actionagent 1.6.0 → 1.6.3

data/app/services/action_agent/agent_tool_roster.rb

@@ -8,11 +8,19 @@ module ActionAgent
   # Three groups, one row vocabulary — the same one the Tools and MCP
   # Services pages use, so a roster reads the way the observability views do:
   #
-  # * **Agent-defined** — tools discovered from the rosters this agent's own
-  #   generations offered (ToolDiscovery's +agent+ origin). Schema-derived:
-  #   the agent class declares them, so they are reported here rather than
-  #   selected. Read-only for the same reason MCP rows are — a checkbox that
-  #   cannot add or remove the tool is a control that changes nothing.
+  # * **Agent-defined** — the tools this agent's own generations offered
+  #   (ToolDiscovery's +agent+ origin), plus every schema tool the host
+  #   declares, on or off. Two kinds share the group. A *schema tool* — one
+  #   a host ActiveAgent::SchemaTools class generates — is offered only while
+  #   +agent.tools+ names it: that is the reading AgentToolbox takes at
+  #   generation time, and the evaluation runner, dashboard runs and the MCP
+  #   facade all follow it. So it is selected here, and switchable, and every
+  #   one has a row because any agent may enable any of them
+  #   (Agent.available_tools) and a tool switched off has to keep its row to
+  #   be switched back on. A tool the agent class declares *in code* is
+  #   offered by the class itself; the dashboard reports it and cannot switch
+  #   it, so its row is read-only — a checkbox that cannot add or remove the
+  #   tool is a control that changes nothing.
   # * **Dashboard** — Agent::AVAILABLE_TOOLS, the capabilities the builder
   #   offers every agent. These are the roster: +agent.tools+ is what
   #   AgentToolbox turns into function schemas at generation time.
@@ -20,8 +28,9 @@ module ActionAgent
   #   agent enables, which is where they are edited.
   #
   # Enablement reads the agent's own configuration: +tools+ for the dashboard
-  # capabilities, +mcp_servers+ for services and their per-server allow-lists
-  # (an entry with no +tools+ key offers everything the server serves).
+  # capabilities and the schema tools, +mcp_servers+ for services and their
+  # per-server allow-lists (an entry with no +tools+ key offers everything
+  # the server serves).
   class AgentToolRoster
     AGENT_DEFINED = "agent_defined"
     DASHBOARD = "dashboard"
@@ -158,23 +167,40 @@ module ActionAgent
       agent_defined_rows + dashboard_rows
     end
 
-    # Schema-derived tools: whatever this agent's generations offered that
-    # is neither an MCP tool nor one of the dashboard's own.
+    # Agent-defined tools: whatever this agent's generations offered that is
+    # neither an MCP tool nor one of the dashboard's own, plus every schema
+    # tool the host declares, whether or not the window saw it.
     def agent_defined_rows
-      detected
+      observed = detected
         .select { |tool| tool[:origin] == ToolDiscovery::ORIGIN_AGENT }
         .reject { |tool| dashboard_function_names.include?(tool[:name]) }
-        .map do |tool|
-          usage_row(tool).merge(
-            key: tool[:name],
-            name: tool[:name],
-            source: AGENT_DEFINED,
-            description: tool[:description],
-            # The agent class declares these; the dashboard reports them.
-            enabled: true,
-            editable: false
-          )
-        end
+        .index_by { |tool| tool[:name] }
+
+      (observed.keys + ActionAgent.schema_tool_names).uniq.map do |name|
+        tool = observed[name]
+        schema = schema_tool?(name)
+
+        usage_row(tool).merge(
+          key: name,
+          name: name,
+          source: AGENT_DEFINED,
+          description: tool&.dig(:description) || schema_tool_description(name),
+          # A schema tool is offered only while the roster names it — the
+          # window may still show it being called, but that is history. A
+          # tool the agent class declares in code is offered by the class.
+          enabled: schema ? saved_tools.include?(name) : true,
+          editable: schema
+        )
+      end
+    end
+
+    def schema_tool?(name)
+      @schema_tool ||= Hash.new { |cache, key| cache[key] = ActionAgent.schema_tool_class_for(key).present? }
+      @schema_tool[name]
+    end
+
+    def schema_tool_description(name)
+      AgentToolbox.schema_tool_definitions(name).first&.dig(:description)
     end
 
