openai 0.22.1 → 0.23.0

data/sig/openai/models/realtime/realtime_audio_input_turn_detection.rbs

@@ -0,0 +1,99 @@
+module OpenAI
+  module Models
+    module Realtime
+      type realtime_audio_input_turn_detection =
+        {
+          create_response: bool,
+          eagerness: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness,
+          idle_timeout_ms: Integer?,
+          interrupt_response: bool,
+          prefix_padding_ms: Integer,
+          silence_duration_ms: Integer,
+          threshold: Float,
+          type: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_
+        }
+
+      class RealtimeAudioInputTurnDetection < OpenAI::Internal::Type::BaseModel
+        attr_reader create_response: bool?
+
+        def create_response=: (bool) -> bool
+
+        attr_reader eagerness: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness?
+
+        def eagerness=: (
+          OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness
+        ) -> OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness
+
+        attr_accessor idle_timeout_ms: Integer?
+
+        attr_reader interrupt_response: bool?
+
+        def interrupt_response=: (bool) -> bool
+
+        attr_reader prefix_padding_ms: Integer?
+
+        def prefix_padding_ms=: (Integer) -> Integer
+
+        attr_reader silence_duration_ms: Integer?
+
+        def silence_duration_ms=: (Integer) -> Integer
+
+        attr_reader threshold: Float?
+
+        def threshold=: (Float) -> Float
+
+        attr_reader type: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_?
+
+        def type=: (
+          OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_
+        ) -> OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_
+
+        def initialize: (
+          ?create_response: bool,
+          ?eagerness: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness,
+          ?idle_timeout_ms: Integer?,
+          ?interrupt_response: bool,
+          ?prefix_padding_ms: Integer,
+          ?silence_duration_ms: Integer,
+          ?threshold: Float,
+          ?type: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_
+        ) -> void
+
+        def to_hash: -> {
+          create_response: bool,
+          eagerness: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness,
+          idle_timeout_ms: Integer?,
+          interrupt_response: bool,
+          prefix_padding_ms: Integer,
+          silence_duration_ms: Integer,
+          threshold: Float,
+          type: OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_
+        }
+
+        type eagerness = :low | :medium | :high | :auto
+
+        module Eagerness
+          extend OpenAI::Internal::Type::Enum
+
+          LOW: :low
+          MEDIUM: :medium
+          HIGH: :high
+          AUTO: :auto
+
+          def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::eagerness]
+        end
+
+        type type_ = :server_vad | :semantic_vad
+
+        module Type
+          extend OpenAI::Internal::Type::Enum
+
+          SERVER_VAD: :server_vad
+          SEMANTIC_VAD: :semantic_vad
+
+          def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeAudioInputTurnDetection::type_]
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/realtime/realtime_conversation_item_assistant_message.rbs

@@ -54,15 +54,25 @@ module OpenAI
 
         type content =
           {
+            audio: String,
             text: String,
+            transcript: String,
             type: OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_
           }
 
         class Content < OpenAI::Internal::Type::BaseModel
+          attr_reader audio: String?
+
+          def audio=: (String) -> String
+
           attr_reader text: String?
 
           def text=: (String) -> String
 
+          attr_reader transcript: String?
+
+          def transcript=: (String) -> String
+
           attr_reader type: OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_?
 
           def type=: (
@@ -70,21 +80,26 @@ module OpenAI
           ) -> OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_
 
           def initialize: (
+            ?audio: String,
             ?text: String,
+            ?transcript: String,
             ?type: OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_
           ) -> void
 
           def to_hash: -> {
+            audio: String,
             text: String,
+            transcript: String,
             type: OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_
           }
 
-          type type_ = :text
+          type type_ = :output_text | :output_audio
 
           module Type
             extend OpenAI::Internal::Type::Enum
 
-            TEXT: :text
+            OUTPUT_TEXT: :output_text
+            OUTPUT_AUDIO: :output_audio
 
             def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeConversationItemAssistantMessage::Content::type_]
           end

data/sig/openai/models/realtime/realtime_conversation_item_user_message.rbs

@@ -55,6 +55,8 @@ module OpenAI
         type content =
           {
             audio: String,
+            detail: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail,
+            image_url: String,
             text: String,
             transcript: String,
             type: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::type_
@@ -65,6 +67,16 @@ module OpenAI
 
           def audio=: (String) -> String
 
+          attr_reader detail: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail?
+
+          def detail=: (
+            OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail
+          ) -> OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail
+
+          attr_reader image_url: String?
+
+          def image_url=: (String) -> String
+
           attr_reader text: String?
 
