Implementation:FlowiseAI Flowise SendMessageAndGetPrediction

Attribute	Value
Source Repository	FlowiseAI/Flowise
Source File	packages/ui/src/api/prediction.js
Domain	Chatflow_Creation
Workflow	Chatflow_Creation
Last Updated	2026-02-12

Overview

Concrete API client functions for sending prediction requests to a Flowise chatflow endpoint. The module exports three functions covering internal non-streaming, internal streaming, and public prediction endpoints. These functions are the primary interface for executing chatflow graphs with user messages and receiving AI-generated responses.

Code Reference

Source Location

• File: packages/ui/src/api/prediction.js
• Lines: L3-5

Signature

const sendMessageAndGetPrediction = (id, input) =>
    client.post(`/internal-prediction/${id}`, input)

const sendMessageAndStreamPrediction = (id, input) =>
    client.post(`/internal-prediction/stream/${id}`, input)

const sendMessageAndGetPredictionPublic = (id, input) =>
    client.post(`/prediction/${id}`, input)

Import

import predictionApi from '@/api/prediction'

I/O Contract

Inputs

Parameter	Type	Required	Description
id	string	Yes	The chatflow ID identifying which flow graph to execute
input	object	Yes	The prediction request payload
input.question	string	Yes	The user's message text to process
input.chatId	string	No	Conversation session ID for multi-turn conversations; omit for the first message
input.uploads	Upload[]	No	Array of file uploads (images, documents) to include with the message
input.overrideConfig	object	No	Runtime configuration overrides for node parameters
input.streaming	boolean	No	Whether to request streaming mode (only relevant for certain endpoints)

Outputs

Field	Type	Description
data.text	string	The generated response text from the chatflow
data.chatMessageId	string	Unique identifier for this message exchange
data.sourceDocuments	Document[]	Retrieved source documents (for RAG-based chatflows)
data.usedTools	ToolUsage[]	Tools that were invoked during processing
data.agentReasoning	Reasoning[]	Agent's step-by-step reasoning trace
data.fileAnnotations	Annotation[]	File references embedded in the response
data.chatId	string	The conversation session ID (assigned by server if not provided)

Implementation Details

Endpoint Variants

Function	HTTP Method	Endpoint	Use Case
sendMessageAndGetPrediction	POST	/api/v1/internal-prediction/{id}	Internal UI testing, non-streaming mode
sendMessageAndStreamPrediction	POST	/api/v1/internal-prediction/stream/{id}	Internal UI testing, streaming mode (SSE)
sendMessageAndGetPredictionPublic	POST	/api/v1/prediction/{id}	Public API access for deployed chatflows

The internal endpoints include the x-request-from: internal header (set by the Axios client in client.js), which distinguishes internal test requests from external API calls on the server side.

Streaming Mode Decision

The decision between streaming and non-streaming is made by querying the chatflow's streaming capability before the first message. The chatflows.js module provides:

const getIsChatflowStreaming = (id) => client.get(`/chatflows-streaming/${id}`)

The chat UI component (ChatMessage) checks this and routes to the appropriate prediction function accordingly.

Session Persistence

The chat ID is persisted in localStorage for conversation continuity:

// From packages/ui/src/utils/genericHelper.js
export const setLocalStorageChatflow = (chatflowid, chatId, saveObj = {}) => {
    const obj = { ...saveObj }
    if (chatId) obj.chatId = chatId
    localStorage.setItem(`${chatflowid}_INTERNAL`, JSON.stringify(obj))
}

export const getLocalStorageChatflow = (chatflowid) => {
    const chatDetails = localStorage.getItem(`${chatflowid}_INTERNAL`)
    if (!chatDetails) return {}
    return JSON.parse(chatDetails)
}

Usage Examples

Non-streaming prediction

import predictionApi from '@/api/prediction'

const chatflowId = 'abc-123'
const response = await predictionApi.sendMessageAndGetPrediction(chatflowId, {
    question: 'What is the capital of France?',
    chatId: 'session-456'
})

const answer = response.data.text           // "The capital of France is Paris."
const sources = response.data.sourceDocuments // [...] if RAG is configured

Streaming prediction

import predictionApi from '@/api/prediction'

const chatflowId = 'abc-123'
const response = await predictionApi.sendMessageAndStreamPrediction(chatflowId, {
    question: 'Explain quantum computing in simple terms.',
    chatId: 'session-456'
})

// The response is processed as an SSE stream by the chat UI component
// Partial tokens are displayed as they arrive

Public endpoint for deployed chatflows

import predictionApi from '@/api/prediction'

const chatflowId = 'deployed-flow-789'
const response = await predictionApi.sendMessageAndGetPredictionPublic(chatflowId, {
    question: 'Help me track my order #12345',
    overrideConfig: {
        temperature: 0.5
    }
})

const answer = response.data.text
const tools = response.data.usedTools  // Tools used during processing

First message in a new conversation (no chatId)

import predictionApi from '@/api/prediction'
import { setLocalStorageChatflow } from '@/utils/genericHelper'

const chatflowId = 'abc-123'

// First message -- no chatId
const response = await predictionApi.sendMessageAndGetPrediction(chatflowId, {
    question: 'Hello, I need help with my account.'
})

// Save the server-assigned chatId for subsequent messages
const chatId = response.data.chatId
setLocalStorageChatflow(chatflowId, chatId)

// Subsequent messages include the chatId
const followUp = await predictionApi.sendMessageAndGetPrediction(chatflowId, {
    question: 'Can you look up my billing history?',
    chatId: chatId
})

Related Pages

• Principle:FlowiseAI_Flowise_Chat_Prediction
• Heuristic:FlowiseAI_Flowise_Document_Loader_Bypass_Optimization