     def dashboard_rows

data/app/services/action_agent/evaluation_runner_service.rb

@@ -149,9 +149,7 @@ module ActionAgent
     end
 
     def sample_generations(model: nil)
-      scope = AgentGeneration
-        .joins(:agent_context)
-        .where(AgentContext.table_name => { contextable: @evaluation.agent })
+      scope = @evaluation.agent.generations
       scope = scope.where(model: model) if model
       scope.order(created_at: :desc).limit(@evaluation.sample_size).to_a
     end
@@ -468,15 +466,17 @@ module ActionAgent
     end
 
     def judge_class
-      provider = judge_provider
-      model = @evaluation.judge_model.presence
-      options = {}
-      options[:model] = model if model
-      options.merge!(owner_provider_options(provider))
-
-      @judge_class ||= Class.new(ActiveAgent::Base) do
-        define_singleton_method(:name) { "EvaluationJudgeAgent" }
-        generate_with provider, **options
+      @judge_class ||= begin
+        provider = judge_provider
+        model = @evaluation.judge_model.presence
+        options = {}
+        options[:model] = model if model
+        options.merge!(owner_provider_options(provider))
+
+        Class.new(ActiveAgent::Base) do
+          define_singleton_method(:name) { "EvaluationJudgeAgent" }
+          generate_with provider, **options
+        end
       end
     end
 

data/app/services/action_agent/scenario_evaluation_runner.rb

@@ -64,7 +64,16 @@ module ActionAgent
       report = if adapter
         raise ArgumentError, "scenario evaluation adapter must be callable" unless adapter.respond_to?(:call)
 
-        adapter.call(evaluation: @evaluation, owner: owner, scenarios: tasks, models: specs, on_result: on_result)
+        # The default path meters each replay itself (see #replay). An adapter
+        # runs the agent in the host's own runtime, out of reach of that call,
+        # so the execution is counted here as each result lands — one per
+        # scenario x model, the same unit. Otherwise a host that adapts the
+        # replay is silently unmetered unless it remembers to meter itself.
+        metered = lambda do |result|
+          ActionAgent.record_usage(owner, :execution)
+          on_result.call(result)
+        end
+        adapter.call(evaluation: @evaluation, owner: owner, scenarios: tasks, models: specs, on_result: metered)
       else
         ensure_judge_defined_kpis! if @evaluation.judge_defined?
         default_report(tasks, specs, on_result)
@@ -122,6 +131,12 @@ module ActionAgent
     def model_specs
       names = Array(@selection[:models]).presence || @evaluation.compare_models
       specs = Evals::ModelSpec.parse_all(names, default_provider: @evaluation.agent.provider, providers: Agent::PROVIDERS + %w[mock])
+      # parse_all resolves a bare name against `providers:` but passes through a
+      # `provider/model` whose provider is not in that list, so the run would
+      # otherwise reach the replay with a provider nothing can serve.
+      unsupported = specs.map(&:provider).uniq - (Agent::PROVIDERS + %w[mock])
+      raise ArgumentError, "unsupported model provider: #{unsupported.to_sentence}" if unsupported.any?
+
       return specs if specs.any?
 
