deeprails 0.18.0 → 0.23.0

data/rbi/deeprails/models/monitor_detail_response.rbi

@@ -8,7 +8,8 @@ module Deeprails
           T.any(Deeprails::MonitorDetailResponse, Deeprails::Internal::AnyHash)
         end
 
-      # An array of capabilities associated with this monitor.
+      # An array of extended AI capabilities associated with this monitor. Can be
+      # `web_search`, `file_search`, and/or `context_awareness`.
       sig { returns(T::Array[Deeprails::MonitorDetailResponse::Capability]) }
       attr_accessor :capabilities
 
@@ -77,7 +78,8 @@ module Deeprails
         ).returns(T.attached_class)
       end
       def self.new(
-        # An array of capabilities associated with this monitor.
+        # An array of extended AI capabilities associated with this monitor. Can be
+        # `web_search`, `file_search`, and/or `context_awareness`.
         capabilities:,
         # The time the monitor was created in UTC.
         created_at:,
@@ -170,8 +172,9 @@ module Deeprails
         attr_accessor :evaluation_status
 
         # 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. When `context_awareness` is enabled,
+        # `context` should be provided.
         sig do
           returns(Deeprails::MonitorDetailResponse::Evaluation::ModelInput)
         end
@@ -290,8 +293,9 @@ module Deeprails
           # Status of the evaluation.
           evaluation_status:,
           # 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. When `context_awareness` is enabled,
+          # `context` should be provided.
           model_input:,
           # Output generated by the LLM to be evaluated.
           model_output:,
@@ -403,6 +407,36 @@ 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::MonitorDetailResponse::Evaluation::ModelInput::Context
+                ]
+              )
+            )
+          end
+          attr_reader :context
+
+          sig do
+            params(
+              context:
+                T::Array[
+                  Deeprails::MonitorDetailResponse::Evaluation::ModelInput::Context::OrHash
+                ]
+            ).void
+          end
+          attr_writer :context
+
           # The ground truth for evaluating Ground Truth Adherence guardrail.
           sig { returns(T.nilable(String)) }
           attr_reader :ground_truth
@@ -417,44 +451,91 @@ 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
-          # 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. When `context_awareness` is enabled,
+          # `context` should be provided.
           sig do
             params(
+              user_prompt: String,
+              context:
+                T::Array[
+                  Deeprails::MonitorDetailResponse::Evaluation::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 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(
               {
+                user_prompt: String,
+                context:
+                  T::Array[
+                    Deeprails::MonitorDetailResponse::Evaluation::ModelInput::Context
+                  ],
                 ground_truth: String,
-                system_prompt: String,
-                user_prompt: String
+                system_prompt: String
               }
             )
           end
           def to_hash
           end
+
+          class Context < Deeprails::Internal::Type::BaseModel
+            OrHash =
+              T.type_alias do
+                T.any(
+                  Deeprails::MonitorDetailResponse::Evaluation::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 evaluation. The run mode allows the user to optimize for speed,

data/rbi/deeprails/models/monitor_event_detail_response.rbi

@@ -11,7 +11,8 @@ module Deeprails
           )
         end
 
-      # The capabilities associated with the monitor event.
+      # The extended AI capabilities associated with the monitor event. Can be
+      # `web_search`, `file_search`, and/or `context_awareness`.
       sig do
         returns(
           T.nilable(T::Array[Deeprails::MonitorEventDetailResponse::Capability])
@@ -156,7 +157,8 @@ module Deeprails
         ).returns(T.attached_class)
       end
       def self.new(
-        # The capabilities associated with the monitor event.
+        # The extended AI capabilities associated with the monitor event. Can be
+        # `web_search`, `file_search`, and/or `context_awareness`.
         capabilities: nil,
         # The time spent on the evaluation in seconds.
         eval_time: nil,

data/rbi/deeprails/models/monitor_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
-      # 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.
       sig { returns(Deeprails::MonitorSubmitEventParams::ModelInput) }
       attr_reader :model_input
 
