deeprails 0.18.0 → 0.23.0

data/rbi/deeprails/models/workflow_event_detail_response.rbi

@@ -11,6 +11,9 @@ module Deeprails
           )
         end
 
+      sig { returns(T::Array[String]) }
+      attr_accessor :analysis_of_failures
+
       # History of evaluations for the event.
       sig do
         returns(
@@ -45,7 +48,12 @@ module Deeprails
       end
       attr_accessor :improvement_action
 
-      # Status of the improvement tool used to improve the event.
+      # Status of the improvement tool used to improve the event. `improvement_required`
+      # indicates that the evaluation is complete and the improvement action is needed
+      # but is not taking place. `improved` and `improvement_failed` indicate when the
+      # improvement action concludes, successfully and unsuccessfully, respectively.
+      # `no_improvement_required` means that the first evaluation passed all its
+      # metrics!
       sig do
         returns(
           T.nilable(
@@ -55,6 +63,9 @@ module Deeprails
       end
       attr_accessor :improvement_tool_status
 
+      sig { returns(T::Array[T.anything]) }
+      attr_accessor :key_improvements
+
       # Status of the event.
       sig do
         returns(Deeprails::WorkflowEventDetailResponse::Status::TaggedSymbol)
@@ -98,8 +109,8 @@ module Deeprails
       end
       attr_writer :automatic_hallucination_tolerance_levels
 
-      # Extended AI capabilities available to the event, if any. Can be `web_search`
-      # and/or `file_search`.
+      # Extended AI capabilities available to the event, if any. Can be `web_search`,
+      # `file_search`, and/or `context_awareness`.
       sig do
         returns(
           T.nilable(
@@ -145,8 +156,17 @@ module Deeprails
       end
       attr_writer :files
 
+      # The maximum number of improvement attempts to be applied to one event before it
+      # is considered failed.
+      sig { returns(T.nilable(Integer)) }
+      attr_reader :max_improvement_attempts
+
+      sig { params(max_improvement_attempts: Integer).void }
+      attr_writer :max_improvement_attempts
+
       sig do
         params(
+          analysis_of_failures: T::Array[String],
           evaluation_history:
             T::Array[
               Deeprails::WorkflowEventDetailResponse::EvaluationHistory::OrHash
@@ -161,6 +181,7 @@ module Deeprails
             T.nilable(
               Deeprails::WorkflowEventDetailResponse::ImprovementToolStatus::OrSymbol
             ),
+          key_improvements: T::Array[T.anything],
           status: Deeprails::WorkflowEventDetailResponse::Status::OrSymbol,
           threshold_type:
             Deeprails::WorkflowEventDetailResponse::ThresholdType::OrSymbol,
@@ -175,10 +196,12 @@ module Deeprails
               Deeprails::WorkflowEventDetailResponse::Capability::OrHash
             ],
           custom_hallucination_threshold_values: T::Hash[Symbol, Float],
-          files: T::Array[Deeprails::WorkflowEventDetailResponse::File::OrHash]
+          files: T::Array[Deeprails::WorkflowEventDetailResponse::File::OrHash],
+          max_improvement_attempts: Integer
         ).returns(T.attached_class)
       end
       def self.new(
+        analysis_of_failures:,
         # History of evaluations for the event.
         evaluation_history:,
         # Evaluation result consisting of average scores and rationales for each of the
@@ -193,8 +216,14 @@ module Deeprails
         improved_model_output:,
         # Type of improvement action used to improve the event.
         improvement_action:,
-        # Status of the improvement tool used to improve the event.
+        # Status of the improvement tool used to improve the event. `improvement_required`
+        # indicates that the evaluation is complete and the improvement action is needed
+        # but is not taking place. `improved` and `improvement_failed` indicate when the
+        # improvement action concludes, successfully and unsuccessfully, respectively.
+        # `no_improvement_required` means that the first evaluation passed all its
+        # metrics!
         improvement_tool_status:,
+        key_improvements:,
         # Status of the event.
         status:,
         # Type of thresholds used to evaluate the event.
@@ -204,21 +233,25 @@ module Deeprails
         # Mapping of guardrail metric names to tolerance values. Values are strings
         # (`low`, `medium`, `high`) representing automatic tolerance levels.
         automatic_hallucination_tolerance_levels: nil,
-        # Extended AI capabilities available to the event, if any. Can be `web_search`
-        # and/or `file_search`.
+        # Extended AI capabilities available to the event, if any. Can be `web_search`,
+        # `file_search`, and/or `context_awareness`.
         capabilities: nil,
         # Mapping of guardrail metric names to threshold values. Values are floating point
         # numbers (0.0-1.0) representing custom thresholds.
         custom_hallucination_threshold_values: nil,
         # List of files available to the event, if any. Will only be present if
         # `file_search` is enabled.
-        files: nil
+        files: nil,
+        # The maximum number of improvement attempts to be applied to one event before it
+        # is considered failed.
+        max_improvement_attempts: nil
       )
       end
 
