openai 0.9.0 → 0.10.0

data/rbi/openai/models/audio/speech_create_params.rbi

@@ -60,13 +60,30 @@ module OpenAI
         attr_writer :response_format
 
         # The speed of the generated audio. Select a value from `0.25` to `4.0`. `1.0` is
-        # the default. Does not work with `gpt-4o-mini-tts`.
+        # the default.
         sig { returns(T.nilable(Float)) }
         attr_reader :speed
 
         sig { params(speed: Float).void }
         attr_writer :speed
 
+        # The format to stream the audio in. Supported formats are `sse` and `audio`.
+        # `sse` is not supported for `tts-1` or `tts-1-hd`.
+        sig do
+          returns(
+            T.nilable(OpenAI::Audio::SpeechCreateParams::StreamFormat::OrSymbol)
+          )
+        end
+        attr_reader :stream_format
+
+        sig do
+          params(
+            stream_format:
+              OpenAI::Audio::SpeechCreateParams::StreamFormat::OrSymbol
+          ).void
+        end
+        attr_writer :stream_format
+
         sig do
           params(
             input: String,
@@ -77,6 +94,8 @@ module OpenAI
             response_format:
               OpenAI::Audio::SpeechCreateParams::ResponseFormat::OrSymbol,
             speed: Float,
+            stream_format:
+              OpenAI::Audio::SpeechCreateParams::StreamFormat::OrSymbol,
             request_options: OpenAI::RequestOptions::OrHash
           ).returns(T.attached_class)
         end
@@ -98,8 +117,11 @@ module OpenAI
           # `wav`, and `pcm`.
           response_format: nil,
           # The speed of the generated audio. Select a value from `0.25` to `4.0`. `1.0` is
-          # the default. Does not work with `gpt-4o-mini-tts`.
+          # the default.
           speed: nil,
+          # The format to stream the audio in. Supported formats are `sse` and `audio`.
+          # `sse` is not supported for `tts-1` or `tts-1-hd`.
+          stream_format: nil,
           request_options: {}
         )
         end
@@ -118,6 +140,8 @@ module OpenAI
               response_format:
                 OpenAI::Audio::SpeechCreateParams::ResponseFormat::OrSymbol,
               speed: Float,
+              stream_format:
+                OpenAI::Audio::SpeechCreateParams::StreamFormat::OrSymbol,
               request_options: OpenAI::RequestOptions
             }
           )
@@ -267,6 +291,39 @@ module OpenAI
           def self.values
           end
         end
+
+        # The format to stream the audio in. Supported formats are `sse` and `audio`.
+        # `sse` is not supported for `tts-1` or `tts-1-hd`.
+        module StreamFormat
+          extend OpenAI::Internal::Type::Enum
+
+          TaggedSymbol =
+            T.type_alias do
+              T.all(Symbol, OpenAI::Audio::SpeechCreateParams::StreamFormat)
+            end
+          OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+          SSE =
+            T.let(
+              :sse,
+              OpenAI::Audio::SpeechCreateParams::StreamFormat::TaggedSymbol
+            )
+          AUDIO =
+            T.let(
+              :audio,
+              OpenAI::Audio::SpeechCreateParams::StreamFormat::TaggedSymbol
+            )
+
+          sig do
+            override.returns(
+              T::Array[
+                OpenAI::Audio::SpeechCreateParams::StreamFormat::TaggedSymbol
+              ]
+            )
+          end
+          def self.values
+          end
+        end
       end
     end
   end

data/rbi/openai/models/audio/transcription.rbi

@@ -28,12 +28,34 @@ module OpenAI
         end
         attr_writer :logprobs
 
+        # Token usage statistics for the request.
+        sig do
+          returns(T.nilable(OpenAI::Audio::Transcription::Usage::Variants))
+        end
+        attr_reader :usage
+
+        sig do
+          params(
+            usage:
+              T.any(
+                OpenAI::Audio::Transcription::Usage::Tokens::OrHash,
+                OpenAI::Audio::Transcription::Usage::Duration::OrHash
+              )
+          ).void
+        end
+        attr_writer :usage
+
         # Represents a transcription response returned by model, based on the provided
         # input.
         sig do
           params(
             text: String,
-            logprobs: T::Array[OpenAI::Audio::Transcription::Logprob::OrHash]
+            logprobs: T::Array[OpenAI::Audio::Transcription::Logprob::OrHash],
+            usage:
+              T.any(
+                OpenAI::Audio::Transcription::Usage::Tokens::OrHash,
+                OpenAI::Audio::Transcription::Usage::Duration::OrHash
+              )
           ).returns(T.attached_class)
         end
         def self.new(
@@ -42,7 +64,9 @@ module OpenAI
           # The log probabilities of the tokens in the transcription. Only returned with the
           # models `gpt-4o-transcribe` and `gpt-4o-mini-transcribe` if `logprobs` is added
           # to the `include` array.
-          logprobs: nil
+          logprobs: nil,
+          # Token usage statistics for the request.
+          usage: nil
         )
         end
 
