flow-uml.txt

@startuml

participant "index.html" as Client
participant "OpenAI" as API

Client -> Client: User clicks "Start Session"
== Session Start ==
Client -> Client: Acquire local audio (getUserMedia())
Client -> Client: Create RTCPeerConnection
Client -> Client: Create DataChannel
Client -> API: addTrack(): attach local audio to RTCPeerConnection
Client -> Client: setLocalDescription()
Client -> API: HTTP POST (localDescription.sdp, API key)
API -> Client: SDP Answer
Client -> Client: setRemoteDescription(answer)
Client -> API: WebRTC: local audio stream
API -> Client: WebRTC: remote audio stream
Client -> Client: ontrack() attaches remote audio stream to <audio>
Client <-> API: DataChannel connected
Client -> Client: DataChannel onopen event fires
Client -> API: DataChannel: "session.update" with session params:\n  instructions, voice, functions, temperature
Client -> API: DataChannel: "response.create" with start instructions
API --> Client: Welcome audio according to start instructions

== During Session ==
Client --> API: User audio requests to agent
API --> Client: Agent audio responses
API -> Client: DataChannel messages:\n  audio buffer changes, transcripts, function calls
Client -> API: DataChannel system message updates

== Session End ==
Client --> API: User asks to end session
API -> Client: "end_session" function call response
Client -> API: DataChannel: "response.create" with end instructions
API --> Client: Goodbye audio
API -> Client: DataChannel: "output_audio_buffer.stopped"
Client -> Client: Close RTCPeerConnection after audio done
@endumlml