       sig do
         override.returns(
           {
+            analysis_of_failures: T::Array[String],
             evaluation_history:
               T::Array[
                 Deeprails::WorkflowEventDetailResponse::EvaluationHistory
@@ -233,6 +266,7 @@ module Deeprails
               T.nilable(
                 Deeprails::WorkflowEventDetailResponse::ImprovementToolStatus::TaggedSymbol
               ),
+            key_improvements: T::Array[T.anything],
             status:
               Deeprails::WorkflowEventDetailResponse::Status::TaggedSymbol,
             threshold_type:
@@ -246,7 +280,8 @@ module Deeprails
             capabilities:
               T::Array[Deeprails::WorkflowEventDetailResponse::Capability],
             custom_hallucination_threshold_values: T::Hash[Symbol, Float],
-            files: T::Array[Deeprails::WorkflowEventDetailResponse::File]
+            files: T::Array[Deeprails::WorkflowEventDetailResponse::File],
+            max_improvement_attempts: Integer
           }
         )
       end
@@ -262,6 +297,12 @@ module Deeprails
             )
           end
 
+        sig { returns(T.nilable(String)) }
+        attr_reader :analysis_of_failures
+
+        sig { params(analysis_of_failures: String).void }
+        attr_writer :analysis_of_failures
+
         sig { returns(T.nilable(String)) }
         attr_reader :attempt
 
@@ -304,6 +345,29 @@ module Deeprails
         sig { params(guardrail_metrics: T::Array[String]).void }
         attr_writer :guardrail_metrics
 
+        sig do
+          returns(
+            T.nilable(
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+            )
+          )
+        end
+        attr_reader :improvement_tool_status
+
+        sig do
+          params(
+            improvement_tool_status:
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::OrSymbol
+          ).void
+        end
+        attr_writer :improvement_tool_status
+
+        sig { returns(T.nilable(T::Array[String])) }
+        attr_reader :key_improvements
+
+        sig { params(key_improvements: T::Array[String]).void }
+        attr_writer :key_improvements
+
         sig { returns(T.nilable(T::Hash[Symbol, T.anything])) }
         attr_reader :model_input
 
@@ -316,12 +380,6 @@ module Deeprails
         sig { params(model_output: String).void }
         attr_writer :model_output
 
-        sig { returns(T.nilable(Time)) }
-        attr_reader :modified_at
-
-        sig { params(modified_at: Time).void }
-        attr_writer :modified_at
-
         sig { returns(T.nilable(String)) }
         attr_reader :nametag
 
