actionagent 1.9.2 → 1.9.4

data/app/views/action_agent/traces/show.html.erb

@@ -1,36 +1,7 @@
-<div class="mb-4">
-  <%= link_to "← Back to traces", traces_path, class: "text-indigo-600 hover:text-indigo-800 text-sm" %>
-</div>
-
-<div class="bg-white dark:bg-gray-800 shadow rounded-lg overflow-hidden">
-  <div class="px-4 py-4 border-b border-gray-200 dark:border-gray-700">
-    <div class="flex items-center justify-between">
-      <div>
-        <h1 class="text-xl font-semibold text-gray-900 dark:text-white">
-          <%= @trace.display_name %>
-        </h1>
-        <p class="text-sm text-gray-500 dark:text-gray-400 font-mono">
-          <%= @trace.trace_id %>
-        </p>
-      </div>
-      <div class="flex items-center gap-4">
-        <% if @trace.error? %>
-          <span class="inline-flex items-center px-3 py-1 rounded-full text-sm font-medium bg-red-100 text-red-800 dark:bg-red-900 dark:text-red-200">
-            Error
-          </span>
-        <% else %>
-          <span class="inline-flex items-center px-3 py-1 rounded-full text-sm font-medium bg-green-100 text-green-800 dark:bg-green-900 dark:text-green-200">
-            Success
-          </span>
-        <% end %>
-        <span class="text-sm text-gray-500">
-          <%= @trace.timestamp&.strftime("%Y-%m-%d %H:%M:%S") %>
-        </span>
-      </div>
-    </div>
-  </div>
-
-  <div class="p-4">
-    <%= render partial: "trace_detail", locals: { trace: @trace } %>
-  </div>
-</div>
+<p><%= link_to "Back to traces", traces_path %></p>
+<article class="panel">
+  <h1><%= @trace.display_name %></h1>
+  <p class="badge <%= @trace.error? ? 'error' : 'success' %>"><%= @trace.error? ? "Error" : "Success" %></p>
+  <p><time><%= @trace.timestamp&.strftime("%Y-%m-%d %H:%M:%S") %></time></p>
+  <%= render "trace_detail", trace: @trace %>
+</article>

