mirrors/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-04-29 19:03:02 +03:00
Code Issues Packages Projects Releases Wiki Activity

[stack 2/4] Align main realtime v2 wire and runtime flow (#14830)

Browse Source 
## Stack Position
2/4. Built on top of #14828.

## Base
- #14828

## Unblocks
- #14829
- #14827

## Scope
- Port the realtime v2 wire parsing, session, app-server, and
conversation runtime behavior onto the split websocket-method base.
- Branch runtime behavior directly on the current realtime session kind
instead of parser-derived flow flags.
- Keep regression coverage in the existing e2e suites.

---------

Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
Ahmed Ibrahim
2026-03-16 21:38:07 -07:00
committed by GitHub
parent 1d85fe79ed
commit fbd7f9b986
28 changed files with 807 additions and 140 deletions
Download Patch File Download Diff File Expand all files Collapse all files

166
codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs
View File

@@ -272,12 +272,12 @@ impl RealtimeWebsocketConnection {
impl RealtimeWebsocketWriter {
pub async fn send_audio_frame(&self, frame: RealtimeAudioFrame) -> Result<(), ApiError> {
self.send_json(RealtimeOutboundMessage::InputAudioBufferAppend { audio: frame.data })
self.send_json(&RealtimeOutboundMessage::InputAudioBufferAppend { audio: frame.data })
.await
}
pub async fn send_conversation_item_create(&self, text: String) -> Result<(), ApiError> {
self.send_json(conversation_item_create_message(self.event_parser, text))
self.send_json(&conversation_item_create_message(self.event_parser, text))
.await
}
@@ -286,7 +286,7 @@ impl RealtimeWebsocketWriter {
handoff_id: String,
output_text: String,
) -> Result<(), ApiError> {
self.send_json(conversation_handoff_append_message(
self.send_json(&conversation_handoff_append_message(
self.event_parser,
handoff_id,
output_text,
@@ -294,6 +294,11 @@ impl RealtimeWebsocketWriter {
.await
}
pub async fn send_response_create(&self) -> Result<(), ApiError> {
self.send_json(&RealtimeOutboundMessage::ResponseCreate)
.await
}
pub async fn send_session_update(
&self,
instructions: String,
@@ -301,7 +306,7 @@ impl RealtimeWebsocketWriter {
) -> Result<(), ApiError> {
let session_mode = normalized_session_mode(self.event_parser, session_mode);
let session = session_update_session(self.event_parser, instructions, session_mode);
self.send_json(RealtimeOutboundMessage::SessionUpdate { session })
self.send_json(&RealtimeOutboundMessage::SessionUpdate { session })
.await
}
@@ -319,11 +324,14 @@ impl RealtimeWebsocketWriter {
Ok(())
}
async fn send_json(&self, message: RealtimeOutboundMessage) -> Result<(), ApiError> {
let payload = serde_json::to_string(&message)
async fn send_json(&self, message: &RealtimeOutboundMessage) -> Result<(), ApiError> {
let payload = serde_json::to_string(message)
.map_err(|err| ApiError::Stream(format!("failed to encode realtime request: {err}")))?;
debug!(?message, "realtime websocket request");
self.send_payload(payload).await
}
pub async fn send_payload(&self, payload: String) -> Result<(), ApiError> {
if self.is_closed.load(Ordering::SeqCst) {
return Err(ApiError::Stream(
"realtime websocket connection is closed".to_string(),
@@ -392,6 +400,7 @@ impl RealtimeWebsocketEvents {
async fn update_active_transcript(&self, event: &mut RealtimeEvent) {
let mut active_transcript = self.active_transcript.lock().await;
match event {
RealtimeEvent::InputAudioSpeechStarted(_) => {}
RealtimeEvent::InputTranscriptDelta(RealtimeTranscriptDelta { delta }) => {
append_transcript_delta(&mut active_transcript.entries, "user", delta);
}
@@ -403,6 +412,7 @@ impl RealtimeWebsocketEvents {
}
RealtimeEvent::SessionUpdated { .. }
| RealtimeEvent::AudioOut(_)
| RealtimeEvent::ResponseCancelled(_)
| RealtimeEvent::ConversationItemAdded(_)
| RealtimeEvent::ConversationItemDone { .. }
| RealtimeEvent::Error(_) => {}
@@ -616,6 +626,8 @@ mod tests {
use crate::endpoint::realtime_websocket::protocol::RealtimeHandoffRequested;
use crate::endpoint::realtime_websocket::protocol::RealtimeTranscriptDelta;
use crate::endpoint::realtime_websocket::protocol::RealtimeTranscriptEntry;
use codex_protocol::protocol::RealtimeInputAudioSpeechStarted;
use codex_protocol::protocol::RealtimeResponseCancelled;
