Use existing sse streamer to stream tts audio before sse client is removed
This commit is contained in:
parent
2247646182
commit
2b5554aafe
|
|
@ -268,4 +268,30 @@ export class SSEStreamer implements IServerSideEventStreamer {
|
|||
client.response.write('message:\ndata:' + JSON.stringify(clientResponse) + '\n\n')
|
||||
}
|
||||
}
|
||||
|
||||
streamTTSDataEvent(chatId: string, audioChunk: string): void {
|
||||
const client = this.clients[chatId]
|
||||
console.log('clients', this.clients)
|
||||
console.log('client', client)
|
||||
if (client) {
|
||||
const clientResponse = {
|
||||
event: 'tts_data',
|
||||
data: audioChunk
|
||||
}
|
||||
client.response.write('message:\ndata:' + JSON.stringify(clientResponse) + '\n\n')
|
||||
}
|
||||
}
|
||||
|
||||
streamTTSEndEvent(chatId: string): void {
|
||||
const client = this.clients[chatId]
|
||||
console.log('clients', this.clients)
|
||||
console.log('client', client)
|
||||
if (client) {
|
||||
const clientResponse = {
|
||||
event: 'tts_end',
|
||||
data: {}
|
||||
}
|
||||
client.response.write('message:\ndata:' + JSON.stringify(clientResponse) + '\n\n')
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -11,8 +11,7 @@ import {
|
|||
IMessage,
|
||||
IServerSideEventStreamer,
|
||||
convertChatHistoryToText,
|
||||
generateFollowUpPrompts,
|
||||
convertTextToSpeech
|
||||
generateFollowUpPrompts
|
||||
} from 'flowise-components'
|
||||
import {
|
||||
IncomingAgentflowInput,
|
||||
|
|
@ -58,6 +57,7 @@ import { ChatMessage } from '../database/entities/ChatMessage'
|
|||
import { Telemetry } from './telemetry'
|
||||
import { getWorkspaceSearchOptions } from '../enterprise/utils/ControllerServiceUtils'
|
||||
import { UsageCacheManager } from '../UsageCacheManager'
|
||||
import { generateTTSForResponseStream, shouldAutoPlayTTS } from './buildChatflow'
|
||||
|
||||
interface IWaitingNode {
|
||||
nodeId: string
|
||||
|
|
@ -136,59 +136,6 @@ interface IExecuteNodeParams {
|
|||
subscriptionId: string
|
||||
}
|
||||
|
||||
// Helper function to check if auto-play TTS is enabled
|
||||
const shouldAutoPlayTTS = (textToSpeechConfig: string | undefined | null): boolean => {
|
||||
if (!textToSpeechConfig) return false
|
||||
try {
|
||||
const config = typeof textToSpeechConfig === 'string' ? JSON.parse(textToSpeechConfig) : textToSpeechConfig
|
||||
// Check each provider to see if any has autoPlay enabled and status true
|
||||
for (const providerKey in config) {
|
||||
const provider = config[providerKey]
|
||||
if (provider && provider.status === true && provider.autoPlay === true) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
} catch (error) {
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
// Helper function to generate TTS for response
|
||||
const generateTTSForResponse = async (
|
||||
responseText: string,
|
||||
textToSpeechConfig: string | undefined,
|
||||
options: ICommonObject
|
||||
): Promise<Buffer | null> => {
|
||||
try {
|
||||
if (!textToSpeechConfig) return null
|
||||
const config = typeof textToSpeechConfig === 'string' ? JSON.parse(textToSpeechConfig) : textToSpeechConfig
|
||||
|
||||
// Find the active provider configuration
|
||||
let activeProviderConfig = null
|
||||
for (const providerKey in config) {
|
||||
const provider = config[providerKey]