@@ -342,6 +400,7 @@ module Deeprails
 
         sig do
           params(
+            analysis_of_failures: String,
             attempt: String,
             created_at: Time,
             error_message: String,
@@ -349,15 +408,18 @@ module Deeprails
             evaluation_status: String,
             evaluation_total_cost: Float,
             guardrail_metrics: T::Array[String],
+            improvement_tool_status:
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::OrSymbol,
+            key_improvements: T::Array[String],
             model_input: T::Hash[Symbol, T.anything],
             model_output: String,
-            modified_at: Time,
             nametag: String,
             progress: Integer,
             run_mode: String
           ).returns(T.attached_class)
         end
         def self.new(
+          analysis_of_failures: nil,
           attempt: nil,
           created_at: nil,
           error_message: nil,
@@ -365,9 +427,10 @@ module Deeprails
           evaluation_status: nil,
           evaluation_total_cost: nil,
           guardrail_metrics: nil,
+          improvement_tool_status: nil,
+          key_improvements: nil,
           model_input: nil,
           model_output: nil,
-          modified_at: nil,
           nametag: nil,
           progress: nil,
           run_mode: nil
@@ -377,6 +440,7 @@ module Deeprails
         sig do
           override.returns(
             {
+              analysis_of_failures: String,
               attempt: String,
               created_at: Time,
               error_message: String,
@@ -384,9 +448,11 @@ module Deeprails
               evaluation_status: String,
               evaluation_total_cost: Float,
               guardrail_metrics: T::Array[String],
+              improvement_tool_status:
+                Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol,
+              key_improvements: T::Array[String],
               model_input: T::Hash[Symbol, T.anything],
               model_output: String,
-              modified_at: Time,
               nametag: String,
               progress: Integer,
               run_mode: String
@@ -395,6 +461,50 @@ module Deeprails
         end
         def to_hash
         end
+
+        module ImprovementToolStatus
+          extend Deeprails::Internal::Type::Enum
+
+          TaggedSymbol =
+            T.type_alias do
+              T.all(
+                Symbol,
+                Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus
+              )
+            end
+          OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+          IMPROVED =
+            T.let(
+              :improved,
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+            )
+          IMPROVEMENT_FAILED =
+            T.let(
+              :improvement_failed,
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+            )
+          NO_IMPROVEMENT_REQUIRED =
+            T.let(
+              :no_improvement_required,
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+            )
+          IMPROVEMENT_REQUIRED =
+            T.let(
+              :improvement_required,
+              Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+            )
+
+          sig do
+            override.returns(
+              T::Array[
+                Deeprails::WorkflowEventDetailResponse::EvaluationHistory::ImprovementToolStatus::TaggedSymbol
+              ]
+            )
+          end
+          def self.values
+          end
+        end
       end
 
       # Type of improvement action used to improve the event.
@@ -437,7 +547,12 @@ module Deeprails
         end
       end
 
-      # Status of the improvement tool used to improve the event.
+      # Status of the improvement tool used to improve the event. `improvement_required`
+      # indicates that the evaluation is complete and the improvement action is needed
+      # but is not taking place. `improved` and `improvement_failed` indicate when the
+      # improvement action concludes, successfully and unsuccessfully, respectively.
+      # `no_improvement_required` means that the first evaluation passed all its
+      # metrics!
       module ImprovementToolStatus
         extend Deeprails::Internal::Type::Enum
 
@@ -455,9 +570,14 @@ module Deeprails
             :improved,
             Deeprails::WorkflowEventDetailResponse::ImprovementToolStatus::TaggedSymbol
           )
-        FAILED_ON_MAX_RETRIES =
+        IMPROVEMENT_FAILED =
           T.let(
-            :"failed on max retries",
+            :improvement_failed,
+            Deeprails::WorkflowEventDetailResponse::ImprovementToolStatus::TaggedSymbol
+          )
+        NO_IMPROVEMENT_REQUIRED =
+          T.let(
+            :no_improvement_required,
             Deeprails::WorkflowEventDetailResponse::ImprovementToolStatus::TaggedSymbol
           )
         IMPROVEMENT_REQUIRED =
@@ -631,19 +751,45 @@ module Deeprails
         sig { params(file_size: Integer).void }
         attr_writer :file_size
 
+        sig { returns(T.nilable(String)) }
+        attr_reader :presigned_url
+
+        sig { params(presigned_url: String).void }
+        attr_writer :presigned_url
+
+        sig { returns(T.nilable(Time)) }
+        attr_reader :presigned_url_expires_at
+
+        sig { params(presigned_url_expires_at: Time).void }
+        attr_writer :presigned_url_expires_at
+
         sig do
           params(
             file_id: String,
             file_name: String,
-            file_size: Integer
+            file_size: Integer,
+            presigned_url: String,
+            presigned_url_expires_at: Time
           ).returns(T.attached_class)
         end
-        def self.new(file_id: nil, file_name: nil, file_size: nil)
+        def self.new(
+          file_id: nil,
+          file_name: nil,
+          file_size: nil,
+          presigned_url: nil,
+          presigned_url_expires_at: nil
+        )
         end
 
         sig do
           override.returns(
-            { file_id: String, file_name: String, file_size: Integer }
+            {
+              file_id: String,
+              file_name: String,
+              file_size: Integer,
+              presigned_url: String,
+              presigned_url_expires_at: Time
+            }
           )
         end
         def to_hash

data/rbi/deeprails/models/workflow_event_response.rbi

@@ -8,6 +8,10 @@ module Deeprails
           T.any(Deeprails::WorkflowEventResponse, Deeprails::Internal::AnyHash)
         end
 