use http::HeaderValue;
use pretty_assertions::assert_eq;
use serde_json::Value;
@@ -660,6 +672,7 @@ mod tests {
sample_rate: 48000,
num_channels: 1,
samples_per_channel: Some(960),
item_id: None,
}))
);
}
@@ -809,10 +822,112 @@ mod tests {
sample_rate: 24_000,
num_channels: 1,
samples_per_channel: None,
item_id: None,
}))
);
}
#[test]
fn parse_realtime_v2_response_audio_delta_with_item_id() {
let payload = json!({
"type": "response.audio.delta",
"delta": "AQID",
"item_id": "item_audio_1"
})
.to_string();
assert_eq!(
parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
Some(RealtimeEvent::AudioOut(RealtimeAudioFrame {
data: "AQID".to_string(),
sample_rate: 24_000,
num_channels: 1,
samples_per_channel: None,
item_id: Some("item_audio_1".to_string()),
}))
);
}
#[test]
fn parse_realtime_v2_speech_started_event() {
let payload = json!({
"type": "input_audio_buffer.speech_started",
"item_id": "item_input_1"
})
.to_string();
assert_eq!(
parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
Some(RealtimeEvent::InputAudioSpeechStarted(
RealtimeInputAudioSpeechStarted {
item_id: Some("item_input_1".to_string()),
}
))
);
}
#[test]
fn parse_realtime_v2_response_cancelled_event() {
let payload = json!({
"type": "response.cancelled",
"response": {"id": "resp_cancelled_1"}
})
.to_string();
assert_eq!(
parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
Some(RealtimeEvent::ResponseCancelled(
RealtimeResponseCancelled {
response_id: Some("resp_cancelled_1".to_string()),
}
))
);
}
#[test]
fn parse_realtime_v2_response_done_handoff_event() {
let payload = json!({
"type": "response.done",
"response": {
"output": [{
"id": "item_123",
"type": "function_call",
"name": "codex",
"call_id": "call_123",
"arguments": "{\"prompt\":\"delegate from done\"}"
}]
}
})
.to_string();
assert_eq!(
parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
Some(RealtimeEvent::HandoffRequested(RealtimeHandoffRequested {
handoff_id: "call_123".to_string(),
item_id: "item_123".to_string(),
input_transcript: "delegate from done".to_string(),
active_transcript: Vec::new(),
}))
);
}
#[test]
fn parse_realtime_v2_response_created_event() {
let payload = json!({
"type": "response.created",
"response": {"id": "resp_created_1"}
})
.to_string();
assert_eq!(
parse_realtime_event(payload.as_str(), RealtimeEventParser::RealtimeV2),
Some(RealtimeEvent::ConversationItemAdded(json!({
"type": "response.created",
"response": {"id": "resp_created_1"}
})))
);
}
#[test]
fn merge_request_headers_matches_http_precedence() {
let mut provider_headers = HeaderMap::new();
@@ -1169,6 +1284,7 @@ mod tests {
sample_rate: 48000,
num_channels: 1,
samples_per_channel: Some(960),
item_id: None,
})
.await
.expect("send audio");
@@ -1196,6 +1312,7 @@ mod tests {
sample_rate: 48000,
num_channels: 1,
samples_per_channel: None,
item_id: None,
})
);
@@ -1285,9 +1402,38 @@ mod tests {
first_json["session"]["type"],
Value::String("realtime".to_string())
);
assert_eq!(first_json["session"]["output_modalities"], json!(["audio"]));
assert_eq!(
first_json["session"]["audio"]["input"]["format"],
json!({
"type": "audio/pcm",
"rate": 24_000,
})
);
assert_eq!(
first_json["session"]["audio"]["input"]["noise_reduction"],
json!({
"type": "near_field",
})
);
assert_eq!(
first_json["session"]["audio"]["input"]["turn_detection"],
json!({
"type": "server_vad",
"interrupt_response": true,
"create_response": true,
})
);
assert_eq!(
first_json["session"]["audio"]["output"]["format"],
json!({
"type": "audio/pcm",
"rate": 24_000,
})
);
assert_eq!(
first_json["session"]["audio"]["output"]["voice"],
Value::String("alloy".to_string())
Value::String("marin".to_string())
);
assert_eq!(
first_json["session"]["tools"][0]["type"],
@@ -1301,6 +1447,10 @@ mod tests {
first_json["session"]["tools"][0]["parameters"]["required"],
json!(["prompt"])
);
assert_eq!(
first_json["session"]["tool_choice"],
Value::String("auto".to_string())
);
ws.send(Message::Text(
json!({
@@ -1511,6 +1661,7 @@ mod tests {
sample_rate: 24_000,
num_channels: 1,
samples_per_channel: Some(480),
item_id: None,
})
.await
.expect("send audio");
@@ -1690,6 +1841,7 @@ mod tests {
sample_rate: 48000,
num_channels: 1,
samples_per_channel: Some(960),
item_id: None,
}),
)
.await