|
||||
if (provider && provider.status === true) {
|
||||
activeProviderConfig = {
|
||||
name: providerKey,
|
||||
credentialId: provider.credentialId,
|
||||
voice: provider.voice,
|
||||
model: provider.model
|
||||
}
|
||||
break
|
||||
}
|
||||
}
|
||||
|
||||
if (!activeProviderConfig) return null
|
||||
|
||||
const audioBuffer = await convertTextToSpeech(responseText, activeProviderConfig, options)
|
||||
return audioBuffer
|
||||
} catch (error) {
|
||||
logger.error(`[server]: TTS generation failed: ${getErrorMessage(error)}`)
|
||||
return null
|
||||
}
|
||||
}
|
||||
|
||||
interface IExecuteAgentFlowParams extends Omit<IExecuteFlowParams, 'incomingInput'> {
|
||||
incomingInput: IncomingAgentflowInput
|
||||
}
|
||||
|
|
@ -2092,7 +2039,6 @@ export const executeAgentFlow = async ({
|
|||
|
||||
if (sessionId) result.sessionId = sessionId
|
||||
|
||||
/*** Auto-play TTS Logic ***/
|
||||
if (shouldAutoPlayTTS(chatflow.textToSpeech) && result.text) {
|
||||
const options = {
|
||||
orgId,
|
||||
|
|
@ -2102,14 +2048,8 @@ export const executeAgentFlow = async ({
|
|||
databaseEntities
|
||||
}
|
||||
|
||||
const audioBuffer = await generateTTSForResponse(result.text, chatflow.textToSpeech, options)
|
||||
if (audioBuffer) {
|
||||
const audioBase64 = audioBuffer.toString('base64')
|
||||
|
||||
// Agent flows are always streamed, so send audio via SSE
|
||||
if (sseStreamer) {
|
||||
sseStreamer.streamAudioEvent(chatId, audioBase64)
|
||||
}
|
||||
await generateTTSForResponseStream(result.text, chatflow.textToSpeech, options, chatId, sseStreamer)
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -7,6 +7,7 @@ import {
|
|||
IFileUpload,
|
||||
convertSpeechToText,
|
||||
convertTextToSpeech,
|
||||
convertTextToSpeechStream,
|
||||
ICommonObject,
|
||||
addSingleFileToStorage,
|
||||
generateFollowUpPrompts,
|
||||
|
|
@ -17,7 +18,8 @@ import {
|
|||
getFileFromUpload,
|
||||
removeSpecificFileFromUpload,
|
||||
EvaluationRunner,
|
||||
handleEscapeCharacters
|
||||
handleEscapeCharacters,
|
||||
IServerSideEventStreamer
|
||||
} from 'flowise-components'
|
||||
import { StatusCodes } from 'http-status-codes'
|
||||
import {
|
||||
|
|
@ -71,12 +73,10 @@ import { executeAgentFlow } from './buildAgentflow'
|
|||
import { Workspace } from '../enterprise/database/entities/workspace.entity'
|
||||
import { Organization } from '../enterprise/database/entities/organization.entity'
|
||||
|
||||
// Helper function to check if auto-play TTS is enabled
|
||||
const shouldAutoPlayTTS = (textToSpeechConfig: string | undefined | null): boolean => {
|
||||
if (!textToSpeechConfig) return false
|
||||
try {
|
||||
const config = typeof textToSpeechConfig === 'string' ? JSON.parse(textToSpeechConfig) : textToSpeechConfig
|
||||
// Check each provider to see if any has autoPlay enabled and status true
|
||||
for (const providerKey in config) {
|
||||
const provider = config[providerKey]
|
||||
if (provider && provider.status === true && provider.autoPlay === true) {
|
||||
|
|
@ -85,21 +85,22 @@ const shouldAutoPlayTTS = (textToSpeechConfig: string | undefined | null): boole
|
|||
}
|
||||
return false
|
||||
} catch (error) {
|
||||
logger.error(`Error parsing textToSpeechConfig: ${getErrorMessage(error)}`)
