Feat: Limit the number of Splitter and Parser operators on the canvas to only one #9869 (#10362)

### What problem does this PR solve? Feat: Limit the number of Splitter and Parser operators on the canvas to only one #9869 ### Type of change - [x] New Feature (non-breaking change which adds functionality)
2025-12-20 12:56:55 +08:00 · 2025-09-29 17:22:40 +08:00
parent 8bc8126848
commit 63cdce660e
7 changed files with 93 additions and 148 deletions
--- a/web/src/locales/en.ts
+++ b/web/src/locales/en.ts
@ -1687,15 +1687,20 @@ This delimiter is used to split the input text into several text pieces echo of
    },
    dataflow: {
      parser: 'Parser',
-      parserDescription: 'Parser',
-      chunker: 'Chunker',
-      chunkerDescription: 'Chunker',
+      parserDescription:
+        'Extracts raw text and structure from files for downstream processing.',
      tokenizer: 'Tokenizer',
-      tokenizerDescription: 'Tokenizer',
-      splitter: 'Splitter',
-      splitterDescription: 'Splitter',
-      hierarchicalMergerDescription: 'Hierarchical merger',
-      hierarchicalMerger: 'Hierarchical merger',
+      tokenizerDescription:
+        'Transforms text into the required data structure (e.g., vector embeddings for Embedding Search) depending on the chosen search method.',
+      splitter: 'Token Splitter',
+      splitterDescription:
+        'Split text into chunks by token length with optional delimiters and overlap.',
+      hierarchicalMergerDescription:
+        'Split documents into sections by title hierarchy with regex rules for finer control.',
+      hierarchicalMerger: 'Title Splitter',
+      extractor: 'Context Generator',
+      extractorDescription:
+        'Use an LLM to extract structured insights from document chunks—such as summaries, classifications, etc.',
      outputFormat: 'Output format',
      lang: 'Language',
      fileFormats: 'File formats',
@ -1713,8 +1718,6 @@ This delimiter is used to split the input text into several text pieces echo of
      exportJson: 'Export JSON',
      viewResult: 'View Result',
      running: 'Running',
-      extractor: 'Extractor',
-      extractorDescription: 'Extractor',
      summary: 'Augmented Context',
      keywords: 'Keywords',
      questions: 'Questions',
--- a/web/src/locales/zh.ts
+++ b/web/src/locales/zh.ts
@ -1605,15 +1605,19 @@ General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于
    },
    dataflow: {
      parser: '解析器',
-      parserDescription: '解析器',
-      chunker: '分块器',
-      chunkerDescription: '分块器',
+      parserDescription: '从文件中提取原始文本和结构以供下游处理。',
      tokenizer: '分词器',
-      tokenizerDescription: '分词器',
-      splitter: '拆分器',
-      splitterDescription: '拆分器',
-      hierarchicalMergerDesription: '分层合并',
-      hierarchicalMerger: '分层合并',
+      tokenizerDescription:
+        '根据所选的搜索方法，将文本转换为所需的数据结构（例如，用于嵌入搜索的向量嵌入）。',
+      splitter: '分词器拆分器',
+      splitterDescription:
+        '根据分词器长度将文本拆分成块，并带有可选的分隔符和重叠。',
+      hierarchicalMergerDescription:
+        '使用正则表达式规则按标题层次结构将文档拆分成多个部分，以实现更精细的控制。',
+      hierarchicalMerger: '标题拆分器',
+      extractor: '提取器',
+      extractorDescription:
+        '使用 LLM 从文档块（例如摘要、分类等）中提取结构化见解。',
      outputFormat: '输出格式',
      lang: '语言',
      fileFormats: '文件格式',
@ -1632,8 +1636,6 @@ General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于
      exportJson: '导出 JSON',
      viewResult: '查看结果',
      running: '运行中',
-      extractor: '提取器',
-      extractorDescription: '提取器',
      summary: '增强上下文',
      keywords: '关键词',
      questions: '问题',
@ -1687,6 +1689,7 @@ General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于
        },
      },
      cancel: '取消',
+      filenameEmbeddingWeight: '文件名嵌入权重',
      switchPromptMessage: '提示词将发生变化，请确认是否放弃已有提示词？',
    },
  },
--- a/web/src/pages/data-flow/canvas/index.tsx
+++ b/web/src/pages/data-flow/canvas/index.tsx
@ -12,6 +12,7 @@ import {
  ControlButton,
  Controls,
  NodeTypes,
+  OnConnectEnd,
  Position,
  ReactFlow,
  ReactFlowInstance,
@ -36,12 +37,13 @@ import {
  useShowDrawer,
 } from '../hooks/use-show-drawer';
 import RunSheet from '../run-sheet';
+import useGraphStore from '../store';
 import { ButtonEdge } from './edge';
 import styles from './index.less';
 import { RagNode } from './node';
 import { BeginNode } from './node/begin-node';
 import { ContextNode } from './node/context-node';
-import { InnerNextStepDropdown } from './node/dropdown/next-step-dropdown';
+import { NextStepDropdown } from './node/dropdown/next-step-dropdown';
 import { HierarchicalMergerNode } from './node/hierarchical-merger-node';
 import NoteNode from './node/note-node';
 import ParserNode from './node/parser-node';
@ -116,6 +118,7 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
  useHideFormSheetOnNodeDeletion({ hideFormDrawer });