1
codex-rs/codex-api/src/endpoint/realtime_websocket/methods_common.rs
View File

@@ -12,7 +12,6 @@ use crate::endpoint::realtime_websocket::protocol::RealtimeSessionMode;
use crate::endpoint::realtime_websocket::protocol::SessionUpdateSession;
pub(super) const REALTIME_AUDIO_SAMPLE_RATE: u32 = 24_000;
pub(super) const REALTIME_AUDIO_FORMAT: &str = "audio/pcm";
pub(super) fn normalized_session_mode(
event_parser: RealtimeEventParser,

23
codex-rs/codex-api/src/endpoint/realtime_websocket/methods_v1.rs
View File

@@ -1,25 +1,27 @@
use crate::endpoint::realtime_websocket::methods_common::REALTIME_AUDIO_FORMAT;
use crate::endpoint::realtime_websocket::methods_common::REALTIME_AUDIO_SAMPLE_RATE;
use crate::endpoint::realtime_websocket::protocol::AudioFormatType;
use crate::endpoint::realtime_websocket::protocol::ConversationContentType;
use crate::endpoint::realtime_websocket::protocol::ConversationItemContent;
use crate::endpoint::realtime_websocket::protocol::ConversationItemPayload;
use crate::endpoint::realtime_websocket::protocol::ConversationItemType;
use crate::endpoint::realtime_websocket::protocol::ConversationMessageItem;
use crate::endpoint::realtime_websocket::protocol::ConversationRole;
use crate::endpoint::realtime_websocket::protocol::RealtimeOutboundMessage;
use crate::endpoint::realtime_websocket::protocol::SessionAudio;
use crate::endpoint::realtime_websocket::protocol::SessionAudioFormat;
use crate::endpoint::realtime_websocket::protocol::SessionAudioInput;
use crate::endpoint::realtime_websocket::protocol::SessionAudioOutput;
use crate::endpoint::realtime_websocket::protocol::SessionAudioVoice;
use crate::endpoint::realtime_websocket::protocol::SessionType;
use crate::endpoint::realtime_websocket::protocol::SessionUpdateSession;
const REALTIME_V1_SESSION_TYPE: &str = "quicksilver";
pub(super) fn conversation_item_create_message(text: String) -> RealtimeOutboundMessage {
RealtimeOutboundMessage::ConversationItemCreate {
item: ConversationItemPayload::Message(ConversationMessageItem {
kind: "message".to_string(),
role: "user".to_string(),
r#type: ConversationItemType::Message,
role: ConversationRole::User,
content: vec![ConversationItemContent {
kind: "text".to_string(),
r#type: ConversationContentType::Text,
text,
}],
}),
@@ -38,20 +40,25 @@ pub(super) fn conversation_handoff_append_message(
pub(super) fn session_update_session(instructions: String) -> SessionUpdateSession {
SessionUpdateSession {
kind: REALTIME_V1_SESSION_TYPE.to_string(),
r#type: SessionType::Quicksilver,
instructions: Some(instructions),
output_modalities: None,
audio: SessionAudio {
input: SessionAudioInput {
format: SessionAudioFormat {
kind: REALTIME_AUDIO_FORMAT.to_string(),
r#type: AudioFormatType::AudioPcm,
rate: REALTIME_AUDIO_SAMPLE_RATE,
},
noise_reduction: None,
turn_detection: None,
},
output: Some(SessionAudioOutput {
format: None,
voice: SessionAudioVoice::Fathom,
}),
},
tools: None,
tool_choice: None,
}
}