@@ -50,7 +74,8 @@ module OpenAI
           override.returns(
             {
               text: String,
-              logprobs: T::Array[OpenAI::Audio::Transcription::Logprob]
+              logprobs: T::Array[OpenAI::Audio::Transcription::Logprob],
+              usage: OpenAI::Audio::Transcription::Usage::Variants
             }
           )
         end
@@ -112,6 +137,191 @@ module OpenAI
           def to_hash
           end
         end
+
+        # Token usage statistics for the request.
+        module Usage
+          extend OpenAI::Internal::Type::Union
+
+          Variants =
+            T.type_alias do
+              T.any(
+                OpenAI::Audio::Transcription::Usage::Tokens,
+                OpenAI::Audio::Transcription::Usage::Duration
+              )
+            end
+
+          class Tokens < OpenAI::Internal::Type::BaseModel
+            OrHash =
+              T.type_alias do
+                T.any(
+                  OpenAI::Audio::Transcription::Usage::Tokens,
+                  OpenAI::Internal::AnyHash
+                )
+              end
+
+            # Number of input tokens billed for this request.
+            sig { returns(Integer) }
+            attr_accessor :input_tokens
+
+            # Number of output tokens generated.
+            sig { returns(Integer) }
+            attr_accessor :output_tokens
+
+            # Total number of tokens used (input + output).
+            sig { returns(Integer) }
+            attr_accessor :total_tokens
+
+            # The type of the usage object. Always `tokens` for this variant.
+            sig { returns(Symbol) }
+            attr_accessor :type
+
+            # Details about the input tokens billed for this request.
+            sig do
+              returns(
+                T.nilable(
+                  OpenAI::Audio::Transcription::Usage::Tokens::InputTokenDetails
+                )
+              )
+            end
+            attr_reader :input_token_details
+
+            sig do
+              params(
+                input_token_details:
+                  OpenAI::Audio::Transcription::Usage::Tokens::InputTokenDetails::OrHash
+              ).void
+            end
+            attr_writer :input_token_details
+
+            # Usage statistics for models billed by token usage.
+            sig do
+              params(
+                input_tokens: Integer,
+                output_tokens: Integer,
+                total_tokens: Integer,
+                input_token_details:
+                  OpenAI::Audio::Transcription::Usage::Tokens::InputTokenDetails::OrHash,
+                type: Symbol
+              ).returns(T.attached_class)
+            end
+            def self.new(
+              # Number of input tokens billed for this request.
+              input_tokens:,
+              # Number of output tokens generated.
+              output_tokens:,
+              # Total number of tokens used (input + output).
+              total_tokens:,
+              # Details about the input tokens billed for this request.
+              input_token_details: nil,
+              # The type of the usage object. Always `tokens` for this variant.
+              type: :tokens
+            )
+            end
+
+            sig do
+              override.returns(
+                {
+                  input_tokens: Integer,
+                  output_tokens: Integer,
+                  total_tokens: Integer,
+                  type: Symbol,
+                  input_token_details:
+                    OpenAI::Audio::Transcription::Usage::Tokens::InputTokenDetails
+                }
+              )
+            end
+            def to_hash
+            end
+
+            class InputTokenDetails < OpenAI::Internal::Type::BaseModel
+              OrHash =
+                T.type_alias do
+                  T.any(
+                    OpenAI::Audio::Transcription::Usage::Tokens::InputTokenDetails,
+                    OpenAI::Internal::AnyHash
+                  )
+                end
+
+              # Number of audio tokens billed for this request.
+              sig { returns(T.nilable(Integer)) }
+              attr_reader :audio_tokens
+
+              sig { params(audio_tokens: Integer).void }
+              attr_writer :audio_tokens
+
+              # Number of text tokens billed for this request.
+              sig { returns(T.nilable(Integer)) }
+              attr_reader :text_tokens
+
+              sig { params(text_tokens: Integer).void }
+              attr_writer :text_tokens
+
+              # Details about the input tokens billed for this request.
+              sig do
+                params(audio_tokens: Integer, text_tokens: Integer).returns(
+                  T.attached_class
+                )
+              end
+              def self.new(
+                # Number of audio tokens billed for this request.
+                audio_tokens: nil,
+                # Number of text tokens billed for this request.
+                text_tokens: nil
+              )
+              end
+
+              sig do
+                override.returns(
+                  { audio_tokens: Integer, text_tokens: Integer }
+                )
+              end
+              def to_hash
+              end
+            end
+          end
+
+          class Duration < OpenAI::Internal::Type::BaseModel
+            OrHash =
+              T.type_alias do
+                T.any(
+                  OpenAI::Audio::Transcription::Usage::Duration,
+                  OpenAI::Internal::AnyHash
+                )
+              end
+
+            # Duration of the input audio in seconds.
+            sig { returns(Float) }
+            attr_accessor :duration
+
+            # The type of the usage object. Always `duration` for this variant.
+            sig { returns(Symbol) }
+            attr_accessor :type
+
+            # Usage statistics for models billed by audio input duration.
+            sig do
+              params(duration: Float, type: Symbol).returns(T.attached_class)
+            end
+            def self.new(
+              # Duration of the input audio in seconds.
+              duration:,
+              # The type of the usage object. Always `duration` for this variant.
+              type: :duration
+            )
+            end
+
+            sig { override.returns({ duration: Float, type: Symbol }) }
+            def to_hash
+            end
+          end
+
+          sig do
+            override.returns(
+              T::Array[OpenAI::Audio::Transcription::Usage::Variants]
+            )
+          end
+          def self.variants
+          end
+        end
       end
     end
   end

