sequenceDiagram
    participant CS as ChannelService (e.g., BaileysStartupService)
    participant OS as OpenaiService
    participant ConfS as ConfigService
    participant Axios as HTTP Client (axios)
    participant OAI as OpenAI API

    CS->>OS: speechToText(msgOrBuffer, updateMediaMessage?)
    activate OS
        OS->>OS: getAudioBufferFromMsg(msgOrBuffer, updateMediaMessage?)
        OS-->>OS: audioBuffer
        OS->>OS: processAudioTranscription(audioBuffer)
        activate OS
            alt client already initialized with API key
                note right of OS: Use existing API key from client
            else
                OS->>ConfS: get("OPENAI").API_KEY
                ConfS-->>OS: apiKey
            end
            OS->>Axios: POST /v1/audio/transcriptions (formData with audioBuffer, model, lang)
            activate Axios
                Axios->>OAI: Transcribe Audio Request
                OAI-->>Axios: Transcription Response
            Axios-->>OS: { text: "transcribed_text" }
            deactivate Axios
        OS-->>OS: "transcribed_text"
        deactivate OS
    OS-->>CS: "transcribed_text"
    deactivate OS

classDiagram
    class BaseChatbotDto {
      +enabled: boolean
      +description: string
      +expire: number
      +keywordFinish: string[]
      +delayMessage: number
      +unknownMessage: string
      +listeningFromMe: boolean
      +stopBotFromMe: boolean
      +keepOpen: boolean
      +debounceTime: number
      +triggerType: TriggerType
      +triggerOperator: TriggerOperator
      +triggerValue: string
      +ignoreJids: string[]
      +splitMessages: boolean
      +timePerChar: number
    }
    class OpenaiDto {
      +openaiCredsId: string
      +botType: string
      +assistantId: string
      +functionUrl: string
      +model: string
      +systemMessages: string[]
      +assistantMessages: string[]
      +userMessages: string[]
      +maxTokens: number
    }
    class DifyDto {
      +botType: $Enums.DifyBotType
      +apiUrl: string
      +apiKey: string
    }
    BaseChatbotDto <|-- OpenaiDto
    BaseChatbotDto <|-- DifyDto

    class BaseChatbotSettingDto {
      +expire: number
      +keywordFinish: string[]
      +delayMessage: number
      +unknownMessage: string
      +listeningFromMe: boolean
      +stopBotFromMe: boolean
      +keepOpen: boolean
      +debounceTime: number
      +ignoreJids: string[]
      +splitMessages: boolean
      +timePerChar: number
      +fallbackId: string
    }
    class OpenaiSettingDto {
      +openaiCredsId: string
      +openaiIdFallback: string
      +speechToText: boolean
    }
    BaseChatbotSettingDto <|-- OpenaiSettingDto

    class BaseChatbotService {
        <<abstract>>
        #prismaRepository: PrismaRepository
        #configService: ConfigService
        #logger: Logger
        +constructor(waMonitor, prismaRepository, loggerName, configService)
        #getBotType(): string
        +createNewSession(instance, data, type)
        #initClient(apiKey: string)
        +sendMessageWhatsApp(instance, remoteJid, message, settings)
        +speechToText(msgOrBuffer, updateMediaMessage) : Promise~string~
        #getAudioBufferFromMsg(msg, updateMediaMessage) : Promise~Buffer~
        #processAudioTranscription(audioBuffer: Buffer) : Promise~string~
    }
    class OpenaiService {
        #client: OpenAI
        +constructor(waMonitor, prismaRepository, configService)
        #getBotType(): string
        +createNewSession(instance, data)
        #initClient(apiKey: string) : OpenAI
        +process(instance, remoteJid, bot, session, settings, content, pushName, msg) : Promise~void~
        #sendMessageToBot(instance, session, settings, bot, remoteJid, pushName, content, msg) : Promise~void~
        #processAssistantMessage(instance, session, openaiBot, remoteJid, pushName, fromMe, content, msg) : Promise~string~
        #processChatCompletionMessage(instance, openaiBot, remoteJid, content, msg) : Promise~string~
        +speechToText(msgOrBuffer, updateMediaMessage) : Promise~string~
    }
    BaseChatbotService <|-- OpenaiService

    class BaseChatbotController {
        <<abstract>>
        #service: BaseChatbotService
        #prismaRepository: PrismaRepository
        #integrationName: string
        +constructor(service, prismaRepository, waMonitor, integrationName)
        #getFallbackBotId(settings: any): string
        #getIntegrationType(): string
        #getAdditionalBotData(data: BaseChatbotDto): Record~string, any~
        #getAdditionalUpdateFields(data: BaseChatbotDto): Record~string, any~
        +createBot(instance, data: BaseChatbotDto)
        +updateBot(instance, botId, data: BaseChatbotDto)
        +emit(data: any)
        #processBot(instance, remoteJid, bot, session, settings, content, pushName, msg) : Promise~void~
    }
    class OpenaiController {
        #openaiService: OpenaiService
        +constructor(openaiService, prismaRepository, waMonitor)
        #getIntegrationType(): string
        #getAdditionalBotData(data: OpenaiDto): Record~string, any~
        #processBot(instance, remoteJid, bot, session, settings, content, pushName, msg) : Promise~void~
        +createOpenaiCreds(instance, data: OpenaiCredsDto)
        +getModels(instance)
    }
    class DifyController {
        #difyService: DifyService
        +constructor(difyService, prismaRepository, waMonitor)
        #getIntegrationType(): string
        #getAdditionalBotData(data: DifyDto): Record~string, any~
        #processBot(instance, remoteJid, bot, session, settings, content, pushName, msg) : Promise~void~
    }
    BaseChatbotController <|-- OpenaiController
    BaseChatbotController <|-- DifyController

    class ChannelStartupService {
        +openaiService: OpenaiService
    }
    class BaileysStartupService {
        +onMessage(received)
    }
    ChannelStartupService <|-- BaileysStartupService
    BaileysStartupService ..> OpenaiService : uses speechToText()