57
codex-rs/codex-api/src/endpoint/realtime_websocket/methods_v2.rs
View File

@@ -1,31 +1,42 @@
use crate::endpoint::realtime_websocket::methods_common::REALTIME_AUDIO_FORMAT;
use crate::endpoint::realtime_websocket::methods_common::REALTIME_AUDIO_SAMPLE_RATE;
use crate::endpoint::realtime_websocket::protocol::AudioFormatType;
use crate::endpoint::realtime_websocket::protocol::ConversationContentType;
use crate::endpoint::realtime_websocket::protocol::ConversationFunctionCallOutputItem;
use crate::endpoint::realtime_websocket::protocol::ConversationItemContent;
use crate::endpoint::realtime_websocket::protocol::ConversationItemPayload;
use crate::endpoint::realtime_websocket::protocol::ConversationItemType;
use crate::endpoint::realtime_websocket::protocol::ConversationMessageItem;
use crate::endpoint::realtime_websocket::protocol::ConversationRole;
use crate::endpoint::realtime_websocket::protocol::NoiseReductionType;
use crate::endpoint::realtime_websocket::protocol::RealtimeOutboundMessage;
use crate::endpoint::realtime_websocket::protocol::RealtimeSessionMode;
use crate::endpoint::realtime_websocket::protocol::SessionAudio;
use crate::endpoint::realtime_websocket::protocol::SessionAudioFormat;
use crate::endpoint::realtime_websocket::protocol::SessionAudioInput;
use crate::endpoint::realtime_websocket::protocol::SessionAudioOutput;
use crate::endpoint::realtime_websocket::protocol::SessionAudioOutputFormat;
use crate::endpoint::realtime_websocket::protocol::SessionAudioVoice;
use crate::endpoint::realtime_websocket::protocol::SessionFunctionTool;
use crate::endpoint::realtime_websocket::protocol::SessionNoiseReduction;
use crate::endpoint::realtime_websocket::protocol::SessionToolType;
use crate::endpoint::realtime_websocket::protocol::SessionTurnDetection;
use crate::endpoint::realtime_websocket::protocol::SessionType;
use crate::endpoint::realtime_websocket::protocol::SessionUpdateSession;
use crate::endpoint::realtime_websocket::protocol::TurnDetectionType;
use serde_json::json;
const REALTIME_V2_SESSION_TYPE: &str = "realtime";
const REALTIME_V2_OUTPUT_MODALITY_AUDIO: &str = "audio";
const REALTIME_V2_TOOL_CHOICE: &str = "auto";
const REALTIME_V2_CODEX_TOOL_NAME: &str = "codex";
const REALTIME_V2_CODEX_TOOL_DESCRIPTION: &str = "Delegate work to Codex and return the result.";
const REALTIME_V2_CODEX_TOOL_DESCRIPTION: &str = "Delegate a request to Codex and return the final result to the user. Use this as the default action. If the user asks to do something next, later, after this, or once current work finishes, call this tool so the work is actually queued instead of merely promising to do it later.";
pub(super) fn conversation_item_create_message(text: String) -> RealtimeOutboundMessage {
