Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
import type {
DocumentPipelineExecutionLogResponse,
IndexingEstimateResponse,
} from '@dify/contracts/api/console/datasets/types.gen'
import type {
PipelineGenerateResponse,
PipelineTemplateDetailResponse,
PipelineTemplateListResponse as PipelineTemplateListContract,
PipelineVariableResponse,
PostRagPipelinesByPipelineIdWorkflowsPublishedDatasourceNodesByNodeIdPreviewResponse,
RagPipelineImportCheckDependenciesResponse,
RagPipelinePublishedRunResponse,
RagPipelinePublishedWorkflowResponse,
RagPipelineStepParametersResponse,
WorkflowAppBlockingResponse,
WorkflowRunNodeExecutionResponse,
} from '@dify/contracts/api/console/rag/types.gen'
import type { Viewport } from 'reactflow'
import type { Dependency } from '@/app/components/plugins/types'
import type { Edge, Node } from '@/app/components/workflow/types'
import type { IconInfo } from '@/models/datasets'
import type {
PipelineExecutionLogResponse,
PipelineTemplate,
PipelineTemplateByIdResponse,
RAGPipelineVariable,
} from '@/models/pipeline'
import type { AppIconType } from '@/types/app'
import type { NodeTracing } from '@/types/workflow'
import { BlockEnum, SupportUploadFileTypes } from '@/app/components/workflow/types'
import { ChunkingMode } from '@/models/datasets'
import { DatasourceType, PipelineInputVarType } from '@/models/pipeline'
import { TransferMethod } from '@/types/app'
const isRecord = (value: unknown): value is Record<string, unknown> =>
typeof value === 'object' && value !== null && !Array.isArray(value)
const getString = (record: Record<string, unknown>, key: string) => {
const value = record[key]
return typeof value === 'string' ? value : undefined
}
const getNumber = (record: Record<string, unknown>, key: string) => {
const value = record[key]
return typeof value === 'number' ? value : undefined
}
const isOneOf = <T extends string>(values: Record<string, T>, value: unknown): value is T =>
typeof value === 'string' && Object.values<string>(values).includes(value)
const APP_ICON_TYPES = { image: 'image', emoji: 'emoji', link: 'link' } satisfies Record<
AppIconType,
AppIconType
>
const toIconInfo = (icon: Record<string, unknown>): IconInfo => {
const iconType = icon.icon_type
return {
icon: getString(icon, 'icon') ?? '',
icon_type: isOneOf(APP_ICON_TYPES, iconType) ? iconType : 'emoji',
icon_background: getString(icon, 'icon_background'),
icon_url: getString(icon, 'icon_url'),
}
}
// Templates exported before a chunk structure was recorded default to general text chunks.
const toChunkingMode = (chunkStructure: string) =>
isOneOf(ChunkingMode, chunkStructure) ? chunkStructure : ChunkingMode.text
// The workflow graph is stored as free-form JSON; keep the nodes and edges React Flow can place.
const isGraphNode = (value: unknown): value is Node =>
isRecord(value) &&
typeof value.id === 'string' &&
isRecord(value.data) &&
isRecord(value.position)
const isGraphEdge = (value: unknown): value is Edge =>
isRecord(value) && typeof value.source === 'string' && typeof value.target === 'string'
const toGraphNodes = (graph: Record<string, unknown>) =>
Array.isArray(graph.nodes) ? graph.nodes.filter(isGraphNode) : []
const toGraphEdges = (graph: Record<string, unknown>) =>
Array.isArray(graph.edges) ? graph.edges.filter(isGraphEdge) : []
const toViewport = (graph: Record<string, unknown>): Viewport => {
const viewport = isRecord(graph.viewport) ? graph.viewport : {}
return {
x: getNumber(viewport, 'x') ?? 0,
y: getNumber(viewport, 'y') ?? 0,
zoom: getNumber(viewport, 'zoom') ?? 1,
}
}
export const selectPipelineTemplateList = (data: PipelineTemplateListContract) => ({
pipeline_templates: data.pipeline_templates.map((template): PipelineTemplate => ({
id: template.id,
name: template.name,
description: template.description,
position: template.position,
icon: toIconInfo(template.icon),
chunk_structure: toChunkingMode(template.chunk_structure),
})),
})
export const selectPipelineTemplate = (
template: PipelineTemplateDetailResponse,
): PipelineTemplateByIdResponse => ({
id: template.id,
name: template.name,
description: template.description,
export_data: template.export_data,
created_by: template.created_by ?? '',
icon_info: toIconInfo(template.icon_info),
chunk_structure: toChunkingMode(template.chunk_structure),
graph: {
nodes: toGraphNodes(template.graph),
edges: toGraphEdges(template.graph),
viewport: toViewport(template.graph),
},
})
export const selectPublishedPipelineNodes = (workflow: RagPipelinePublishedWorkflowResponse) =>
workflow ? toGraphNodes(workflow.graph) : []
const toDefaultValue = (value: unknown) =>
typeof value === 'string' || typeof value === 'number' || typeof value === 'boolean'
? value
: undefined
const toPipelineVariable = (variable: PipelineVariableResponse): RAGPipelineVariable[] => {
if (!isOneOf(PipelineInputVarType, variable.type)) return []
return [
{
belong_to_node_id: variable.belong_to_node_id,
type: variable.type,
label: variable.label,
variable: variable.variable,
required: variable.required,
max_length: variable.max_length ?? undefined,
default_value: toDefaultValue(variable.default_value),
placeholder: variable.placeholder ?? undefined,
unit: variable.unit ?? undefined,
tooltips: variable.tooltips ?? undefined,
options: variable.options ?? undefined,
allowed_file_types: variable.allowed_file_types?.filter((type) =>
isOneOf(SupportUploadFileTypes, type),
),
allowed_file_extensions: variable.allowed_file_extensions ?? undefined,
allowed_file_upload_methods: variable.allowed_file_upload_methods?.filter((method) =>
isOneOf(TransferMethod, method),
),
},
]
}
// Variables of a type the input forms cannot render are left out.