  const { visible, hideModal, showModal } = useSetModalState();
+
  const [dropdownPosition, setDropdownPosition] = useState({ x: 0, y: 0 });

  const isConnectedRef = useRef(false);
@ -128,6 +131,8 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {

  const { setActiveDropdown, clearActiveDropdown } = useDropdownManager();

+  const { hasChildNode } = useGraphStore((state) => state);
+
  const onPaneClick = useCallback(() => {
    hideFormDrawer();
    if (visible && !preventCloseRef.current) {
@ -159,7 +164,7 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
    isConnectedRef.current = true;
  };

-  const OnConnectStart = (event: any, params: any) => {
+  const onConnectStart = (event: any, params: any) => {
    isConnectedRef.current = false;

    if (params && params.nodeId && params.handleId) {
@ -172,7 +177,12 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
    }
  };

-  const OnConnectEnd = (event: MouseEvent | TouchEvent) => {
+  const onConnectEnd: OnConnectEnd = (event, connectionState) => {
+    const nodeId = connectionState.fromNode?.id;
+    // Events triggered by Handle are directly interrupted
+    if (connectionState.toNode !== null || (nodeId && hasChildNode(nodeId))) {
+      return;
+    }
    if ('clientX' in event && 'clientY' in event) {
      const { clientX, clientY } = event;
      setDropdownPosition({ x: clientX, y: clientY });
@ -220,8 +230,8 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
          onConnect={onConnect}
          nodeTypes={nodeTypes}
          edgeTypes={edgeTypes}
-          onConnectStart={OnConnectStart}
-          onConnectEnd={OnConnectEnd}
+          onConnectStart={onConnectStart}
+          onConnectEnd={onConnectEnd}
          onNodeClick={onNodeClick}
          onPaneClick={onPaneClick}
          onInit={setReactFlowInstance}
@ -268,7 +278,7 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
              isFromConnectionDrag: true,
            }}
          >
-            <InnerNextStepDropdown
+            <NextStepDropdown
              hideModal={() => {
                hideModal();
                clearActiveDropdown();
@ -276,7 +286,7 @@ function DataFlowCanvas({ drawerVisible, hideDrawer, showLogSheet }: IProps) {
              position={dropdownPosition}
            >
              <span></span>
-            </InnerNextStepDropdown>
+            </NextStepDropdown>
          </HandleContext.Provider>
        )}
      </AgentInstanceContext.Provider>
--- a/web/src/pages/data-flow/canvas/node/dropdown/next-step-dropdown.tsx
+++ b/web/src/pages/data-flow/canvas/node/dropdown/next-step-dropdown.tsx
@ -12,6 +12,7 @@ import {
 } from '@/components/ui/tooltip';
 import { IModalProps } from '@/interfaces/common';
 import { useGetNodeDescription, useGetNodeName } from '@/pages/data-flow/hooks';
+import useGraphStore from '@/pages/data-flow/store';
 import { Position } from '@xyflow/react';
 import { t } from 'i18next';
 import {
@ -20,6 +21,7 @@ import {
  memo,
  useContext,
  useEffect,
+  useMemo,
  useRef,
 } from 'react';
 import { Operator } from '../../../constant';
@ -110,6 +112,26 @@ function OperatorItemList({
  return <ul className="space-y-2">{operators.map(renderOperatorItem)}</ul>;
 }