+      # The ID of the billing request for the event.
+      sig { returns(String) }
+      attr_accessor :billing_request_id
+
       # The time the event was created in UTC.
       sig { returns(Time) }
       attr_accessor :created_at
@@ -26,6 +30,7 @@ module Deeprails
 
       sig do
         params(
+          billing_request_id: String,
           created_at: Time,
           event_id: String,
           status: Deeprails::WorkflowEventResponse::Status::OrSymbol,
@@ -33,6 +38,8 @@ module Deeprails
         ).returns(T.attached_class)
       end
       def self.new(
+        # The ID of the billing request for the event.
+        billing_request_id:,
         # The time the event was created in UTC.
         created_at:,
         # A unique workflow event ID.
@@ -47,6 +54,7 @@ module Deeprails
       sig do
         override.returns(
           {
+            billing_request_id: String,
             created_at: Time,
             event_id: String,
             status: Deeprails::WorkflowEventResponse::Status::TaggedSymbol,

data/rbi/deeprails/models.rbi

@@ -11,6 +11,9 @@ module Deeprails
 
   DefendRetrieveWorkflowParams = Deeprails::Models::DefendRetrieveWorkflowParams
 
+  DefendSubmitAndStreamEventParams =
+    Deeprails::Models::DefendSubmitAndStreamEventParams
+
   DefendSubmitEventParams = Deeprails::Models::DefendSubmitEventParams
 
   DefendUpdateResponse = Deeprails::Models::DefendUpdateResponse

data/rbi/deeprails/resources/defend.rbi

@@ -17,6 +17,7 @@ module Deeprails
               Symbol,
               Deeprails::DefendCreateWorkflowParams::AutomaticHallucinationToleranceLevel::OrSymbol
             ],
+          context_awareness: T::Boolean,
           custom_hallucination_threshold_values: T::Hash[Symbol, Float],
           description: String,
           file_search: T::Array[String],
@@ -45,6 +46,13 @@ module Deeprails
         # `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
         # `comprehensive_safety`.
         automatic_hallucination_tolerance_levels: nil,
+        # Context includes any structured information that directly relates to the model’s
+        # input and expected output—e.g., the recent turn-by-turn history between an AI
+        # tutor and a student, facts or state passed through an agentic workflow, or other
+        # domain-specific signals your system already knows and wants the model to
+        # condition on. This field determines whether to enable context awareness for this
+        # workflow's evaluations. Defaults to false.
+        context_awareness: nil,
         # Mapping of guardrail metrics to floating point threshold values. Possible
         # metrics are `correctness`, `completeness`, `instruction_adherence`,
         # `context_adherence`, `ground_truth_adherence`, or `comprehensive_safety`.
@@ -98,6 +106,42 @@ module Deeprails
       )
       end
 
