = BuiltinNodeTypes.ITERATION_START
+    type: NodeType = NodeType.ITERATION_START
 
 
 class IterationState(BaseIterationState):

api/dify_graph/nodes/iteration/iteration_node.py

@@ -9,8 +9,8 @@ from typing_extensions import TypeIs
 from dify_graph.constants import CONVERSATION_VARIABLE_NODE_ID
 from dify_graph.entities.graph_config import NodeConfigDictAdapter
 from dify_graph.enums import (
-    BuiltinNodeTypes,
     NodeExecutionType,
+    NodeType,
     WorkflowNodeExecutionMetadataKey,
     WorkflowNodeExecutionStatus,
 )
@@ -62,7 +62,7 @@ class IterationNode(LLMUsageTrackingMixin, Node[IterationNodeData]):
     Iteration Node.
     """
 
-    node_type = BuiltinNodeTypes.ITERATION
+    node_type = NodeType.ITERATION
     execution_type = NodeExecutionType.CONTAINER
 
     @classmethod
@@ -485,9 +485,12 @@ class IterationNode(LLMUsageTrackingMixin, Node[IterationNodeData]):
 
             # variable selector to variable mapping
             try:
+                # Get node class
+                from dify_graph.nodes.node_mapping import get_node_type_classes_mapping
+
                 typed_sub_node_config = NodeConfigDictAdapter.validate_python(sub_node_config)
                 node_type = typed_sub_node_config["data"].type
-                node_mapping = Node.get_node_type_classes_mapping()
+                node_mapping = get_node_type_classes_mapping()
                 if node_type not in node_mapping:
                     continue
                 node_version = str(typed_sub_node_config["data"].version)
@@ -560,7 +563,7 @@ class IterationNode(LLMUsageTrackingMixin, Node[IterationNodeData]):
             raise IterationIndexNotFoundError(f"iteration {self._node_id} current index not found")
         current_index = index_variable.value
         for event in rst:
-            if isinstance(event, GraphNodeEventBase) and event.node_type == BuiltinNodeTypes.ITERATION_START:
+            if isinstance(event, GraphNodeEventBase) and event.node_type == NodeType.ITERATION_START:
                 continue
 
             if isinstance(event, GraphNodeEventBase):

api/dify_graph/nodes/iteration/iteration_start_node.py

@@ -1,4 +1,4 @@
-from dify_graph.enums import BuiltinNodeTypes, WorkflowNodeExecutionStatus
+from dify_graph.enums import NodeType, WorkflowNodeExecutionStatus
 from dify_graph.node_events import NodeRunResult
 from dify_graph.nodes.base.node import Node
 from dify_graph.nodes.iteration.entities import IterationStartNodeData
@@ -9,7 +9,7 @@ class IterationStartNode(Node[IterationStartNodeData]):
     Iteration Start Node.
     """
 
-    node_type = BuiltinNodeTypes.ITERATION_START
+    node_type = NodeType.ITERATION_START
 
     @classmethod
     def version(cls) -> str: