openai 0.90.0 → 0.91.0

data/responses-websocket.md

@@ -1,9 +1,11 @@
 # Responses WebSocket workflows
 
-Requires `openai` 0.86.0 or later and the optional `async-websocket` gem:
+Install `openai` and the optional `async-websocket` gem. The `multiplex` workflow
+uses `OpenAI::Responses::Session`; versions with only `client.responses.connect`
+can use the raw connection patterns below.
 
 ```ruby
-gem "openai", ">= 0.86.0"
+gem "openai"
 gem "async-websocket"
 ```
 
@@ -58,6 +60,73 @@ event types arrive as `OpenAI::Responses::UnknownServerEvent`. Preserve or handl
 them explicitly where your application needs them. Do not assume every event
 has a response, text delta, or lane ID.
 
+## Managed sessions
+
+`OpenAI::Responses::Session.open` adds routed lanes and final-response collection
+using the same authentication, custom headers, and connection options as
+`client.responses.connect`. With `OPENAI_API_KEY` set, this complete example
+opens a session and closes its reader and transport when the block exits:
+
+```ruby
+require "openai"
+
+limits = OpenAI::Responses::SessionLimits.new(
+  max_lanes: 8,
+  max_events_per_lane: 128,
+  max_events: 512,
+  max_bytes_per_lane: 16 * 1024 * 1024,
+  max_bytes: 32 * 1024 * 1024,
+  max_response_bytes: 64 * 1024 * 1024
+)
+
+OpenAI::Responses::Session.open(
+  client: OpenAI::Client.new,
+  limits: limits,
+  request_options: {extra_headers: {"X-Application-Request" => "example"}}
+) do |session|
+  lane = session.lane("conversation")
+  lane.send_event(type: "response.create", model: "gpt-5.2", input: "Say hello.")
+  response = lane.get_final_response
+  puts response.status
+  # Supply previous_response_id: response.id explicitly for continuation.
+end
+```
+
+These required budgets are application choices, not service limits or changes
+to raw connection defaults. The default and all detached lane IDs count toward `max_lanes` until reconnect. Queue
+budgets apply per lane and across the session; byte counts measure serialized
+event data, not total process memory. `max_response_bytes` bounds cumulative
+compatible event bytes retained for collection. Choose budgets for legitimate
+large output, including images and tool results. Queue overflow closes the
+session's connection; collection overflow leaves raw events and other lanes usable.
+
+Register lanes before sending and consume each response before creating another
+on that lane. One Async task reads the physical connection; each lane has one
+consumer, and all session operations use the owning Ruby thread. `lane.receive`
+returns original typed or unknown events. `lane.get_final_response` consumes
+remaining events and returns the current completed, failed, or incomplete response;
+inspect its status. Collect a parent's final response before receiving its successor's
+`response.created`, which starts accumulation for the successor. The helper does not
+retain a history of completed responses; callers can retain the raw terminal events. Protocol errors raise `OpenAI::Responses::RequestError` with the
+original event in `error.event`. Raw commands other than `response.create` leave
+subsequent unsequenced, unscoped errors uncorrelated until explicit reconnect. Consuming
+such an error during an active default response fails that lane; named lanes remain
+usable. Errors received before the first raw submission retain ordinary create-retry
+behavior, regardless of later commands sent before consuming them. Observe `session.default` for connection-level
+errors and events with no registered lane. Closing a named lane discards its
+queue and routes future events for that ID to the default lane. Closing a lane
+does not cancel its server work. Its ID remains reserved until reconnect,
+including after terminal events: steering can create automatic successors.
+Continue using the same open lane for sequential responses. Closing the default
+lane opts out of unclaimed events; it cannot be replaced before reconnect.
+
+Recovery is explicit: `session.reconnect(client: fresh_client,
+request_options: fresh_options, restore: callback)` re-evaluates authentication
+and custom headers. Old lane handles raise `StateLostError`; the callback
+registers fresh lanes and sends application-selected restoration state. No
+previous create is replayed. Choose stored IDs or full history as described in
+[reconnect guidance](#reconnect-without-replaying-uncertain-work).
+
 ## Sequential tool turns
 
 The `tools` workflow forces a call to one known, harmless application function.
@@ -87,14 +156,45 @@ route a lane-scoped error to just that lane, while treating a connection-scoped
 error as affecting the connection. Track separate response IDs when multiple
 requests are outstanding within a lane; a lane alone is not a request ID.
 
-The server executes same-lane requests in FIFO order. The SDK does not enforce
-lane grammar, schedule lanes, or manage a per-lane queue. Consult the platform
-guide for the current concurrent-response and named-lane limits.
+The server executes same-lane requests in FIFO order without overlap. Named IDs
+contain 1–256 ASCII letters, digits, underscores, hyphens or periods; an empty string
+is invalid. The raw connection forwards IDs without enforcing this grammar.
+The optional [session helper](#managed-sessions)
+routes events into local queues and allows one consumed response at a time per
+lane; it does not schedule server work. Named-lane terminal events and request
+errors echo the ID; default-lane events omit it.
+
+The service allows 16 active responses and queues additional creates. It accepts
+32 distinct named IDs per connection; the default lane does not count. These
+are separate from the session's application budgets. Detaching a local lane
+does not reset the server's distinct-ID count.
 
-To fork, send an existing response ID on a different lane. With `store=false`,
+To fork, send an existing response ID on a different lane. With `store=false` or ZDR,
 wait for the fork's `response.in_progress` before advancing the source lane so
 its parent remains available in the connection-local cache.
 
+The server keeps recent response state in a connection-local cache. With
+`store=true`, an older response may be loaded from persisted state. With
+`store=false` or ZDR, an uncached ID returns `previous_response_not_found`.
+A same-lane continuation returning a 4xx or 5xx evicts its referenced cached
+parent; an errored cross-lane fork preserves the shared parent for the source
+lane. Do not blindly retry the same parent after a cache miss; use the explicit
+context-restoration approach below when needed.
+
+## Warmup and compaction
+
+For optional warmup, send `connection.response.create(..., generate: false)`,
+consume the terminal response, and save its ID. A later create can use that ID
+as `previous_response_id` with new input. Warmup prepares state without model
+output; the existing create helper forwards `generate` as an extra field.
+
+With automatic compaction configured through `context_management`, continue
+using the latest response ID and only new input items. Standalone
+`client.responses.compact(...)` instead returns a compacted input window. Send
+its complete `output` as input to a new WebSocket chain, omitting or nulling
+`previous_response_id`; do not use the compaction object's ID as a response ID
+or prune items from its output.
+
 ## Reconnect without replaying uncertain work
 
 Both reconnect examples intentionally close a healthy connection **after** a
@@ -154,15 +254,22 @@ acknowledgment, and it does not guarantee that a response stopped executing.
   event. Stop using that connection; do not automatically resend it.
 - A transport read failure poisons the connection. A server `error` event is
   delivered as an event and does not by itself poison it.
+  Handle `previous_response_not_found`, `invalid_stream_id`,
+  `websocket_stream_limit_reached`, and `websocket_connection_limit_reached`
+  through the original error event. Observe `session.default` when using named
+  session lanes so connection-scoped errors are not left unread.
 - Malformed JSON raises `ResponsesProtocolError` without exposing the payload;
   the connection permits later reads. Decide whether continuing is appropriate
   for your application.
 - Raw `send_event` supports generated client models or hashes, including
   `response.steer`. There is no `connection.response.steer` convenience method.
-- The platform documents `generate: false` warmup. It can be forwarded through
-  create's keyword rest even though 0.86.0 has no named generated `generate`
-  keyword. Follow the platform contract; do not send HTTP-only `stream` or
-  `background` flags just because generated models expose them.
+  Steering accepts only `type`, `previous_response_id`, and `input`, without
+  `stream_id`. Accepted input is committed at the successor's `response.created`.
+  If `response.steer.pending` requires tool output or approval, fill its
+  `required_input` stubs and send one create on the parent's lane with that
+  parent ID. Reuse saved results; do not rerun tools or resend accepted input.
+- Follow the platform contract; do not send HTTP-only `stream` or `background`
+  flags just because generated models expose them.
 - X.509 workload identity and provider runtimes are not supported by this
   entry point. A nonempty `request_options.extra_query` or nonzero
   `request_options.max_retries` is rejected. Supported workload identity has a

data/sig/openai/helpers/responses_websocket/connection.rbs

@@ -31,6 +31,11 @@ module OpenAI
           OpenAI::Models::Responses::connection_client_event event
         ) -> nil
 
+        # @api private
+        def encode_client_event: (
+          OpenAI::Models::Responses::connection_client_event event
+        ) -> ::Hash[Symbol, untyped]
+
         # @api private
         def send_raw: (String data) -> nil
 

data/sig/openai/helpers/responses_websocket/session.rbs

@@ -0,0 +1,123 @@
+module OpenAI
+  module Models
+    module Responses
+      class SessionError < OpenAI::Errors::Error
+        def cause: -> nil
+      end
+
+      class BufferError < SessionError
+      end
+
+      class StateLostError < SessionError
+      end
+
+      class RequestError < SessionError
+        attr_reader event: connection_server_event
+
+        # @api private
+        def initialize: (connection_server_event event) -> void
+      end
+
+      class SessionLimits
+        attr_reader max_lanes: Integer
+        attr_reader max_events_per_lane: Integer
+        attr_reader max_events: Integer
+        attr_reader max_bytes_per_lane: Integer
+        attr_reader max_bytes: Integer
+        attr_reader max_response_bytes: Integer
+        def initialize: (
+          max_lanes: Integer,
+          max_events_per_lane: Integer,
+          max_events: Integer,
+          max_bytes_per_lane: Integer,
+          max_bytes: Integer,
+          max_response_bytes: Integer
+        ) -> void
+      end
+
+      class SessionLane
+        attr_reader stream_id: String?
+        def send_event: (connection_client_event event) -> void
+        def receive: -> connection_server_event
+        def get_final_response: -> OpenAI::Responses::Response
+        def close: -> void
+
+        # @api private
+        def initialize: (
+          Session session,
+          String? stream_id,
+          Hash[String, untyped] pending_steers
+        ) -> void
+
+        # @api private
+        def enqueue: (
+          connection_server_event event,
+          Integer bytes,
+          ?bool raw_command_sent
+        ) -> bool?
+
+        # @api private
+        def discard: (?Exception? error) -> void
+
+        # @api private
+        def wake: -> void
+      end
+
+      class Session
+        attr_reader limits: SessionLimits
+        attr_reader default: SessionLane
+        def self.open: [Result] (
+          client: OpenAI::Client,
+          limits: SessionLimits,
+          ?request_options: OpenAI::request_opts?,
+          ?websocket_base_url: String?,
+          ?transport: top,
+          ?transport_options: ::Hash[Symbol, top]
+        ) {
+          (Session session) -> Result
+        } -> Result
+        def lane: (String? stream_id) -> SessionLane
+        def reconnect: (
+          restore: ^(Session) -> top,
+          ?client: OpenAI::Client,
+          ?request_options: OpenAI::request_opts?
+        ) -> self
+        def close: -> void
+
+        # @api private
+        def initialize: (
+          task: untyped,
+          client: OpenAI::Client,
+          limits: SessionLimits,
+          request_options: OpenAI::request_opts?,
+          websocket_base_url: String?,
+          transport: top,
+          transport_options: ::Hash[Symbol, top]
+        ) -> void
+
+        # @api private
+        def start: -> void
+
+        # @api private
+        def encode_client_event: (
+          connection_client_event event
+        ) -> ::Hash[Symbol, untyped]
+
+        # @api private
+        def send_event: (connection_client_event event) -> void
+
+        # @api private
+        def check_reader!: -> void
+
+        # @api private
+        def assert_owner!: -> void
+
+        # @api private
+        def consumed: (Integer bytes) -> void
+
+        # @api private
+        def detach: (SessionLane value) -> void
+      end
+    end
+  end
+end

data/sig/openai/models/beta/agents/session_update_params.rbs

@@ -5,6 +5,7 @@ module OpenAI
         type session_update_params =
           {
             session_id: String,
+            ?agent: OpenAI::Beta::Agents::SessionUpdateParams::Agent,
             ?metadata: ::Hash[Symbol, String]?,
             ?request_options: OpenAI::request_opts
           }
@@ -15,19 +16,117 @@ module OpenAI
 
           attr_accessor session_id: String
 
+          attr_reader agent: OpenAI::Beta::Agents::SessionUpdateParams::Agent?
+
+          def agent=: (
+            OpenAI::Beta::Agents::SessionUpdateParams::Agent
+          ) -> OpenAI::Beta::Agents::SessionUpdateParams::Agent
+
           attr_accessor metadata: ::Hash[Symbol, String]?
 
           def initialize: (
             session_id: String,
+            ?agent: OpenAI::Beta::Agents::SessionUpdateParams::Agent,
             ?metadata: ::Hash[Symbol, String]?,
             ?request_options: OpenAI::request_opts
           ) -> void
 
           def to_hash: -> {
             session_id: String,
+            ?agent: OpenAI::Beta::Agents::SessionUpdateParams::Agent,
             ?metadata: ::Hash[Symbol, String]?,
             ?request_options: OpenAI::RequestOptions
           }
+
+          type agent =
+            {
+              ?model: String,
+              ?reasoning: OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning,
+              ?service_tier: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::service_tier?
+            }
+
+          class Agent < OpenAI::Internal::Type::BaseModel
+            attr_reader model: String?
+
+            def model=: (String) -> String
+
+            attr_reader reasoning: OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning?
+
+            def reasoning=: (
+              OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning
+            ) -> OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning
+
+            attr_accessor service_tier: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::service_tier?
+
+            def initialize: (
+              ?model: String,
+              ?reasoning: OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning,
+              ?service_tier: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::service_tier?
+            ) -> void
+
+            def to_hash: -> {
+              ?model: String,
+              ?reasoning: OpenAI::Beta::Agents::SessionUpdateParams::Agent::Reasoning,
+              ?service_tier: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::service_tier?
+            }
+
+            type reasoning =
+              {
+                ?effort: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::Reasoning::effort?
+              }
+
+            class Reasoning < OpenAI::Internal::Type::BaseModel
+              attr_accessor effort: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::Reasoning::effort?
+
+              def initialize: (
+                ?effort: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::Reasoning::effort?
+              ) -> void
+
+              def to_hash: -> {
+                ?effort: OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::Reasoning::effort?
+              }
+
+              type effort =
+                :none | :minimal | :low | :medium | :high | :xhigh | :max
+
+              module Effort
+                extend OpenAI::Internal::Type::Enum
+
+                NONE: :none
+                MINIMAL: :minimal
+                LOW: :low
+                MEDIUM: :medium
+                HIGH: :high
+                XHIGH: :xhigh
+                MAX: :max
+
+                def self?.values: -> ::Array[OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::Reasoning::effort]
+              end
+            end
+
+            type service_tier = :auto | :default | :flex | :priority | :fast
+
+            module ServiceTier
+              extend OpenAI::Internal::Type::Enum
+
+              # Selects the service tier automatically.
+              AUTO: :auto
+
+              # Uses the default service tier.
+              DEFAULT: :default
+
+              # Uses the flex service tier.
+              FLEX: :flex
+
+              # Uses the priority service tier.
+              PRIORITY: :priority
+
+              # Uses the fast service tier.
+              FAST: :fast
+
+              def self?.values: -> ::Array[OpenAI::Models::Beta::Agents::SessionUpdateParams::Agent::service_tier]
+            end
+          end
         end
       end
     end

data/sig/openai/models/beta/beta_response.rbs

@@ -278,7 +278,6 @@ module OpenAI
           | :"gpt-5.1"
           | :"gpt-5.1-2025-11-13"
           | :"gpt-5.1-codex"
-          | :"gpt-5.1-mini"
           | :"gpt-5.1-chat-latest"
           | :"gpt-5"
           | :"gpt-5-mini"
@@ -309,6 +308,8 @@ module OpenAI
           | :"gpt-4o-2024-11-20"
           | :"gpt-4o-2024-08-06"
           | :"gpt-4o-2024-05-13"
+          | :"gpt-audio-mini"
+          | :"gpt-audio-mini-2025-12-15"
           | :"gpt-4o-audio-preview"
           | :"gpt-4o-audio-preview-2024-10-01"
           | :"gpt-4o-audio-preview-2024-12-17"
@@ -342,6 +343,7 @@ module OpenAI
           | :"gpt-3.5-turbo-1106"
           | :"gpt-3.5-turbo-0125"
           | :"gpt-3.5-turbo-16k-0613"
+          | :"gpt-5.1-mini"
           | :"o1-pro"
           | :"o1-pro-2025-03-19"
           | :"o3-pro"
@@ -388,7 +390,6 @@ module OpenAI
           GPT_5_1: :"gpt-5.1"
           GPT_5_1_2025_11_13: :"gpt-5.1-2025-11-13"
           GPT_5_1_CODEX: :"gpt-5.1-codex"
-          GPT_5_1_MINI: :"gpt-5.1-mini"
           GPT_5_1_CHAT_LATEST: :"gpt-5.1-chat-latest"
           GPT_5: :"gpt-5"
           GPT_5_MINI: :"gpt-5-mini"
@@ -419,6 +420,8 @@ module OpenAI
           GPT_4O_2024_11_20: :"gpt-4o-2024-11-20"
           GPT_4O_2024_08_06: :"gpt-4o-2024-08-06"
           GPT_4O_2024_05_13: :"gpt-4o-2024-05-13"
+          GPT_AUDIO_MINI: :"gpt-audio-mini"
+          GPT_AUDIO_MINI_2025_12_15: :"gpt-audio-mini-2025-12-15"
           GPT_4O_AUDIO_PREVIEW: :"gpt-4o-audio-preview"
           GPT_4O_AUDIO_PREVIEW_2024_10_01: :"gpt-4o-audio-preview-2024-10-01"
           GPT_4O_AUDIO_PREVIEW_2024_12_17: :"gpt-4o-audio-preview-2024-12-17"
@@ -452,6 +455,7 @@ module OpenAI
           GPT_3_5_TURBO_1106: :"gpt-3.5-turbo-1106"
           GPT_3_5_TURBO_0125: :"gpt-3.5-turbo-0125"
           GPT_3_5_TURBO_16K_0613: :"gpt-3.5-turbo-16k-0613"
+          GPT_5_1_MINI: :"gpt-5.1-mini"
           O1_PRO: :"o1-pro"
           O1_PRO_2025_03_19: :"o1-pro-2025-03-19"
           O3_PRO: :"o3-pro"

data/sig/openai/models/beta/beta_response_compaction_compacting_event.rbs

@@ -0,0 +1,58 @@
+module OpenAI
+  module Models
+    class BetaResponseCompactionCompactingEvent = Beta::BetaResponseCompactionCompactingEvent
+
+    module Beta
+      type beta_response_compaction_compacting_event =
+        {
+          item_id: String,
+          output_index: Integer,
+          sequence_number: Integer,
+          ?type: :"response.compaction.compacting",
+          ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent
+        }
+
+      class BetaResponseCompactionCompactingEvent < OpenAI::Internal::Type::BaseModel
+        attr_accessor item_id: String
+
+        attr_accessor output_index: Integer
+
+        attr_accessor sequence_number: Integer
+
+        attr_accessor type: :"response.compaction.compacting"
+
+        attr_reader agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent?
+
+        def agent=: (
+          OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent
+        ) -> OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent
+
+        def initialize: (
+          item_id: String,
+          output_index: Integer,
+          sequence_number: Integer,
+          ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent,
+          ?type: :"response.compaction.compacting"
+        ) -> void
+
+        def to_hash: -> {
+          item_id: String,
+          output_index: Integer,
+          sequence_number: Integer,
+          ?type: :"response.compaction.compacting",
+          ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent
+        }
+
+        type agent = { agent_name: String }
+
+        class Agent < OpenAI::Internal::Type::BaseModel
+          attr_accessor agent_name: String
+
+          def initialize: (agent_name: String) -> void
+
+          def to_hash: -> { agent_name: String }
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/beta/beta_response_stream_event.rbs

@@ -13,6 +13,7 @@ module OpenAI
         | OpenAI::Beta::BetaResponseCodeInterpreterCallCompletedEvent
         | OpenAI::Beta::BetaResponseCodeInterpreterCallInProgressEvent
         | OpenAI::Beta::BetaResponseCodeInterpreterCallInterpretingEvent
+        | OpenAI::Beta::BetaResponseCompactionCompactingEvent
         | OpenAI::Beta::BetaResponseCompletedEvent
         | OpenAI::Beta::BetaResponseContentPartAddedEvent
         | OpenAI::Beta::BetaResponseContentPartDoneEvent

data/sig/openai/models/beta/beta_responses_client_event.rbs

@@ -271,7 +271,6 @@ module OpenAI
             | :"gpt-5.1"
             | :"gpt-5.1-2025-11-13"
             | :"gpt-5.1-codex"
-            | :"gpt-5.1-mini"
             | :"gpt-5.1-chat-latest"
             | :"gpt-5"
             | :"gpt-5-mini"
@@ -302,6 +301,8 @@ module OpenAI
             | :"gpt-4o-2024-11-20"
             | :"gpt-4o-2024-08-06"
             | :"gpt-4o-2024-05-13"
+            | :"gpt-audio-mini"
+            | :"gpt-audio-mini-2025-12-15"
             | :"gpt-4o-audio-preview"
             | :"gpt-4o-audio-preview-2024-10-01"
             | :"gpt-4o-audio-preview-2024-12-17"
@@ -335,6 +336,7 @@ module OpenAI
             | :"gpt-3.5-turbo-1106"
             | :"gpt-3.5-turbo-0125"
             | :"gpt-3.5-turbo-16k-0613"
+            | :"gpt-5.1-mini"
             | :"o1-pro"
             | :"o1-pro-2025-03-19"
             | :"o3-pro"
@@ -381,7 +383,6 @@ module OpenAI
             GPT_5_1: :"gpt-5.1"
             GPT_5_1_2025_11_13: :"gpt-5.1-2025-11-13"
             GPT_5_1_CODEX: :"gpt-5.1-codex"
-            GPT_5_1_MINI: :"gpt-5.1-mini"
             GPT_5_1_CHAT_LATEST: :"gpt-5.1-chat-latest"
             GPT_5: :"gpt-5"
             GPT_5_MINI: :"gpt-5-mini"
@@ -412,6 +413,8 @@ module OpenAI
             GPT_4O_2024_11_20: :"gpt-4o-2024-11-20"
             GPT_4O_2024_08_06: :"gpt-4o-2024-08-06"
             GPT_4O_2024_05_13: :"gpt-4o-2024-05-13"
+            GPT_AUDIO_MINI: :"gpt-audio-mini"
+            GPT_AUDIO_MINI_2025_12_15: :"gpt-audio-mini-2025-12-15"
             GPT_4O_AUDIO_PREVIEW: :"gpt-4o-audio-preview"
             GPT_4O_AUDIO_PREVIEW_2024_10_01: :"gpt-4o-audio-preview-2024-10-01"
             GPT_4O_AUDIO_PREVIEW_2024_12_17: :"gpt-4o-audio-preview-2024-12-17"
@@ -445,6 +448,7 @@ module OpenAI
             GPT_3_5_TURBO_1106: :"gpt-3.5-turbo-1106"
             GPT_3_5_TURBO_0125: :"gpt-3.5-turbo-0125"
             GPT_3_5_TURBO_16K_0613: :"gpt-3.5-turbo-16k-0613"
+            GPT_5_1_MINI: :"gpt-5.1-mini"
             O1_PRO: :"o1-pro"
             O1_PRO_2025_03_19: :"o1-pro-2025-03-19"
             O3_PRO: :"o3-pro"
@@ -591,6 +595,7 @@ module OpenAI
             {
               ?comparison_response_id: String?,
               ?mode: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::mode,
+              ?prewarm: bool,
               ?ttl: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::ttl
             }
 