           def text=: (String) -> String
@@ -81,6 +93,8 @@ module OpenAI
 
           def initialize: (
             ?audio: String,
+            ?detail: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail,
+            ?image_url: String,
             ?text: String,
             ?transcript: String,
             ?type: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::type_
@@ -88,18 +102,33 @@ module OpenAI
 
           def to_hash: -> {
             audio: String,
+            detail: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail,
+            image_url: String,
             text: String,
             transcript: String,
             type: OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::type_
           }
 
-          type type_ = :input_text | :input_audio
+          type detail = :auto | :low | :high
+
+          module Detail
+            extend OpenAI::Internal::Type::Enum
+
+            AUTO: :auto
+            LOW: :low
+            HIGH: :high
+
+            def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::detail]
+          end
+
+          type type_ = :input_text | :input_audio | :input_image
 
           module Type
             extend OpenAI::Internal::Type::Enum
 
             INPUT_TEXT: :input_text
             INPUT_AUDIO: :input_audio
+            INPUT_IMAGE: :input_image
 
             def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeConversationItemUserMessage::Content::type_]
           end

data/sig/openai/models/realtime/realtime_response.rbs

@@ -4,18 +4,16 @@ module OpenAI
       type realtime_response =
         {
           id: String,
+          audio: OpenAI::Realtime::RealtimeResponse::Audio,
           conversation_id: String,
           max_output_tokens: OpenAI::Models::Realtime::RealtimeResponse::max_output_tokens,
           metadata: OpenAI::Models::metadata?,
-          modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality],
           object: OpenAI::Models::Realtime::RealtimeResponse::object,
           output: ::Array[OpenAI::Models::Realtime::conversation_item],
-          output_audio_format: OpenAI::Models::Realtime::RealtimeResponse::output_audio_format,
+          output_modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality],
           status: OpenAI::Models::Realtime::RealtimeResponse::status,
           status_details: OpenAI::Realtime::RealtimeResponseStatus,
-          temperature: Float,
-          usage: OpenAI::Realtime::RealtimeResponseUsage,
-          voice: OpenAI::Models::Realtime::RealtimeResponse::voice
+          usage: OpenAI::Realtime::RealtimeResponseUsage
         }
 
       class RealtimeResponse < OpenAI::Internal::Type::BaseModel
@@ -23,6 +21,12 @@ module OpenAI
 
         def id=: (String) -> String
 
+        attr_reader audio: OpenAI::Realtime::RealtimeResponse::Audio?
+
+        def audio=: (
+          OpenAI::Realtime::RealtimeResponse::Audio
+        ) -> OpenAI::Realtime::RealtimeResponse::Audio
+
         attr_reader conversation_id: String?
 
         def conversation_id=: (String) -> String
@@ -35,12 +39,6 @@ module OpenAI
 
         attr_accessor metadata: OpenAI::Models::metadata?
 
-        attr_reader modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality]?
-
-        def modalities=: (
-          ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality]
-        ) -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality]
-
         attr_reader object: OpenAI::Models::Realtime::RealtimeResponse::object?
 
         def object=: (
@@ -53,11 +51,11 @@ module OpenAI
           ::Array[OpenAI::Models::Realtime::conversation_item]
         ) -> ::Array[OpenAI::Models::Realtime::conversation_item]
 
-        attr_reader output_audio_format: OpenAI::Models::Realtime::RealtimeResponse::output_audio_format?
+        attr_reader output_modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality]?
 
-        def output_audio_format=: (
-          OpenAI::Models::Realtime::RealtimeResponse::output_audio_format
-        ) -> OpenAI::Models::Realtime::RealtimeResponse::output_audio_format
+        def output_modalities=: (
+          ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality]
+        ) -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality]
 
         attr_reader status: OpenAI::Models::Realtime::RealtimeResponse::status?
 