+const singleOperators = [
+  Operator.Tokenizer,
+  Operator.Splitter,
+  Operator.HierarchicalMerger,
+  Operator.Parser,
+];
+// Limit the number of operators of a certain type on the canvas to only one
+function useRestrictSingleOperatorOnCanvas() {
+  const list: Operator[] = [];
+  const { findNodeByName } = useGraphStore((state) => state);
+
+  singleOperators.forEach((operator) => {
+    if (!findNodeByName(operator)) {
+      list.push(operator);
+    }
+  });
+
+  return list;
+}
+
 function AccordionOperators({
  isCustomDropdown = false,
  mousePosition,
@ -117,15 +139,16 @@ function AccordionOperators({
  isCustomDropdown?: boolean;
  mousePosition?: { x: number; y: number };
 }) {
+  const singleOperators = useRestrictSingleOperatorOnCanvas();
+  const operators = useMemo(() => {
+    const list = [...singleOperators];
+    list.push(Operator.Extractor);
+    return list;
+  }, [singleOperators]);
+
  return (
    <OperatorItemList
-      operators={[
-        Operator.Parser,
-        Operator.Tokenizer,
-        Operator.Splitter,
-        Operator.HierarchicalMerger,
-        Operator.Extractor,
-      ]}
+      operators={operators}
      isCustomDropdown={isCustomDropdown}
      mousePosition={mousePosition}
    ></OperatorItemList>
--- a/web/src/pages/data-flow/canvas/node/handle.tsx
+++ b/web/src/pages/data-flow/canvas/node/handle.tsx
@ -4,8 +4,9 @@ import { Handle, HandleProps } from '@xyflow/react';
 import { Plus } from 'lucide-react';
 import { useMemo } from 'react';
 import { HandleContext } from '../../context';
+import useGraphStore from '../../store';
 import { useDropdownManager } from '../context';
-import { InnerNextStepDropdown } from './dropdown/next-step-dropdown';
+import { NextStepDropdown } from './dropdown/next-step-dropdown';

 export function CommonHandle({
  className,
@ -17,6 +18,8 @@ export function CommonHandle({
  const { canShowDropdown, setActiveDropdown, clearActiveDropdown } =
    useDropdownManager();

+  const { hasChildNode } = useGraphStore((state) => state);
+
  const value = useMemo(
    () => ({
      nodeId,
@ -39,6 +42,10 @@ export function CommonHandle({
        onClick={(e) => {
          e.stopPropagation();