data/rbi/openai/models/audio/transcription_text_done_event.rbi

@@ -43,6 +43,19 @@ module OpenAI
         end
         attr_writer :logprobs
 
+        # Usage statistics for models billed by token usage.
+        sig do
+          returns(T.nilable(OpenAI::Audio::TranscriptionTextDoneEvent::Usage))
+        end
+        attr_reader :usage
+
+        sig do
+          params(
+            usage: OpenAI::Audio::TranscriptionTextDoneEvent::Usage::OrHash
+          ).void
+        end
+        attr_writer :usage
+
         # Emitted when the transcription is complete. Contains the complete transcription
         # text. Only emitted when you
         # [create a transcription](https://platform.openai.com/docs/api-reference/audio/create-transcription)
@@ -54,6 +67,7 @@ module OpenAI
               T::Array[
                 OpenAI::Audio::TranscriptionTextDoneEvent::Logprob::OrHash
               ],
+            usage: OpenAI::Audio::TranscriptionTextDoneEvent::Usage::OrHash,
             type: Symbol
           ).returns(T.attached_class)
         end
@@ -65,6 +79,8 @@ module OpenAI
           # [create a transcription](https://platform.openai.com/docs/api-reference/audio/create-transcription)
           # with the `include[]` parameter set to `logprobs`.
           logprobs: nil,
+          # Usage statistics for models billed by token usage.
+          usage: nil,
           # The type of the event. Always `transcript.text.done`.
           type: :"transcript.text.done"
         )
@@ -76,7 +92,8 @@ module OpenAI
               text: String,
               type: Symbol,
               logprobs:
-                T::Array[OpenAI::Audio::TranscriptionTextDoneEvent::Logprob]
+                T::Array[OpenAI::Audio::TranscriptionTextDoneEvent::Logprob],
+              usage: OpenAI::Audio::TranscriptionTextDoneEvent::Usage
             }
           )
         end
@@ -138,6 +155,134 @@ module OpenAI
           def to_hash
           end
         end
+
+        class Usage < OpenAI::Internal::Type::BaseModel
+          OrHash =
+            T.type_alias do
+              T.any(
+                OpenAI::Audio::TranscriptionTextDoneEvent::Usage,
+                OpenAI::Internal::AnyHash
+              )
+            end
+
+          # Number of input tokens billed for this request.
+          sig { returns(Integer) }
+          attr_accessor :input_tokens
+
+          # Number of output tokens generated.
+          sig { returns(Integer) }
+          attr_accessor :output_tokens
+
+          # Total number of tokens used (input + output).
+          sig { returns(Integer) }
+          attr_accessor :total_tokens
+
+          # The type of the usage object. Always `tokens` for this variant.
+          sig { returns(Symbol) }
+          attr_accessor :type
+
+          # Details about the input tokens billed for this request.
+          sig do
+            returns(
+              T.nilable(
+                OpenAI::Audio::TranscriptionTextDoneEvent::Usage::InputTokenDetails
+              )
+            )
+          end
+          attr_reader :input_token_details
+
+          sig do
+            params(
+              input_token_details:
+                OpenAI::Audio::TranscriptionTextDoneEvent::Usage::InputTokenDetails::OrHash
+            ).void
+          end
+          attr_writer :input_token_details
+
+          # Usage statistics for models billed by token usage.
+          sig do
+            params(
+              input_tokens: Integer,
+              output_tokens: Integer,
+              total_tokens: Integer,
+              input_token_details:
+                OpenAI::Audio::TranscriptionTextDoneEvent::Usage::InputTokenDetails::OrHash,
+              type: Symbol
+            ).returns(T.attached_class)
+          end
+          def self.new(
+            # Number of input tokens billed for this request.
+            input_tokens:,
+            # Number of output tokens generated.
+            output_tokens:,
+            # Total number of tokens used (input + output).
+            total_tokens:,
+            # Details about the input tokens billed for this request.
+            input_token_details: nil,
+            # The type of the usage object. Always `tokens` for this variant.
+            type: :tokens
+          )
+          end
+
+          sig do
+            override.returns(
+              {
+                input_tokens: Integer,
+                output_tokens: Integer,
+                total_tokens: Integer,
+                type: Symbol,
+                input_token_details:
+                  OpenAI::Audio::TranscriptionTextDoneEvent::Usage::InputTokenDetails
+              }
+            )
+          end
+          def to_hash
+          end
+
+          class InputTokenDetails < OpenAI::Internal::Type::BaseModel
+            OrHash =
+              T.type_alias do
+                T.any(
+                  OpenAI::Audio::TranscriptionTextDoneEvent::Usage::InputTokenDetails,
+                  OpenAI::Internal::AnyHash
+                )
+              end
+
+            # Number of audio tokens billed for this request.
+            sig { returns(T.nilable(Integer)) }
+            attr_reader :audio_tokens
+
+            sig { params(audio_tokens: Integer).void }
+            attr_writer :audio_tokens
+
+            # Number of text tokens billed for this request.
+            sig { returns(T.nilable(Integer)) }
+            attr_reader :text_tokens
+
+            sig { params(text_tokens: Integer).void }
+            attr_writer :text_tokens
+
+            # Details about the input tokens billed for this request.
+            sig do
+              params(audio_tokens: Integer, text_tokens: Integer).returns(
+                T.attached_class
+              )
+            end
+            def self.new(
+              # Number of audio tokens billed for this request.
+              audio_tokens: nil,
+              # Number of text tokens billed for this request.
+              text_tokens: nil
+            )
+            end
+
+            sig do
+              override.returns({ audio_tokens: Integer, text_tokens: Integer })
+            end
+            def to_hash
+            end
+          end
+        end
       end
     end
   end