RealtimeOutboundMessage::ConversationItemCreate {
item: ConversationItemPayload::Message(ConversationMessageItem {
kind: "message".to_string(),
role: "user".to_string(),
r#type: ConversationItemType::Message,
role: ConversationRole::User,
content: vec![ConversationItemContent {
kind: "input_text".to_string(),
r#type: ConversationContentType::InputText,
text,
}],
}),
@@ -38,7 +49,7 @@ pub(super) fn conversation_handoff_append_message(
) -> RealtimeOutboundMessage {
RealtimeOutboundMessage::ConversationItemCreate {
item: ConversationItemPayload::FunctionCallOutput(ConversationFunctionCallOutputItem {
kind: "function_call_output".to_string(),
r#type: ConversationItemType::FunctionCallOutput,
call_id: handoff_id,
output: output_text,
}),
@@ -51,21 +62,34 @@ pub(super) fn session_update_session(
) -> SessionUpdateSession {
match session_mode {
RealtimeSessionMode::Conversational => SessionUpdateSession {
kind: REALTIME_V2_SESSION_TYPE.to_string(),
r#type: SessionType::Realtime,
instructions: Some(instructions),
output_modalities: Some(vec![REALTIME_V2_OUTPUT_MODALITY_AUDIO.to_string()]),
audio: SessionAudio {
input: SessionAudioInput {
format: SessionAudioFormat {
kind: REALTIME_AUDIO_FORMAT.to_string(),
r#type: AudioFormatType::AudioPcm,
rate: REALTIME_AUDIO_SAMPLE_RATE,
},
noise_reduction: Some(SessionNoiseReduction {
r#type: NoiseReductionType::NearField,
}),
turn_detection: Some(SessionTurnDetection {
r#type: TurnDetectionType::ServerVad,
interrupt_response: true,
create_response: true,
}),
},
output: Some(SessionAudioOutput {
voice: SessionAudioVoice::Alloy,
format: Some(SessionAudioOutputFormat {
r#type: AudioFormatType::AudioPcm,
rate: REALTIME_AUDIO_SAMPLE_RATE,
}),
voice: SessionAudioVoice::Marin,
}),
},
tools: Some(vec![SessionFunctionTool {
kind: "function".to_string(),
r#type: SessionToolType::Function,
name: REALTIME_V2_CODEX_TOOL_NAME.to_string(),
description: REALTIME_V2_CODEX_TOOL_DESCRIPTION.to_string(),
parameters: json!({
@@ -73,27 +97,32 @@ pub(super) fn session_update_session(
"properties": {
"prompt": {
"type": "string",
"description": "Prompt text for the delegated Codex task."
"description": "The user request to delegate to Codex."
}
},
"required": ["prompt"],
"additionalProperties": false
}),
}]),
tool_choice: Some(REALTIME_V2_TOOL_CHOICE.to_string()),
},
RealtimeSessionMode::Transcription => SessionUpdateSession {
kind: "transcription".to_string(),
r#type: SessionType::Transcription,
instructions: None,
output_modalities: None,
audio: SessionAudio {
input: SessionAudioInput {
format: SessionAudioFormat {
kind: REALTIME_AUDIO_FORMAT.to_string(),
r#type: AudioFormatType::AudioPcm,
rate: REALTIME_AUDIO_SAMPLE_RATE,
},
noise_reduction: None,
turn_detection: None,
},
output: None,
},
tools: None,
tool_choice: None,
},
}
}

103
codex-rs/codex-api/src/endpoint/realtime_websocket/protocol.rs
View File

@@ -39,6 +39,8 @@ pub(super) enum RealtimeOutboundMessage {
handoff_id: String,
output_text: String,
},
#[serde(rename = "response.create")]
ResponseCreate,
#[serde(rename = "session.update")]
SessionUpdate { session: SessionUpdateSession },
#[serde(rename = "conversation.item.create")]
@@ -48,12 +50,24 @@ pub(super) enum RealtimeOutboundMessage {
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionUpdateSession {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) r#type: SessionType,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) instructions: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) output_modalities: Option<Vec<String>>,
pub(super) audio: SessionAudio,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) tools: Option<Vec<SessionFunctionTool>>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) tool_choice: Option<String>,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum SessionType {
Quicksilver,
Realtime,
Transcription,
}
#[derive(Debug, Clone, Serialize)]
@@ -66,17 +80,29 @@ pub(super) struct SessionAudio {
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionAudioInput {
pub(super) format: SessionAudioFormat,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) noise_reduction: Option<SessionNoiseReduction>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) turn_detection: Option<SessionTurnDetection>,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionAudioFormat {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) r#type: AudioFormatType,
pub(super) rate: u32,
}
#[derive(Debug, Clone, Copy, Serialize)]
pub(super) enum AudioFormatType {
#[serde(rename = "audio/pcm")]
AudioPcm,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionAudioOutput {
#[serde(skip_serializing_if = "Option::is_none")]
pub(super) format: Option<SessionAudioOutputFormat>,
pub(super) voice: SessionAudioVoice,
}
@@ -84,18 +110,64 @@ pub(super) struct SessionAudioOutput {
pub(super) enum SessionAudioVoice {
#[serde(rename = "fathom")]
Fathom,
#[serde(rename = "alloy")]
Alloy,
#[serde(rename = "marin")]
Marin,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionNoiseReduction {
#[serde(rename = "type")]
pub(super) r#type: NoiseReductionType,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum NoiseReductionType {
NearField,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionTurnDetection {
#[serde(rename = "type")]
pub(super) r#type: TurnDetectionType,
pub(super) interrupt_response: bool,
pub(super) create_response: bool,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum TurnDetectionType {
ServerVad,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionAudioOutputFormat {
#[serde(rename = "type")]
pub(super) r#type: AudioFormatType,
pub(super) rate: u32,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct ConversationMessageItem {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) role: String,
pub(super) r#type: ConversationItemType,
pub(super) role: ConversationRole,
pub(super) content: Vec<ConversationItemContent>,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum ConversationItemType {
Message,
FunctionCallOutput,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum ConversationRole {
User,
}
#[derive(Debug, Clone, Serialize)]
#[serde(untagged)]
pub(super) enum ConversationItemPayload {
@@ -106,7 +178,7 @@ pub(super) enum ConversationItemPayload {
#[derive(Debug, Clone, Serialize)]
pub(super) struct ConversationFunctionCallOutputItem {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) r#type: ConversationItemType,
pub(super) call_id: String,
pub(super) output: String,
}
@@ -114,19 +186,32 @@ pub(super) struct ConversationFunctionCallOutputItem {
#[derive(Debug, Clone, Serialize)]
pub(super) struct ConversationItemContent {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) r#type: ConversationContentType,
pub(super) text: String,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum ConversationContentType {
Text,
InputText,
}
#[derive(Debug, Clone, Serialize)]
pub(super) struct SessionFunctionTool {
#[serde(rename = "type")]
pub(super) kind: String,
pub(super) r#type: SessionToolType,
pub(super) name: String,
pub(super) description: String,
pub(super) parameters: Value,
}
#[derive(Debug, Clone, Copy, Serialize)]
#[serde(rename_all = "snake_case")]
pub(super) enum SessionToolType {
Function,
}
pub(super) fn parse_realtime_event(
payload: &str,
event_parser: RealtimeEventParser,

1
codex-rs/codex-api/src/endpoint/realtime_websocket/protocol_v1.rs
View File

@@ -35,6 +35,7 @@ pub(super) fn parse_realtime_event_v1(payload: &str) -> Option<RealtimeEvent> {
.get("samples_per_channel")
.and_then(Value::as_u64)
.and_then(|value| u32::try_from(value).ok()),
item_id: None,
}))
}
"conversation.input_transcript.delta" => {