@@ -40,8 +40,8 @@ module Deeprails
 
       # 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`.
       sig do
         returns(
           T.nilable(Deeprails::MonitorSubmitEventParams::RunMode::OrSymbol)
@@ -67,8 +67,8 @@ 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
-        # 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:,
@@ -76,8 +76,8 @@ 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: {}
       )
@@ -106,6 +106,34 @@ 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::MonitorSubmitEventParams::ModelInput::Context]
+            )
+          )
+        end
+        attr_reader :context
+
+        sig do
+          params(
+            context:
+              T::Array[
+                Deeprails::MonitorSubmitEventParams::ModelInput::Context::OrHash
+              ]
+          ).void
+        end
+        attr_writer :context
+
         # The ground truth for evaluating Ground Truth Adherence guardrail.
         sig { returns(T.nilable(String)) }
         attr_reader :ground_truth
@@ -120,46 +148,96 @@ 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
-        # 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.
         sig do
           params(
+            user_prompt: String,
+            context:
+              T::Array[
+                Deeprails::MonitorSubmitEventParams::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 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::MonitorSubmitEventParams::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::MonitorSubmitEventParams::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 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`.
       module RunMode
         extend Deeprails::Internal::Type::Enum
 
@@ -169,6 +247,11 @@ module Deeprails
           end
         OrSymbol = T.type_alias { T.any(Symbol, String) }
 
+        PRECISION_PLUS_CODEX =
+          T.let(
+            :precision_plus_codex,
+            Deeprails::MonitorSubmitEventParams::RunMode::TaggedSymbol
+          )
         PRECISION_PLUS =
           T.let(
             :precision_plus,

data/rbi/deeprails/models/monitor_update_params.rbi

@@ -11,14 +11,41 @@ module Deeprails
           T.any(Deeprails::MonitorUpdateParams, Deeprails::Internal::AnyHash)
         end
 
-      # Description of the monitor.
+      # New description of the monitor.
       sig { returns(T.nilable(String)) }
       attr_reader :description
 
       sig { params(description: String).void }
       attr_writer :description
 
-      # 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.
+      sig { returns(T.nilable(T::Array[String])) }
+      attr_reader :file_search
+
+      sig { params(file_search: T::Array[String]).void }
+      attr_writer :file_search
+
+      # An array of the new guardrail metrics that model input and output pairs will be
+      # evaluated on.
+      sig do
+        returns(
+          T.nilable(
+            T::Array[Deeprails::MonitorUpdateParams::GuardrailMetric::OrSymbol]
+          )
+        )
+      end
+      attr_reader :guardrail_metrics
+
+      sig do
+        params(
+          guardrail_metrics:
+            T::Array[Deeprails::MonitorUpdateParams::GuardrailMetric::OrSymbol]
+        ).void
+      end
+      attr_writer :guardrail_metrics
+
+      # New name of the monitor.
       sig { returns(T.nilable(String)) }
       attr_reader :name
 
@@ -37,22 +64,41 @@ module Deeprails
       end
       attr_writer :status
 
+      # Whether to enable web search for this monitor'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(
           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(T.attached_class)
       end
       def self.new(
-        # 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
@@ -61,8 +107,14 @@ module Deeprails
         override.returns(
           {
             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
           }
         )
@@ -70,6 +122,57 @@ module Deeprails
       def to_hash
       end
 
+      module GuardrailMetric
+        extend Deeprails::Internal::Type::Enum
+
+        TaggedSymbol =
+          T.type_alias do
+            T.all(Symbol, Deeprails::MonitorUpdateParams::GuardrailMetric)
+          end
+        OrSymbol = T.type_alias { T.any(Symbol, String) }
+
+        CORRECTNESS =
+          T.let(
+            :correctness,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+        COMPLETENESS =
+          T.let(
+            :completeness,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+        INSTRUCTION_ADHERENCE =
+          T.let(
+            :instruction_adherence,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+        CONTEXT_ADHERENCE =
+          T.let(
+            :context_adherence,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+        GROUND_TRUTH_ADHERENCE =
+          T.let(
+            :ground_truth_adherence,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+        COMPREHENSIVE_SAFETY =
+          T.let(
+            :comprehensive_safety,
+            Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+          )
+
+        sig do
+          override.returns(
+            T::Array[
+              Deeprails::MonitorUpdateParams::GuardrailMetric::TaggedSymbol
+            ]
+          )
+        end
+        def self.values
+        end
+      end
+
       # Status of the monitor. Can be `active` or `inactive`. Inactive monitors no
       # longer record and evaluate events.
       module Status