+          if (hasChildNode(nodeId)) {
+            return;
+          }
+
          if (!canShowDropdown()) {
            return;
          }
@ -49,14 +56,14 @@ export function CommonHandle({
      >
        <Plus className="size-3 pointer-events-none text-text-title-invert" />
        {visible && (
-          <InnerNextStepDropdown
+          <NextStepDropdown
            hideModal={() => {
              hideModal();
              clearActiveDropdown();
            }}
          >
            <span></span>
-          </InnerNextStepDropdown>
+          </NextStepDropdown>
        )}
      </Handle>
    </HandleContext.Provider>
--- a/web/src/pages/data-flow/canvas/node/tokenizer-node.tsx
+++ b/web/src/pages/data-flow/canvas/node/tokenizer-node.tsx
@ -4,7 +4,7 @@ import { memo } from 'react';
 import { NodeHandleId } from '../../constant';
 import { needsSingleStepDebugging } from '../../utils';
 import { CommonHandle } from './handle';
-import { LeftHandleStyle, RightHandleStyle } from './handle-icon';
+import { LeftHandleStyle } from './handle-icon';
 import NodeHeader from './node-header';
 import { NodeWrapper } from './node-wrapper';
 import { ToolBar } from './toolbar';
@ -31,15 +31,6 @@ function TokenizerNode({
          style={LeftHandleStyle}
          nodeId={id}
        ></CommonHandle>
-        <CommonHandle
-          type="source"
-          position={Position.Right}
-          isConnectable={isConnectable}
-          id={NodeHandleId.Start}
-          style={RightHandleStyle}
-          nodeId={id}
-          isConnectableEnd={false}
-        ></CommonHandle>
        <NodeHeader id={id} name={data.name} label={data.label}></NodeHeader>
      </NodeWrapper>
    </ToolBar>
--- a/web/src/pages/data-flow/store.ts
+++ b/web/src/pages/data-flow/store.ts
@ -14,7 +14,6 @@ import {
  applyEdgeChanges,
  applyNodeChanges,
 } from '@xyflow/react';
-import { omit } from 'lodash';
 import differenceWith from 'lodash/differenceWith';
 import intersectionWith from 'lodash/intersectionWith';
 import lodashSet from 'lodash/set';
@ -59,7 +58,6 @@ export type RFState = {
  updateNode: (node: RAGFlowNodeType) => void;
  addEdge: (connection: Connection) => void;
  getEdge: (id: string) => Edge | undefined;
-  updateFormDataOnConnect: (connection: Connection) => void;
  updateSwitchFormData: (
    source: string,
    sourceHandle?: string | null,
@ -67,7 +65,6 @@ export type RFState = {
    isConnecting?: boolean,
  ) => void;
  duplicateNode: (id: string, name: string) => void;
-  duplicateIterationNode: (id: string, name: string) => void;
  deleteEdge: () => void;
  deleteEdgeById: (id: string) => void;
  deleteNodeById: (id: string) => void;
@ -89,6 +86,7 @@ export type RFState = {
  ) => void; // Deleting a condition of a classification operator will delete the related edge
  findAgentToolNodeById: (id: string | null) => string | undefined;
  selectNodeIds: (nodeIds: string[]) => void;
+  hasChildNode: (nodeId: string) => boolean;
 };

 // this is our useStore hook that we can use in our components to get parts of the store and call actions
@ -126,11 +124,9 @@ const useGraphStore = create<RFState>()(
        setEdges(mapEdgeMouseEvent(edges, edgeId, false));
      },
      onConnect: (connection: Connection) => {
-        const { updateFormDataOnConnect } = get();
        set({
          edges: addEdge(connection, get().edges),
        });
-        updateFormDataOnConnect(connection);
      },
      onSelectionChange: ({ nodes, edges }: OnSelectionChangeParams) => {
        set({
@ -217,37 +213,14 @@ const useGraphStore = create<RFState>()(
        set({
          edges: addEdge(connection, get().edges),
        });
-        //  TODO: This may not be reasonable. You need to choose between listening to changes in the form.
-        get().updateFormDataOnConnect(connection);
      },
      getEdge: (id: string) => {
        return get().edges.find((x) => x.id === id);
      },
-      updateFormDataOnConnect: (connection: Connection) => {
-        const { getOperatorTypeFromId, updateSwitchFormData } = get();
-        const { source, target, sourceHandle } = connection;
-        const operatorType = getOperatorTypeFromId(source);
-        if (source) {
-          switch (operatorType) {
-            case Operator.Switch: {
-              updateSwitchFormData(source, sourceHandle, target, true);
-              break;
-            }
-            default:
-              break;
-          }
-        }
-      },
      duplicateNode: (id: string, name: string) => {
-        const { getNode, addNode, generateNodeName, duplicateIterationNode } =
-          get();
+        const { getNode, addNode, generateNodeName } = get();
        const node = getNode(id);