60
codex-rs/codex-api/src/endpoint/realtime_websocket/protocol_v2.rs
View File

@@ -5,6 +5,8 @@ use crate::endpoint::realtime_websocket::protocol_common::parse_transcript_delta
use codex_protocol::protocol::RealtimeAudioFrame;
use codex_protocol::protocol::RealtimeEvent;
use codex_protocol::protocol::RealtimeHandoffRequested;
use codex_protocol::protocol::RealtimeInputAudioSpeechStarted;
use codex_protocol::protocol::RealtimeResponseCancelled;
use serde_json::Map as JsonMap;
use serde_json::Value;
use tracing::debug;
@@ -19,7 +21,9 @@ pub(super) fn parse_realtime_event_v2(payload: &str) -> Option<RealtimeEvent> {
match message_type.as_str() {
"session.updated" => parse_session_updated_event(&parsed),
"response.output_audio.delta" => parse_output_audio_delta_event(&parsed),
"response.output_audio.delta" | "response.audio.delta" => {
parse_output_audio_delta_event(&parsed)
}
"conversation.item.input_audio_transcription.delta" => {
parse_transcript_delta_event(&parsed, "delta").map(RealtimeEvent::InputTranscriptDelta)
}
@@ -30,11 +34,37 @@ pub(super) fn parse_realtime_event_v2(payload: &str) -> Option<RealtimeEvent> {
"response.output_text.delta" | "response.output_audio_transcript.delta" => {
parse_transcript_delta_event(&parsed, "delta").map(RealtimeEvent::OutputTranscriptDelta)
}
"input_audio_buffer.speech_started" => Some(RealtimeEvent::InputAudioSpeechStarted(
RealtimeInputAudioSpeechStarted {
item_id: parsed
.get("item_id")
.and_then(Value::as_str)
.map(str::to_string),
},
)),
"conversation.item.added" => parsed
.get("item")
.cloned()
.map(RealtimeEvent::ConversationItemAdded),
"conversation.item.done" => parse_conversation_item_done_event(&parsed),
"response.created" => Some(RealtimeEvent::ConversationItemAdded(parsed)),
"response.done" => parse_response_done_event(parsed),
"response.cancelled" => Some(RealtimeEvent::ResponseCancelled(
RealtimeResponseCancelled {
response_id: parsed
.get("response")
.and_then(Value::as_object)
.and_then(|response| response.get("id"))
.and_then(Value::as_str)
.map(str::to_string)
.or_else(|| {
parsed
.get("response_id")
.and_then(Value::as_str)
.map(str::to_string)
}),
},
)),
"error" => parse_error_event(&parsed),
_ => {
debug!("received unsupported realtime v2 event type: {message_type}, data: {payload}");
@@ -67,6 +97,10 @@ fn parse_output_audio_delta_event(parsed: &Value) -> Option<RealtimeEvent> {
.get("samples_per_channel")
.and_then(Value::as_u64)
.and_then(|value| u32::try_from(value).ok()),
item_id: parsed
.get("item_id")
.and_then(Value::as_str)
.map(str::to_string),
}))
}
@@ -82,6 +116,30 @@ fn parse_conversation_item_done_event(parsed: &Value) -> Option<RealtimeEvent> {
.map(|item_id| RealtimeEvent::ConversationItemDone { item_id })
}
fn parse_response_done_event(parsed: Value) -> Option<RealtimeEvent> {
if let Some(handoff) = parse_response_done_handoff_requested_event(&parsed) {
return Some(handoff);
}
Some(RealtimeEvent::ConversationItemAdded(parsed))
}
fn parse_response_done_handoff_requested_event(parsed: &Value) -> Option<RealtimeEvent> {
let item = parsed
.get("response")
.and_then(Value::as_object)
.and_then(|response| response.get("output"))
.and_then(Value::as_array)?
.iter()
.find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call")
&& item.get("name").and_then(Value::as_str) == Some(CODEX_TOOL_NAME)
})?
.as_object()?;
parse_handoff_requested_event(item)
}
fn parse_handoff_requested_event(item: &JsonMap<String, Value>) -> Option<RealtimeEvent> {
let item_type = item.get("type").and_then(Value::as_str);
let item_name = item.get("name").and_then(Value::as_str);