deeprails 0.18.0 → 0.23.0

data/rbi/deeprails/models/defend_submit_event_params.rbi

@@ -15,8 +15,8 @@ module Deeprails
         end
 
       # 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.
       sig { returns(Deeprails::DefendSubmitEventParams::ModelInput) }
       attr_reader :model_input
 
@@ -37,8 +37,8 @@ module Deeprails
 
       # 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`.
       sig { returns(Deeprails::DefendSubmitEventParams::RunMode::OrSymbol) }
       attr_accessor :run_mode
 
@@ -61,8 +61,8 @@ module Deeprails
       end
       def self.new(
         # 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:,
@@ -70,8 +70,8 @@ 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,
@@ -103,6 +103,34 @@ module Deeprails
             )
           end
 
+        # The user prompt used to generate the output.
+        sig { returns(String) }
+        attr_accessor :user_prompt
+
+        # 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.
+        sig do
+          returns(
+            T.nilable(
+              T::Array[Deeprails::DefendSubmitEventParams::ModelInput::Context]
+            )
+          )
+        end
+        attr_reader :context
+
+        sig do
+          params(
+            context:
+              T::Array[
+                Deeprails::DefendSubmitEventParams::ModelInput::Context::OrHash
+              ]
+          ).void
+        end
+        attr_writer :context
+
         # The ground truth for evaluating the Ground Truth Adherence guardrail.
         sig { returns(T.nilable(String)) }
         attr_reader :ground_truth
@@ -117,46 +145,96 @@ module Deeprails
         sig { params(system_prompt: String).void }
         attr_writer :system_prompt
 
-        # The user prompt used to generate the output.
-        sig { returns(T.nilable(String)) }
-        attr_reader :user_prompt
-
-        sig { params(user_prompt: String).void }
-        attr_writer :user_prompt
-
         # 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.
         sig do
           params(
+            user_prompt: String,
+            context:
+              T::Array[
+                Deeprails::DefendSubmitEventParams::ModelInput::Context::OrHash
+              ],
             ground_truth: String,
-            system_prompt: String,
-            user_prompt: String
+            system_prompt: String
           ).returns(T.attached_class)
         end
         def self.new(
+          # The user prompt used to generate the output.
+          user_prompt:,
+          # 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.
+          context: nil,
           # The ground truth for evaluating the Ground Truth Adherence guardrail.
           ground_truth: nil,
           # The system prompt used to generate the output.
-          system_prompt: nil,
-          # The user prompt used to generate the output.
-          user_prompt: nil
+          system_prompt: nil
         )
         end
 
         sig do
           override.returns(
-            { ground_truth: String, system_prompt: String, user_prompt: String }
+            {
+              user_prompt: String,
+              context:
+                T::Array[
+                  Deeprails::DefendSubmitEventParams::ModelInput::Context
+                ],
+              ground_truth: String,
+              system_prompt: String
+            }
           )
         end
         def to_hash
         end
+
+        class Context < Deeprails::Internal::Type::BaseModel
+          OrHash =
+            T.type_alias do
+              T.any(
+                Deeprails::DefendSubmitEventParams::ModelInput::Context,
+                Deeprails::Internal::AnyHash
+              )
+            end
+
+          # The content of the message.
+          sig { returns(T.nilable(String)) }
+          attr_reader :content
+
+          sig { params(content: String).void }
+          attr_writer :content
+
+          # The role of the speaker.
+          sig { returns(T.nilable(String)) }
+          attr_reader :role
+
+          sig { params(role: String).void }
+          attr_writer :role
+
+          sig do
+            params(content: String, role: String).returns(T.attached_class)
+          end
+          def self.new(
+            # The content of the message.
+            content: nil,
+            # The role of the speaker.
+            role: nil
+          )
+          end
+
+          sig { override.returns({ content: String, role: String }) }
+          def to_hash
+          end
+        end
       end
 
       # 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`.
       module RunMode
         extend Deeprails::Internal::Type::Enum
 
@@ -166,6 +244,11 @@ module Deeprails
           end
         OrSymbol = T.type_alias { T.any(Symbol, String) }
 