@@ -71,54 +69,119 @@ module OpenAI
           OpenAI::Realtime::RealtimeResponseStatus
         ) -> OpenAI::Realtime::RealtimeResponseStatus
 
-        attr_reader temperature: Float?
-
-        def temperature=: (Float) -> Float
-
         attr_reader usage: OpenAI::Realtime::RealtimeResponseUsage?
 
         def usage=: (
           OpenAI::Realtime::RealtimeResponseUsage
         ) -> OpenAI::Realtime::RealtimeResponseUsage
 
-        attr_reader voice: OpenAI::Models::Realtime::RealtimeResponse::voice?
-
-        def voice=: (
-          OpenAI::Models::Realtime::RealtimeResponse::voice
-        ) -> OpenAI::Models::Realtime::RealtimeResponse::voice
-
         def initialize: (
           ?id: String,
+          ?audio: OpenAI::Realtime::RealtimeResponse::Audio,
           ?conversation_id: String,
           ?max_output_tokens: OpenAI::Models::Realtime::RealtimeResponse::max_output_tokens,
           ?metadata: OpenAI::Models::metadata?,
-          ?modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality],
           ?object: OpenAI::Models::Realtime::RealtimeResponse::object,
           ?output: ::Array[OpenAI::Models::Realtime::conversation_item],
-          ?output_audio_format: OpenAI::Models::Realtime::RealtimeResponse::output_audio_format,
+          ?output_modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality],
           ?status: OpenAI::Models::Realtime::RealtimeResponse::status,
           ?status_details: OpenAI::Realtime::RealtimeResponseStatus,
-          ?temperature: Float,
-          ?usage: OpenAI::Realtime::RealtimeResponseUsage,
-          ?voice: OpenAI::Models::Realtime::RealtimeResponse::voice
+          ?usage: OpenAI::Realtime::RealtimeResponseUsage
         ) -> void
 
         def to_hash: -> {
           id: String,
+          audio: OpenAI::Realtime::RealtimeResponse::Audio,
           conversation_id: String,
           max_output_tokens: OpenAI::Models::Realtime::RealtimeResponse::max_output_tokens,
           metadata: OpenAI::Models::metadata?,
-          modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality],
           object: OpenAI::Models::Realtime::RealtimeResponse::object,
           output: ::Array[OpenAI::Models::Realtime::conversation_item],
-          output_audio_format: OpenAI::Models::Realtime::RealtimeResponse::output_audio_format,
+          output_modalities: ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality],
           status: OpenAI::Models::Realtime::RealtimeResponse::status,
           status_details: OpenAI::Realtime::RealtimeResponseStatus,
-          temperature: Float,
-          usage: OpenAI::Realtime::RealtimeResponseUsage,
-          voice: OpenAI::Models::Realtime::RealtimeResponse::voice
+          usage: OpenAI::Realtime::RealtimeResponseUsage
         }
 
+        type audio =
+          { output: OpenAI::Realtime::RealtimeResponse::Audio::Output }
+
+        class Audio < OpenAI::Internal::Type::BaseModel
+          attr_reader output: OpenAI::Realtime::RealtimeResponse::Audio::Output?
+
+          def output=: (
+            OpenAI::Realtime::RealtimeResponse::Audio::Output
+          ) -> OpenAI::Realtime::RealtimeResponse::Audio::Output
+
+          def initialize: (
+            ?output: OpenAI::Realtime::RealtimeResponse::Audio::Output
+          ) -> void
+
+          def to_hash: -> {
+            output: OpenAI::Realtime::RealtimeResponse::Audio::Output
+          }
+
+          type output =
+            {
+              format_: OpenAI::Models::Realtime::realtime_audio_formats,
+              voice: OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice
+            }
+
+          class Output < OpenAI::Internal::Type::BaseModel
+            attr_reader format_: OpenAI::Models::Realtime::realtime_audio_formats?
+
+            def format_=: (
+              OpenAI::Models::Realtime::realtime_audio_formats
+            ) -> OpenAI::Models::Realtime::realtime_audio_formats
+
+            attr_reader voice: OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice?
+
+            def voice=: (
+              OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice
+            ) -> OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice
+
+            def initialize: (
+              ?format_: OpenAI::Models::Realtime::realtime_audio_formats,
+              ?voice: OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice
+            ) -> void
+
+            def to_hash: -> {
+              format_: OpenAI::Models::Realtime::realtime_audio_formats,
+              voice: OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice
+            }
+
+            type voice =
+              String
+              | :alloy
+              | :ash
+              | :ballad
+              | :coral
+              | :echo
+              | :sage
+              | :shimmer
+              | :verse
+              | :marin
+              | :cedar
+
+            module Voice
+              extend OpenAI::Internal::Type::Union
+
+              def self?.variants: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::Audio::Output::voice]
+
+              ALLOY: :alloy
+              ASH: :ash
+              BALLAD: :ballad
+              CORAL: :coral
+              ECHO: :echo
+              SAGE: :sage
+              SHIMMER: :shimmer
+              VERSE: :verse
+              MARIN: :marin
+              CEDAR: :cedar
+            end
+          end
+        end
+
         type max_output_tokens = Integer | :inf
 
         module MaxOutputTokens
@@ -127,17 +190,6 @@ module OpenAI
           def self?.variants: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::max_output_tokens]
         end
 