data/rbi/openai/models/audio/transcription_verbose.rbi

@@ -37,6 +37,15 @@ module OpenAI
         end
         attr_writer :segments
 
+        # Usage statistics for models billed by audio input duration.
+        sig { returns(T.nilable(OpenAI::Audio::TranscriptionVerbose::Usage)) }
+        attr_reader :usage
+
+        sig do
+          params(usage: OpenAI::Audio::TranscriptionVerbose::Usage::OrHash).void
+        end
+        attr_writer :usage
+
         # Extracted words and their corresponding timestamps.
         sig { returns(T.nilable(T::Array[OpenAI::Audio::TranscriptionWord])) }
         attr_reader :words
@@ -54,6 +63,7 @@ module OpenAI
             language: String,
             text: String,
             segments: T::Array[OpenAI::Audio::TranscriptionSegment::OrHash],
+            usage: OpenAI::Audio::TranscriptionVerbose::Usage::OrHash,
             words: T::Array[OpenAI::Audio::TranscriptionWord::OrHash]
           ).returns(T.attached_class)
         end
@@ -66,6 +76,8 @@ module OpenAI
           text:,
           # Segments of the transcribed text and their corresponding details.
           segments: nil,
+          # Usage statistics for models billed by audio input duration.
+          usage: nil,
           # Extracted words and their corresponding timestamps.
           words: nil
         )
@@ -78,12 +90,47 @@ module OpenAI
               language: String,
               text: String,
               segments: T::Array[OpenAI::Audio::TranscriptionSegment],
+              usage: OpenAI::Audio::TranscriptionVerbose::Usage,
               words: T::Array[OpenAI::Audio::TranscriptionWord]
             }
           )
         end
         def to_hash
         end
+
+        class Usage < OpenAI::Internal::Type::BaseModel
+          OrHash =
+            T.type_alias do
+              T.any(
+                OpenAI::Audio::TranscriptionVerbose::Usage,
+                OpenAI::Internal::AnyHash
+              )
+            end
+
+          # Duration of the input audio in seconds.
+          sig { returns(Float) }
+          attr_accessor :duration
+
+          # The type of the usage object. Always `duration` for this variant.
+          sig { returns(Symbol) }
+          attr_accessor :type
+
+          # Usage statistics for models billed by audio input duration.
+          sig do
+            params(duration: Float, type: Symbol).returns(T.attached_class)
+          end
+          def self.new(
+            # Duration of the input audio in seconds.
+            duration:,
+            # The type of the usage object. Always `duration` for this variant.
+            type: :duration
+          )
+          end
+
+          sig { override.returns({ duration: Float, type: Symbol }) }
+          def to_hash
+          end
+        end
       end
     end
   end