+        PRECISION_PLUS_CODEX =
+          T.let(
+            :precision_plus_codex,
+            Deeprails::DefendSubmitEventParams::RunMode::TaggedSymbol
+          )
         PRECISION_PLUS =
           T.let(
             :precision_plus,

data/rbi/deeprails/models/defend_update_response.rbi

@@ -21,11 +21,19 @@ module Deeprails
       sig { returns(String) }
       attr_accessor :workflow_id
 
+      # The name of the workflow.
+      sig { returns(T.nilable(String)) }
+      attr_reader :name
+
+      sig { params(name: String).void }
+      attr_writer :name
+
       sig do
         params(
           modified_at: Time,
           status: Deeprails::DefendUpdateResponse::Status::OrSymbol,
-          workflow_id: String
+          workflow_id: String,
+          name: String
         ).returns(T.attached_class)
       end
       def self.new(
@@ -35,7 +43,9 @@ module Deeprails
         # workflows will not accept events.
         status:,
         # A unique workflow ID.
-        workflow_id:
+        workflow_id:,
+        # The name of the workflow.
+        name: nil
       )
       end
 
@@ -44,7 +54,8 @@ module Deeprails
           {
             modified_at: Time,
             status: Deeprails::DefendUpdateResponse::Status::TaggedSymbol,
-            workflow_id: String
+            workflow_id: String,
+            name: String
           }
         )
       end

data/rbi/deeprails/models/defend_update_workflow_params.rbi

@@ -14,32 +14,193 @@ module Deeprails
           )
         end
 
-      # 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`.
+      sig do
+        returns(
+          T.nilable(
+            T::Hash[
+              Symbol,
+              Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::OrSymbol
+            ]
+          )
+        )
+      end
+      attr_reader :automatic_hallucination_tolerance_levels
+
+      sig do
+        params(
+          automatic_hallucination_tolerance_levels:
+            T::Hash[
+              Symbol,
+              Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::OrSymbol
+            ]
+        ).void
+      end
+      attr_writer :automatic_hallucination_tolerance_levels
+
+      # Whether to enable context awareness for this workflow's evaluations.
+      sig { returns(T.nilable(T::Boolean)) }
+      attr_reader :context_awareness
+
+      sig { params(context_awareness: T::Boolean).void }
+      attr_writer :context_awareness
+
+      # 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`.
+      sig { returns(T.nilable(T::Hash[Symbol, Float])) }
+      attr_reader :custom_hallucination_threshold_values
+
+      sig do
+        params(
+          custom_hallucination_threshold_values: T::Hash[Symbol, Float]
+        ).void
+      end
+      attr_writer :custom_hallucination_threshold_values
+
+      # New description for the workflow.
       sig { returns(T.nilable(String)) }
       attr_reader :description
 
       sig { params(description: String).void }
       attr_writer :description
 
-      # 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.
+      sig { returns(T.nilable(T::Array[String])) }
+      attr_reader :file_search
+
+      sig { params(file_search: T::Array[String]).void }
+      attr_writer :file_search
+
+      # 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.
+      sig do
+        returns(
+          T.nilable(
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::OrSymbol
+          )
+        )
+      end
+      attr_reader :improvement_action
+
+      sig do
+        params(
+          improvement_action:
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::OrSymbol
+        ).void
+      end
+      attr_writer :improvement_action
+
+      # Max. number of improvement action attempts until a given event passes the
+      # guardrails. Defaults to 10.
+      sig { returns(T.nilable(Integer)) }
+      attr_reader :max_improvement_attempts
+
+      sig { params(max_improvement_attempts: Integer).void }
+      attr_writer :max_improvement_attempts
+
+      # New name for the workflow.
       sig { returns(T.nilable(String)) }
       attr_reader :name
 