+      # Use this endpoint to create a new event for a guardrail workflow with real-time
+      # streaming feedback via Server-Sent Events (SSE).
+      sig do
+        params(
+          workflow_id: String,
+          model_input: T::Hash[Symbol, T.anything],
+          model_output: String,
+          model_used: String,
+          run_mode:
+            Deeprails::DefendSubmitAndStreamEventParams::RunMode::OrSymbol,
+          stream: T::Boolean,
+          nametag: String,
+          request_options: Deeprails::RequestOptions::OrHash
+        ).returns(Deeprails::Internal::Stream[String])
+      end
+      def submit_and_stream_event_streaming(
+        # Path param: The ID of the workflow to create the event for.
+        workflow_id,
+        # Body param: The input provided to the model (e.g., prompt, messages).
+        model_input:,
+        # Body param: The output generated by the model to be evaluated.
+        model_output:,
+        # Body param: The model that generated the output (e.g., "gpt-4", "claude-3").
+        model_used:,
+        # Body param: The evaluation run mode. Streaming only supports fast, precision,
+        # and precision_codex.
+        run_mode:,
+        # Query param: Enable SSE streaming for real-time token feedback. Only supported
+        # for single-model run modes (fast, precision, precision_codex).
+        stream: nil,
+        # Body param: Optional tag to identify this event.
+        nametag: nil,
+        request_options: {}
+      )
+      end
+
       # Use this endpoint to submit a model input and output pair to a workflow for
       # evaluation
       sig do
@@ -115,8 +159,8 @@ module Deeprails
         # Workflow ID associated with this event.
         workflow_id,
         # A dictionary of inputs sent to the LLM to generate output. The dictionary must
-        # contain at least a `user_prompt` field or a `system_prompt` field. For the
-        # ground_truth_adherence guardrail metric, `ground_truth` should be provided.
+        # contain a `user_prompt` field. For the ground_truth_adherence guardrail metric,
+        # `ground_truth` should be provided.
         model_input:,
         # Output generated by the LLM to be evaluated.
         model_output:,
@@ -124,8 +168,8 @@ module Deeprails
         model_used:,
         # Run mode for the workflow event. The run mode allows the user to optimize for
         # speed, accuracy, and cost by determining which models are used to evaluate the
-        # event. Available run modes include `precision_plus`, `precision`, `smart`, and
-        # `economy`. Defaults to `smart`.
+        # event. Available run modes include `precision_plus_codex`, `precision_plus`,
+        # `precision`, `smart`, and `economy`. Defaults to `smart`.
         run_mode:,
         # An optional, user-defined tag for the event.
         nametag: nil,
@@ -137,18 +181,64 @@ module Deeprails
       sig do
         params(
           workflow_id: String,
+          automatic_hallucination_tolerance_levels:
+            T::Hash[
+              Symbol,
+              Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::OrSymbol
+            ],
+          context_awareness: T::Boolean,
+          custom_hallucination_threshold_values: T::Hash[Symbol, Float],
           description: String,
+          file_search: T::Array[String],
+          improvement_action:
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::OrSymbol,
+          max_improvement_attempts: Integer,
           name: String,
+          threshold_type:
+            Deeprails::DefendUpdateWorkflowParams::ThresholdType::OrSymbol,
+          web_search: T::Boolean,
           request_options: Deeprails::RequestOptions::OrHash
         ).returns(Deeprails::DefendUpdateResponse)
       end
       def update_workflow(
         # The ID of the workflow to edit.
         workflow_id,
-        # Description for the workflow.
+        # New mapping of guardrail metrics to hallucination tolerance levels (either
+        # `low`, `medium`, or `high`) to be used when `threshold_type` is set to
+        # `automatic`. Possible metrics are `completeness`, `instruction_adherence`,
+        # `context_adherence`, `ground_truth_adherence`, or `comprehensive_safety`.
+        automatic_hallucination_tolerance_levels: nil,
+        # Whether to enable context awareness for this workflow's evaluations.
+        context_awareness: nil,
+        # New mapping of guardrail metrics to floating point threshold values to be used
+        # when `threshold_type` is set to `custom`. Possible metrics are `correctness`,
+        # `completeness`, `instruction_adherence`, `context_adherence`,
+        # `ground_truth_adherence`, or `comprehensive_safety`.
+        custom_hallucination_threshold_values: nil,
+        # New description for the workflow.
         description: nil,
-        # Name of the workflow.
+        # An array of file IDs to search in the workflow's evaluations. Files must be
+        # uploaded via the DeepRails API first.
+        file_search: nil,
+        # The new action used to improve outputs that fail one or more guardrail metrics
+        # for the workflow events. May be `regen`, `fixit`, or `do_nothing`. ReGen runs
+        # the user's input prompt with minor induced variance. FixIt attempts to directly
+        # address the shortcomings of the output using the guardrail failure rationale. Do
+        # Nothing does not attempt any improvement.
+        improvement_action: nil,
+        # Max. number of improvement action attempts until a given event passes the
+        # guardrails. Defaults to 10.
+        max_improvement_attempts: nil,
+        # New name for the workflow.
         name: nil,
+        # New type of thresholds to use for the workflow, either `automatic` or `custom`.
+        # Automatic thresholds are assigned internally after the user specifies a
+        # qualitative tolerance for the metrics, whereas custom metrics allow the user to
+        # set the threshold for each metric as a floating point number between 0.0 and
+        # 1.0.
+        threshold_type: nil,
+        # Whether to enable web search for this workflow's evaluations.
+        web_search: nil,
         request_options: {}
       )
       end

