Implementation:Langgenius Dify FetchPromptTemplate

Knowledge Sources	Domains	Last Updated
Dify	LLM_Applications, Frontend, API	2026-02-12 00:00 GMT

Overview

Description

fetchPromptTemplate is a frontend service function that retrieves recommended prompt template configurations from the backend. Given the application mode, model mode, model name, and whether a knowledge base (dataset) is connected, it returns pre-built prompt templates appropriate for both chat-mode and completion-mode interactions. This allows the platform to provide sensible defaults and scaffolding for prompt design based on the current application context.

The function queries the /app/prompt-templates endpoint with the configuration parameters, and the backend returns both a ChatPromptConfig (for chat-mode models) and a CompletionPromptConfig (for completion-mode models), plus any recommended stop sequences.

Additionally, the useDebugConfigurationContext hook (defined in web/context/debug-configuration.ts, lines 31-109) provides the React context that holds the full prompt editing state during the configuration workflow. This context stores the prompt mode (simple vs. advanced), current advanced prompt content, conversation history roles, block insertion status, model configuration, dataset connections, completion parameters, and all feature toggles. It serves as the centralized state management layer for the prompt design and debug configuration UI.

Usage

Call fetchPromptTemplate when the user changes the application mode, model, or dataset connection status to retrieve appropriate prompt template recommendations. Use useDebugConfigurationContext in any React component within the configuration workflow to access or modify the current prompt state.

Code Reference

Source Location

web/service/debug.ts, lines 86-100

Signature

export const fetchPromptTemplate = ({
  appMode,
  mode,
  modelName,
  hasSetDataSet,
}: {
  appMode: AppModeEnum
  mode: ModelModeType
  modelName: string
  hasSetDataSet: boolean
}) => {
  return get<Promise<{
    chat_prompt_config: ChatPromptConfig
    completion_prompt_config: CompletionPromptConfig
    stop: []
  }>>('/app/prompt-templates', {
    params: {
      app_mode: appMode,
      model_mode: mode,
      model_name: modelName,
      has_context: hasSetDataSet,
    },
  })
}

Related: useDebugConfigurationContext (web/context/debug-configuration.ts, lines 31-109):

type IDebugConfiguration = {
  appId: string
  mode: AppModeEnum
  modelModeType: ModelModeType
  promptMode: PromptMode
  setPromptMode: (promptMode: PromptMode) => void
  isAdvancedMode: boolean
  chatPromptConfig: ChatPromptConfig
  completionPromptConfig: CompletionPromptConfig
  currentAdvancedPrompt: PromptItem | PromptItem[]
  setCurrentAdvancedPrompt: (prompt: PromptItem | PromptItem[], isUserChanged?: boolean) => void
  conversationHistoriesRole: ConversationHistoriesRole
  setConversationHistoriesRole: (role: ConversationHistoriesRole) => void
  hasSetBlockStatus: BlockStatus
  modelConfig: ModelConfig
  setModelConfig: (modelConfig: ModelConfig) => void
  completionParams: FormValue
  setCompletionParams: (completionParams: FormValue) => void
  dataSets: DataSet[]
  setDataSets: (dataSet: DataSet[]) => void
  // ... additional state properties
}

export const useDebugConfigurationContext = () => useContext(DebugConfigurationContext)

Import

import { fetchPromptTemplate } from '@/service/debug'
import { useDebugConfigurationContext } from '@/context/debug-configuration'

I/O Contract

Inputs

Parameter	Type	Required	Description
appMode	AppModeEnum	Yes	The application mode: 'completion', 'workflow', 'chat', 'advanced-chat', or 'agent-chat'.
mode	ModelModeType	Yes	The model interaction mode: 'chat', 'completion', or (unset).
modelName	string	Yes	The name of the selected model (e.g., 'gpt-3.5-turbo', 'gpt-4').
hasSetDataSet	boolean	Yes	Whether a knowledge base (dataset) has been connected to the application. When true, the returned template includes context injection placeholders.

Outputs

Field	Type	Description
chat_prompt_config	ChatPromptConfig	A prompt configuration for chat-mode models, containing an array of PromptItem objects with role ('system', 'user', 'assistant') and text fields.
completion_prompt_config	CompletionPromptConfig	A prompt configuration for completion-mode models, containing a single PromptItem (prompt text) and conversation_histories_role with user_prefix and assistant_prefix strings.
stop	[]	An array of recommended stop sequences (may be empty).

Usage Examples

Fetching a prompt template for a chat app with a connected dataset

import { fetchPromptTemplate } from '@/service/debug'
import { AppModeEnum, ModelModeType } from '@/types/app'

const templates = await fetchPromptTemplate({
  appMode: AppModeEnum.CHAT,
  mode: ModelModeType.chat,
  modelName: 'gpt-4',
  hasSetDataSet: true,
})

// Use the chat prompt config for a chat-mode model
console.log(templates.chat_prompt_config.prompt)
// [{ role: 'system', text: '...' }, { role: 'user', text: '...' }]

Accessing prompt state in a React component

import { useDebugConfigurationContext } from '@/context/debug-configuration'

function PromptEditor() {
  const {
    promptMode,
    setPromptMode,
    currentAdvancedPrompt,
    setCurrentAdvancedPrompt,
    chatPromptConfig,
    completionPromptConfig,
    modelConfig,
  } = useDebugConfigurationContext()

  // Read and modify prompt state through the context
}

Related Pages

• Principle:Langgenius_Dify_Prompt_Template_Design