       sig { params(name: String).void }
       attr_writer :name
 
+      # 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.
+      sig do
+        returns(
+          T.nilable(
+            Deeprails::DefendUpdateWorkflowParams::ThresholdType::OrSymbol
+          )
+        )
+      end
+      attr_reader :threshold_type
+
+      sig do
+        params(
+          threshold_type:
+            Deeprails::DefendUpdateWorkflowParams::ThresholdType::OrSymbol
+        ).void
+      end
+      attr_writer :threshold_type
+
+      # Whether to enable web search for this workflow's evaluations.
+      sig { returns(T.nilable(T::Boolean)) }
+      attr_reader :web_search
+
+      sig { params(web_search: T::Boolean).void }
+      attr_writer :web_search
+
       sig do
         params(
+          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(T.attached_class)
       end
       def self.new(
-        # 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
@@ -47,14 +208,147 @@ module Deeprails
       sig do
         override.returns(
           {
+            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
           }
         )
       end
       def to_hash
       end
+
+      module AutomaticHallucinationToleranceLevel
+        extend Deeprails::Internal::Type::Enum
+
+        TaggedSymbol =
+          T.type_alias do
+            T.all(
+              Symbol,
+              Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel
+            )
+          end
+        OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+        LOW =
+          T.let(
+            :low,
+            Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::TaggedSymbol
+          )
+        MEDIUM =
+          T.let(
+            :medium,
+            Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::TaggedSymbol
+          )
+        HIGH =
+          T.let(
+            :high,
+            Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::TaggedSymbol
+          )
+
+        sig do
+          override.returns(
+            T::Array[
+              Deeprails::DefendUpdateWorkflowParams::AutomaticHallucinationToleranceLevel::TaggedSymbol
+            ]
+          )
+        end
+        def self.values
+        end
+      end
+
+      # 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.
+      module ImprovementAction
+        extend Deeprails::Internal::Type::Enum
+
+        TaggedSymbol =
+          T.type_alias do
+            T.all(
+              Symbol,
+              Deeprails::DefendUpdateWorkflowParams::ImprovementAction
+            )
+          end
+        OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+        REGEN =
+          T.let(
+            :regen,
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::TaggedSymbol
+          )
+        FIXIT =
+          T.let(
+            :fixit,
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::TaggedSymbol
+          )
+        DO_NOTHING =
+          T.let(
+            :do_nothing,
+            Deeprails::DefendUpdateWorkflowParams::ImprovementAction::TaggedSymbol
+          )
+
+        sig do
+          override.returns(
+            T::Array[
+              Deeprails::DefendUpdateWorkflowParams::ImprovementAction::TaggedSymbol
+            ]
+          )
+        end
+        def self.values
+        end
+      end
+
+      # 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.
+      module ThresholdType
+        extend Deeprails::Internal::Type::Enum
+
+        TaggedSymbol =
+          T.type_alias do
+            T.all(Symbol, Deeprails::DefendUpdateWorkflowParams::ThresholdType)
+          end
+        OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+        AUTOMATIC =
+          T.let(
+            :automatic,
+            Deeprails::DefendUpdateWorkflowParams::ThresholdType::TaggedSymbol
+          )
+        CUSTOM =
+          T.let(
+            :custom,
+            Deeprails::DefendUpdateWorkflowParams::ThresholdType::TaggedSymbol
+          )
+
+        sig do
+          override.returns(
+            T::Array[
+              Deeprails::DefendUpdateWorkflowParams::ThresholdType::TaggedSymbol
+            ]
+          )
+        end
+        def self.values
+        end
+      end
     end
   end
 end

data/rbi/deeprails/models/monitor_create_params.rbi

@@ -26,6 +26,18 @@ module Deeprails
       sig { returns(String) }
       attr_accessor :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.
+      sig { returns(T.nilable(T::Boolean)) }
+      attr_reader :context_awareness
+
+      sig { params(context_awareness: T::Boolean).void }
+      attr_writer :context_awareness
+
       # Description of the new monitor.
       sig { returns(T.nilable(String)) }
       attr_reader :description
@@ -53,6 +65,7 @@ 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,
@@ -67,6 +80,13 @@ 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
@@ -86,6 +106,7 @@ module Deeprails
                 Deeprails::MonitorCreateParams::GuardrailMetric::OrSymbol
               ],
             name: String,
+            context_awareness: T::Boolean,
             description: String,
             file_search: T::Array[String],
             web_search: T::Boolean,