       [ Evals::ModelSpec.new(label: @evaluation.agent.model, provider: @evaluation.agent.provider, model: @evaluation.agent.model) ]

data/config/routes.rb

@@ -161,6 +161,9 @@ ActionAgent::Engine.routes.draw do
   # dashboard API key rather than a session, so it sits outside the api
   # namespace's session-authenticated controllers.
   post "mcp", to: "api/mcp#create"
+  match "mcp", to: ->(_env) { [ 405, { "Allow" => "POST" }, [] ] },
+    via: [ :get, :delete ],
+    constraints: ->(request) { request.delete? || !request.format.html? }
 
   # MCP Streamable HTTP (2025-03-26): a client MAY open the server-to-client
   # SSE stream with GET, and ends a session with DELETE. This facade offers

data/lib/action_agent/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module ActionAgent
-  VERSION = "1.6.0"
+  VERSION = "1.6.3"
 end

data/lib/generators/action_agent/install_generator.rb

@@ -57,6 +57,12 @@ module ActionAgent
         )
       end
 
+      # Agent releases arrived after both tables shipped; the migration guards
+      # every column, so it is emitted for any install that lacks it.
+      unless existing_migration?("add_agent_releases")
+        migration_template("add_agent_releases.rb.erb", "db/migrate/add_agent_releases.rb")
+      end
+
       # The rest of the dashboard — agents, runs, versions, conversations,
       # evaluations, sandboxes, recordings, keys. Skippable for an app that
       # only wants to be a trace sink.

data/lib/generators/action_agent/templates/action_agent.rb.erb

@@ -31,6 +31,14 @@ ActionAgent.configure do |config|
   #   config.authentication_method = nil
   #
   config.authentication_method = nil
+  #
+  # Where a browser is sent when it asks for a dashboard page without a valid
+  # session; unset, it gets a minimal session-expired page. API clients get a
+  # bare 401 either way. The sign-out path is what the header's "Sign out"
+  # item POSTs to (with _method=delete); unset, the item is not shown.
+  #
+  # config.sign_in_path = "/sign_in"
+  # config.sign_out_path = "/sign_out"
 
   # ==========================================================================
   # Ingest API authentication

data/lib/generators/action_agent/templates/add_agent_releases.rb.erb

@@ -0,0 +1,50 @@
+# frozen_string_literal: true
+
+# Agent releases: a version cut from the agent's code carries the digest of
+# what the model was given and the deploy it shipped in, and every trace,
+# run and evaluation run records the version it ran under. Emitted for an
+# install whose dashboard tables predate releases; the create-table
+# migration carries the same columns for a fresh install.
+#
+# Each step is guarded so the migration is safe to re-run against a database
+# that already has some of these columns.
+class AddAgentReleases < ActiveRecord::Migration<%= migration_version %>
+  def up
+    add_column_unless_exists :active_agent_agents, :release_digest, :string
+
+    add_column_unless_exists :active_agent_agent_versions, :release_digest, :string
+    add_column_unless_exists :active_agent_agent_versions, :revision, :string
+    add_index_unless_exists :active_agent_agent_versions, [ :agent_id, :release_digest ]
+
+    %i[active_agent_telemetry_traces active_agent_agent_runs active_agent_evaluation_runs].each do |table|
+      add_column_unless_exists table, :agent_version_id, :bigint
+      add_index_unless_exists table, :agent_version_id
+    end
+  end
+
+  def down
+    remove_column :active_agent_agents, :release_digest if column_exists?(:active_agent_agents, :release_digest)
+    remove_column :active_agent_agent_versions, :release_digest if column_exists?(:active_agent_agent_versions, :release_digest)
+    remove_column :active_agent_agent_versions, :revision if column_exists?(:active_agent_agent_versions, :revision)
+
+    %i[active_agent_telemetry_traces active_agent_agent_runs active_agent_evaluation_runs].each do |table|
+      remove_column table, :agent_version_id if column_exists?(table, :agent_version_id)
+    end
+  end
+
+  private
+
+  def add_column_unless_exists(table, column, type)
+    return unless table_exists?(table)
+    return if column_exists?(table, column)
+
+    add_column table, column, type
+  end
+
+  def add_index_unless_exists(table, columns)
+    return unless table_exists?(table)
+    return if index_exists?(table, columns)
+
+    add_index table, columns
+  end
+end