@@ -603,6 +608,10 @@ module OpenAI
               OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::mode
             ) -> OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::mode
 
+            attr_reader prewarm: bool?
+
+            def prewarm=: (bool) -> bool
+
             attr_reader ttl: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::ttl?
 
             def ttl=: (
@@ -612,12 +621,14 @@ module OpenAI
             def initialize: (
               ?comparison_response_id: String?,
               ?mode: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::mode,
+              ?prewarm: bool,
               ?ttl: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::ttl
             ) -> void
 
             def to_hash: -> {
               ?comparison_response_id: String?,
               ?mode: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::mode,
+              ?prewarm: bool,
               ?ttl: OpenAI::Models::Beta::BetaResponsesClientEvent::ResponseCreate::PromptCacheOptions::ttl
             }
 

data/sig/openai/models/beta/beta_responses_server_event.rbs

@@ -19,6 +19,7 @@ module OpenAI
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseCodeInterpreterCallWsCompleted
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseCodeInterpreterCallInWsProgress
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseCodeInterpreterCallWsInterpreting
+        | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseCompactionWsCompacting
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseWsCompleted
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseContentPartWsAdded
         | OpenAI::Beta::BetaResponsesServerEvent::BetaResponseContentPartWsDone
@@ -365,6 +366,40 @@ module OpenAI
           }
         end
 
+        type beta_response_compaction_ws_compacting =
+          {
+            item_id: String,
+            output_index: Integer,
+            sequence_number: Integer,
+            ?type: :"response.compaction.compacting",
+            ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent,
+            ?stream_id: String
+          }
+
+        class BetaResponseCompactionWsCompacting < OpenAI::Models::Beta::BetaResponseCompactionCompactingEvent
+          def stream_id: -> String?
+
+          def stream_id=: (String _) -> String
+
+          def initialize: (
+            item_id: String,
+            output_index: Integer,
+            sequence_number: Integer,
+            ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent,
+            ?stream_id: String,
+            ?type: :"response.compaction.compacting"
+          ) -> void
+
+          def to_hash: -> {
+            item_id: String,
+            output_index: Integer,
+            sequence_number: Integer,
+            ?type: :"response.compaction.compacting",
+            ?agent: OpenAI::Beta::BetaResponseCompactionCompactingEvent::Agent,
+            ?stream_id: String
+          }
+        end
+
         type beta_response_ws_completed =
           {
             response: OpenAI::Beta::BetaResponse,