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
// Helper function to generate TTS for response
|
||||
const generateTTSForResponse = async (
|
||||
const generateTTSForResponseStream = async (
|
||||
responseText: string,
|
||||
textToSpeechConfig: string | undefined,
|
||||
options: ICommonObject
|
||||
): Promise<Buffer | null> => {
|
||||
options: ICommonObject,
|
||||
chatId: string,
|
||||
sseStreamer: IServerSideEventStreamer
|
||||
): Promise<void> => {
|
||||
try {
|
||||
if (!textToSpeechConfig) return null
|
||||
if (!textToSpeechConfig) return
|
||||
const config = typeof textToSpeechConfig === 'string' ? JSON.parse(textToSpeechConfig) : textToSpeechConfig
|
||||
|
||||
// Find the active provider configuration
|
||||
let activeProviderConfig = null
|
||||
for (const providerKey in config) {
|
||||
const provider = config[providerKey]
|
||||
|
|
@ -114,13 +115,24 @@ const generateTTSForResponse = async (
|
|||
}
|
||||
}
|
||||
|
||||
if (!activeProviderConfig) return null
|
||||
if (!activeProviderConfig) return
|
||||
|
||||
const audioBuffer = await convertTextToSpeech(responseText, activeProviderConfig, options)
|
||||
return audioBuffer
|
||||
await convertTextToSpeechStream(
|
||||
responseText,
|
||||
activeProviderConfig,
|
||||
options,
|
||||
(chunk: Buffer) => {
|
||||
const audioBase64 = chunk.toString('base64')
|
||||
logger.info(`Received TTS chunk: ${audioBase64}`)
|
||||
sseStreamer.streamTTSDataEvent(chatId, audioBase64)
|
||||
},
|
||||
() => {
|
||||
sseStreamer.streamTTSEndEvent(chatId)
|
||||
}
|
||||
)
|
||||
} catch (error) {
|
||||
logger.error(`[server]: TTS generation failed: ${getErrorMessage(error)}`)
|
||||
return null
|
||||
logger.error(`[server]: TTS streaming failed: ${getErrorMessage(error)}`)
|
||||
sseStreamer.streamTTSEndEvent(chatId)
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -880,8 +892,6 @@ export const executeFlow = async ({
|
|||
if (Object.keys(setVariableNodesOutput).length) result.flowVariables = setVariableNodesOutput
|
||||
|
||||
if (shouldAutoPlayTTS(chatflow.textToSpeech) && result.text) {
|
||||
logger.info('[server]: Generating TTS for response')
|
||||
logger.info(`[server/executeFlow]: TTS config: ${JSON.stringify(chatflow.textToSpeech)}`)
|
||||
const options = {
|
||||
orgId,
|
||||
chatflowid,
|
||||
|
|
@ -890,15 +900,10 @@ export const executeFlow = async ({
|
|||
databaseEntities
|
||||
}
|
||||
|
||||
const audioBuffer = await generateTTSForResponse(result.text, chatflow.textToSpeech, options)
|
||||
if (audioBuffer) {
|
||||
const audioBase64 = audioBuffer.toString('base64')
|
||||
|
||||
if (streaming && sseStreamer) {
|
||||
sseStreamer.streamAudioEvent(chatId, audioBase64)
|
||||
} else {
|
||||
result.audioData = audioBase64
|
||||
}
|
||||
await generateTTSForResponseStream(result.text, chatflow.textToSpeech, options, chatId, sseStreamer)
|
||||
} else if (sseStreamer) {
|
||||
await generateTTSForResponseStream(result.text, chatflow.textToSpeech, options, chatId, sseStreamer)
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -1129,3 +1134,5 @@ const incrementFailedMetricCounter = (metricsProvider: IMetricsProvider, isInter
|
|||
)
|
||||
}
|
||||
}
|
||||
|
||||
export { shouldAutoPlayTTS, generateTTSForResponseStream }
|
||||
|
|
|
|||
Loading…
Reference in New Issue