-        if (node?.data.label === Operator.Iteration) {
-          duplicateIterationNode(id, name);
-          return;
-        }
-
        addNode({
          ...(node || {}),
          data: {
@ -257,35 +230,6 @@ const useGraphStore = create<RFState>()(
          ...generateDuplicateNode(node?.position, node?.data?.label),
        });
      },
-      duplicateIterationNode: (id: string, name: string) => {
-        const { getNode, generateNodeName, nodes } = get();
-        const node = getNode(id);
-
-        const iterationNode: RAGFlowNodeType = {
-          ...(node || {}),
-          data: {
-            ...(node?.data || { label: Operator.Iteration, form: {} }),
-            name: generateNodeName(name),
-          },
-          ...generateDuplicateNode(node?.position, node?.data?.label),
-        };
-
-        const children = nodes
-          .filter((x) => x.parentId === node?.id)
-          .map((x) => ({
-            ...(x || {}),
-            data: {
-              ...duplicateNodeForm(x?.data),
-              name: generateNodeName(x.data.name),
-            },
-            ...omit(generateDuplicateNode(x?.position, x?.data?.label), [
-              'position',
-            ]),
-            parentId: iterationNode.id,
-          }));
-
-        set({ nodes: nodes.concat(iterationNode, ...children) });
-      },
      deleteEdge: () => {
        const { edges, selectedEdgeIds } = get();
        set({
@ -295,55 +239,15 @@ const useGraphStore = create<RFState>()(
        });
      },
      deleteEdgeById: (id: string) => {
-        const {
-          edges,
-          updateNodeForm,
-          getOperatorTypeFromId,
-          updateSwitchFormData,
-        } = get();
-        const currentEdge = edges.find((x) => x.id === id);
+        const { edges } = get();

-        if (currentEdge) {
-          const { source, sourceHandle, target } = currentEdge;
-          const operatorType = getOperatorTypeFromId(source);
-          // After deleting the edge, set the corresponding field in the node's form field to undefined
-          switch (operatorType) {
-            case Operator.Relevant:
-              updateNodeForm(source, {
-                [sourceHandle as string]: undefined,
-              });
-              break;
-            // case Operator.Categorize:
-            //   if (sourceHandle)
-            //     updateNodeForm(source, undefined, [
-            //       'category_description',
-            //       sourceHandle,
-            //       'to',
-            //     ]);
-            //   break;
-            case Operator.Switch: {
-              updateSwitchFormData(source, sourceHandle, target, false);
-              break;
-            }
-            default:
-              break;
-          }
-        }
        set({
          edges: edges.filter((edge) => edge.id !== id),
        });
      },
      deleteNodeById: (id: string) => {
-        const {
-          nodes,
-          edges,
-          getOperatorTypeFromId,
-          deleteAgentDownstreamNodesById,
-        } = get();
-        if (getOperatorTypeFromId(id) === Operator.Agent) {
-          deleteAgentDownstreamNodesById(id);
-          return;
-        }
+        const { nodes, edges } = get();
+
        set({
          nodes: nodes.filter((node) => node.id !== id),
          edges: edges
@ -526,6 +430,10 @@ const useGraphStore = create<RFState>()(
          })),
        );
      },
+      hasChildNode: (nodeId) => {
+        const { edges } = get();
+        return edges.some((edge) => edge.source === nodeId);
+      },
    })),
    { name: 'graph', trace: true },
  ),