export const selectPipelineStepParameters = (data: RagPipelineStepParametersResponse) => ({
variables: data.variables.flatMap(toPipelineVariable),
})
export const selectPipelineExecutionLog = (
log: DocumentPipelineExecutionLogResponse,
): PipelineExecutionLogResponse => ({
datasource_info: isRecord(log.datasource_info) ? log.datasource_info : {},
input_data: isRecord(log.input_data) ? log.input_data : {},
datasource_node_id: log.datasource_node_id ?? '',
datasource_type: isOneOf(DatasourceType, log.datasource_type)
? log.datasource_type
: DatasourceType.localFile,
})
export const selectPipelineDependencies = (
data: RagPipelineImportCheckDependenciesResponse,
): Dependency[] =>
(data.leaked_dependencies ?? []).map((dependency) => ({
...dependency,
// The dependency store represents an absent version with undefined.
value: { ...dependency.value, version: dependency.value.version ?? undefined },
}))
export const isPipelineProcessingRun = (
response: RagPipelinePublishedRunResponse,
): response is PipelineGenerateResponse => 'batch' in response
// Preview runs return the indexing estimate as the workflow outputs.
const isFileIndexingEstimate = (value: unknown): value is IndexingEstimateResponse =>
isRecord(value) && typeof value.total_segments === 'number' && Array.isArray(value.preview)
export const selectPreviewRunEstimate = (response: WorkflowAppBlockingResponse) =>
isFileIndexingEstimate(response.data.outputs) ? response.data.outputs : undefined
export const selectOnlineDocumentPreview = (
data: PostRagPipelinesByPipelineIdWorkflowsPublishedDatasourceNodesByNodeIdPreviewResponse,
) => ({ content: getString(data, 'content') ?? '' })
const BLOCKS: Record<string, BlockEnum> = BlockEnum
export const selectDatasourceSingleRun = (
execution: WorkflowRunNodeExecutionResponse,
): NodeTracing => {
const metadata = isRecord(execution.execution_metadata) ? execution.execution_metadata : undefined
return {
id: execution.id,
index: execution.index ?? 0,
predecessor_node_id: execution.predecessor_node_id ?? '',
node_id: execution.node_id ?? '',
node_type: isOneOf(BLOCKS, execution.node_type) ? execution.node_type : BlockEnum.DataSource,
title: execution.title ?? '',
inputs: execution.inputs,
inputs_truncated: execution.inputs_truncated ?? false,
process_data: execution.process_data,
process_data_truncated: execution.process_data_truncated ?? false,
outputs: isRecord(execution.outputs) ? execution.outputs : undefined,
outputs_truncated: execution.outputs_truncated ?? false,
status: execution.status ?? '',
error: execution.error ?? undefined,
elapsed_time: execution.elapsed_time ?? 0,
execution_metadata: metadata
? {
total_tokens: getNumber(metadata, 'total_tokens') ?? 0,
total_price: getNumber(metadata, 'total_price') ?? 0,
currency: getString(metadata, 'currency') ?? '',
}
: undefined,
// Datasource nodes never run inside an iteration or loop.
metadata: { iterator_length: 0, iterator_index: 0, loop_length: 0, loop_index: 0 },
created_at: execution.created_at ?? 0,
created_by: execution.created_by_account ?? { id: '', name: '', email: '' },
finished_at: execution.finished_at ?? 0,
extras: execution.extras,
retry_index: execution.retry_index ?? undefined,
}
}