data/lib/tasks/action_agent.rake

@@ -7,3 +7,36 @@ namespace :action_agent do
     puts "Agent template library: #{ActionAgent::AgentTemplate.count} templates"
   end
 end
+
+namespace :action_agent do
+  namespace :agents do
+    desc "Cut a dashboard version for every agent whose code changed (run on deploy; REVISION=<git sha> to label it)"
+    task :release, [ :revision ] => :environment do |_task, args|
+      Rails.application.eager_load!
+      result = ActionAgent::AgentRelease.call(revision: args[:revision] || ENV["REVISION"], released_by: ENV["RELEASED_BY"])
+
+      puts "Release #{result.revision.presence || '(no revision)'}"
+      result.rows.each do |row|
+        if row.skipped
+          puts format("  %-28s skipped — %s", row.agent.name, row.skipped)
+        elsif row.cut
+          puts format("  %-28s v%-3d cut   %s", row.agent.name, row.version.version_number, row.version.change_summary)
+        else
+          puts format("  %-28s v%-3d unchanged (%s)", row.agent.name, row.version.version_number, row.version.release_digest)
+        end
+      end
+      puts "#{result.cut.size} cut, #{result.rows.size - result.cut.size - result.skipped.size} unchanged, #{result.skipped.size} skipped"
+    end
+
+    desc "List each agent's current version and release"
+    task versions: :environment do
+      ActionAgent::Agent.order(:name).find_each do |agent|
+        version = agent.latest_version
+        release = agent.latest_release
+        puts format("%-28s v%-3s %s%s", agent.name, version&.version_number || "-",
+                    release ? "release #{release.release_digest}" : "no release",
+                    release&.revision.present? ? " · #{release.revision}" : "")
+      end
+    end
+  end
+end

metadata

@@ -1,13 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: actionagent
 version: !ruby/object:Gem::Version
-  version: 1.6.0
+  version: 1.6.3
 platform: ruby
 authors:
 - Justin Bowen
+autorequire:
 bindir: bin
 cert_chain: []
-date: 1980-01-02 00:00:00.000000000 Z
+date: 2026-09-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: activeagent
@@ -192,6 +193,7 @@ files:
 - app/serializers/action_agent/trace_interaction_serializer.rb
 - app/services/action_agent/agent_execution_service.rb
 - app/services/action_agent/agent_registrar.rb
+- app/services/action_agent/agent_release.rb
 - app/services/action_agent/agent_scorecard.rb
 - app/services/action_agent/agent_tool_roster.rb
 - app/services/action_agent/agent_toolbox.rb
@@ -226,6 +228,7 @@ files:
 - lib/generators/action_agent/install_generator.rb
 - lib/generators/action_agent/templates/action_agent.rb.erb
 - lib/generators/action_agent/templates/add_agent_id_to_active_agent_telemetry_traces.rb.erb
+- lib/generators/action_agent/templates/add_agent_releases.rb.erb
 - lib/generators/action_agent/templates/create_active_agent_dashboard_tables.rb.erb
 - lib/generators/action_agent/templates/create_active_agent_evaluation_scenarios.rb.erb
 - lib/generators/action_agent/templates/create_active_agent_telemetry_traces.rb.erb
@@ -238,6 +241,7 @@ metadata:
   documentation_uri: https://docs.activeagents.ai/framework/self-hosted-observability
   source_code_uri: https://github.com/activeagents/activeagent
   rubygems_mfa_required: 'true'
+post_install_message:
 rdoc_options: []
 require_paths:
 - lib
@@ -252,7 +256,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.6.9
+rubygems_version: 3.5.22
+signing_key:
 specification_version: 4
 summary: The Active Agent dashboard, as a mountable Rails engine
 test_files: []