-        type modality = :text | :audio
-
-        module Modality
-          extend OpenAI::Internal::Type::Enum
-
-          TEXT: :text
-          AUDIO: :audio
-
-          def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::modality]
-        end
-
         type object = :"realtime.response"
 
         module Object
@@ -148,16 +200,15 @@ module OpenAI
           def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::object]
         end
 
-        type output_audio_format = :pcm16 | :g711_ulaw | :g711_alaw
+        type output_modality = :text | :audio
 
-        module OutputAudioFormat
+        module OutputModality
           extend OpenAI::Internal::Type::Enum
 
-          PCM16: :pcm16
-          G711_ULAW: :g711_ulaw
-          G711_ALAW: :g711_alaw
+          TEXT: :text
+          AUDIO: :audio
 
-          def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_audio_format]
+          def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::output_modality]
         end
 
         type status =
@@ -174,36 +225,6 @@ module OpenAI
 
           def self?.values: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::status]
         end
-
-        type voice =
-          String
-          | :alloy
-          | :ash
-          | :ballad
-          | :coral
-          | :echo
-          | :sage
-          | :shimmer
-          | :verse
-          | :marin
-          | :cedar
-
-        module Voice
-          extend OpenAI::Internal::Type::Union
-
-          def self?.variants: -> ::Array[OpenAI::Models::Realtime::RealtimeResponse::voice]
-
-          ALLOY: :alloy
-          ASH: :ash
-          BALLAD: :ballad
-          CORAL: :coral
-          ECHO: :echo
-          SAGE: :sage
-          SHIMMER: :shimmer
-          VERSE: :verse
-          MARIN: :marin
-          CEDAR: :cedar
-        end
       end
     end
   end

data/sig/openai/models/realtime/realtime_response_create_audio_output.rbs

@@ -0,0 +1,84 @@
+module OpenAI
+  module Models
+    module Realtime
+      type realtime_response_create_audio_output =
+        { output: OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output }
+
+      class RealtimeResponseCreateAudioOutput < OpenAI::Internal::Type::BaseModel
+        attr_reader output: OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output?
+
+        def output=: (
+          OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output
+        ) -> OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output
+
+        def initialize: (
+          ?output: OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output
+        ) -> void
+
+        def to_hash: -> {
+          output: OpenAI::Realtime::RealtimeResponseCreateAudioOutput::Output
+        }
+
+        type output =
+          {
+            format_: OpenAI::Models::Realtime::realtime_audio_formats,
+            voice: OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice
+          }
+
+        class Output < OpenAI::Internal::Type::BaseModel
+          attr_reader format_: OpenAI::Models::Realtime::realtime_audio_formats?
+
+          def format_=: (
+            OpenAI::Models::Realtime::realtime_audio_formats
+          ) -> OpenAI::Models::Realtime::realtime_audio_formats
+
+          attr_reader voice: OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice?
+
+          def voice=: (
+            OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice
+          ) -> OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice
+
+          def initialize: (
+            ?format_: OpenAI::Models::Realtime::realtime_audio_formats,
+            ?voice: OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice
+          ) -> void
+
+          def to_hash: -> {
+            format_: OpenAI::Models::Realtime::realtime_audio_formats,
+            voice: OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice
+          }
+
+          type voice =
+            String
+            | :alloy
+            | :ash
+            | :ballad
+            | :coral
+            | :echo
+            | :sage
+            | :shimmer
+            | :verse
+            | :marin
+            | :cedar
+
+          module Voice
+            extend OpenAI::Internal::Type::Union
+
+            def self?.variants: -> ::Array[OpenAI::Models::Realtime::RealtimeResponseCreateAudioOutput::Output::voice]
+
+            ALLOY: :alloy
+            ASH: :ash
+            BALLAD: :ballad
+            CORAL: :coral
+            ECHO: :echo
+            SAGE: :sage
+            SHIMMER: :shimmer
+            VERSE: :verse
+            MARIN: :marin
+            CEDAR: :cedar
+          end
+        end
+      end
+    end
+  end
+end