data/rbi/deeprails/resources/monitor.rbi

@@ -10,6 +10,7 @@ module Deeprails
           guardrail_metrics:
             T::Array[Deeprails::MonitorCreateParams::GuardrailMetric::OrSymbol],
           name: String,
+          context_awareness: T::Boolean,
           description: String,
           file_search: T::Array[String],
           web_search: T::Boolean,
@@ -24,6 +25,13 @@ module Deeprails
         guardrail_metrics:,
         # Name of the new monitor.
         name:,
+        # Context includes any structured information that directly relates to the model’s
+        # input and expected output—e.g., the recent turn-by-turn history between an AI
+        # tutor and a student, facts or state passed through an agentic workflow, or other
+        # domain-specific signals your system already knows and wants the model to
+        # condition on. This field determines whether to enable context awareness for this
+        # monitor's evaluations. Defaults to false.
+        context_awareness: nil,
         # Description of the new monitor.
         description: nil,
         # An array of file IDs to search in the monitor's evaluations. Files must be
@@ -54,27 +62,39 @@ module Deeprails
       )
       end
 
-      # Use this endpoint to update the name, description, or status of an existing
-      # monitor
+      # Use this endpoint to update the name, status, and/or other details of an
+      # existing monitor.
       sig do
         params(
           monitor_id: String,
           description: String,
+          file_search: T::Array[String],
+          guardrail_metrics:
+            T::Array[Deeprails::MonitorUpdateParams::GuardrailMetric::OrSymbol],
           name: String,
           status: Deeprails::MonitorUpdateParams::Status::OrSymbol,
+          web_search: T::Boolean,
           request_options: Deeprails::RequestOptions::OrHash
         ).returns(Deeprails::MonitorUpdateResponse)
       end
       def update(
         # The ID of the monitor to edit.
         monitor_id,
-        # Description of the monitor.
+        # New description of the monitor.
         description: nil,
-        # Name of the monitor.
+        # An array of file IDs to search in the monitor's evaluations. Files must be
+        # uploaded via the DeepRails API first.
+        file_search: nil,
+        # An array of the new guardrail metrics that model input and output pairs will be
+        # evaluated on.
+        guardrail_metrics: nil,
+        # New name of the monitor.
         name: nil,
         # Status of the monitor. Can be `active` or `inactive`. Inactive monitors no
         # longer record and evaluate events.
         status: nil,
+        # Whether to enable web search for this monitor's evaluations.
+        web_search: nil,
         request_options: {}
       )
       end
@@ -112,8 +132,8 @@ module Deeprails
         # The ID of the monitor associated with this event.
         monitor_id,
         # A dictionary of inputs sent to the LLM to generate output. The dictionary must
-        # contain at least a `user_prompt` field or a `system_prompt` field. For
-        # ground_truth_adherence guardrail metric, `ground_truth` should be provided.
+        # contain a `user_prompt` field. For ground_truth_adherence guardrail metric,
+        # `ground_truth` should be provided.
         model_input:,
         # Output generated by the LLM to be evaluated.
         model_output:,
@@ -121,8 +141,8 @@ module Deeprails
         nametag: nil,
         # Run mode for the monitor event. The run mode allows the user to optimize for
         # speed, accuracy, and cost by determining which models are used to evaluate the
-        # event. Available run modes include `precision_plus`, `precision`, `smart`, and
-        # `economy`. Defaults to `smart`.
+        # event. Available run modes include `precision_plus_codex`, `precision_plus`,
+        # `precision`, `smart`, and `economy`. Defaults to `smart`.
         run_mode: nil,
         request_options: {}
       )

data/sig/deeprails/internal/stream.rbs

@@ -0,0 +1,9 @@
+module Deeprails
+  module Internal
+    class Stream[Elem]
+      include Deeprails::Internal::Type::BaseStream[Deeprails::Internal::Util::server_sent_event, Elem]
+
+      private def iterator: -> Enumerable[Elem]
+    end
+  end
+end