data/app/views/layouts/action_agent/application.html.erb

@@ -1,94 +1,34 @@
 <!DOCTYPE html>
-<html lang="en" class="h-full">
+<html lang="en">
 <head>
   <meta charset="UTF-8">
   <meta name="viewport" content="width=device-width, initial-scale=1.0">
-  <title>ActiveAgent Dashboard</title>
-  <script src="https://cdn.tailwindcss.com"></script>
-  <script src="https://unpkg.com/@hotwired/turbo@7.3.0/dist/turbo.es2017-esm.js" type="module"></script>
-  <script src="https://unpkg.com/@hotwired/stimulus@3.2.2/dist/stimulus.umd.js"></script>
-  <style>
-    [data-turbo-preview] { opacity: 0.5; }
-    .span-bar { transition: all 0.2s ease; }
-    .span-bar:hover { transform: scaleY(1.1); }
-  </style>
-  <script>
-    // Initialize Stimulus
-    const application = Stimulus.Application.start();
-
-    // Traces controller for expandable rows
-    application.register("traces", class extends Stimulus.Controller {
-      static targets = ["row", "detail"]
-
-      toggle(event) {
-        const traceId = event.currentTarget.dataset.traceId;
-        const detail = document.getElementById(`detail-${traceId}`);
-        if (detail) {
-          detail.classList.toggle("hidden");
-        }
-      }
-    });
-
-    // Auto-refresh controller
-    application.register("refresh", class extends Stimulus.Controller {
-      static values = { interval: Number }
-
-      connect() {
-        if (this.intervalValue > 0) {
-          this.startRefresh();
-        }
-      }
-
-      disconnect() {
-        this.stopRefresh();
-      }
-
-      startRefresh() {
-        this.timer = setInterval(() => {
-          Turbo.visit(window.location.href, { action: "replace" });
-        }, this.intervalValue);
-      }
-
-      stopRefresh() {
-        if (this.timer) clearInterval(this.timer);
-      }
-    });
-  </script>
+  <title>ActiveAgent Trace Console</title>
+  <% if @refresh_interval %>
+    <meta http-equiv="refresh" content="<%= @refresh_interval %>">
+  <% end %>
+  <%= render "action_agent/traces/styles" %>
 </head>
-<body class="h-full bg-gray-50 dark:bg-gray-900">
-  <div class="min-h-full">
-    <!-- Header -->
-    <nav class="bg-white dark:bg-gray-800 border-b border-gray-200 dark:border-gray-700">
-      <div class="mx-auto max-w-7xl px-4 sm:px-6 lg:px-8">
-        <div class="flex h-16 items-center justify-between">
-          <div class="flex items-center">
-            <div class="flex-shrink-0">
-              <span class="text-xl font-bold text-indigo-600 dark:text-indigo-400">
-                ActiveAgent
-              </span>
-            </div>
-            <div class="ml-10 flex items-baseline space-x-4">
-              <%= link_to "Traces",
-                  traces_path,
-                  class: "px-3 py-2 rounded-md text-sm font-medium #{request.path == traces_path ? 'bg-indigo-100 text-indigo-700 dark:bg-indigo-900 dark:text-indigo-200' : 'text-gray-600 hover:bg-gray-100 dark:text-gray-300 dark:hover:bg-gray-700'}" %>
-              <%= link_to "Metrics",
-                  metrics_traces_path,
-                  class: "px-3 py-2 rounded-md text-sm font-medium #{request.path == metrics_traces_path ? 'bg-indigo-100 text-indigo-700 dark:bg-indigo-900 dark:text-indigo-200' : 'text-gray-600 hover:bg-gray-100 dark:text-gray-300 dark:hover:bg-gray-700'}" %>
-            </div>
-          </div>
-          <div class="text-sm text-gray-500 dark:text-gray-400">
-            Local Telemetry
-          </div>
-        </div>
-      </div>
+<body class="trace-console">
+  <header>
+    <nav aria-label="Trace console">
+      <strong>ActiveAgent</strong>
+      <%= link_to "Traces", traces_path %>
+      <%= link_to "Metrics", metrics_traces_path %>
+      <span class="muted">Local Telemetry</span>
     </nav>
-
-    <!-- Main content -->
-    <main>
-      <div class="mx-auto max-w-7xl py-6 px-4 sm:px-6 lg:px-8">
-        <%= yield %>
-      </div>
-    </main>
-  </div>
+  </header>
+  <main>
+    <div class="refresh-controls">
+      <% if @refresh_interval %>
+        <span>Refreshes every <%= @refresh_interval %> seconds.</span>
+        <%= link_to "Stop refreshing", "#{request.path}?#{request.query_parameters.except('refresh').to_query}" %>
+      <% else %>
+        <%= link_to "Auto-refresh", "#{request.path}?#{request.query_parameters.merge('refresh' => '30').to_query}" %>
+      <% end %>
+      <%= link_to "Refresh now", request.path + (request.query_string.present? ? "?#{request.query_string}" : "") %>
+    </div>
+    <%= yield %>
+  </main>
 </body>
 </html>

data/config/routes.rb

@@ -29,8 +29,13 @@ ActionAgent::Engine.routes.draw do
 
     # The collector for evaluation reports an application ran itself
     # (ActiveAgent::Evals::Publisher), at <mount>/api/evaluation_reports.
-    # Authenticated like trace ingest, with a bearer token.
+    # Authenticated like trace ingest, with a bearer token. GET
+    # <mount>/api/evaluation_reports/:run_id exports a stored run in the
+    # envelope the collector takes. A run_id may hold dots, so the path has
+    # no format extension.
     resources :evaluation_reports, only: [ :create ]
+    get "evaluation_reports/:run_id", to: "evaluation_reports#show", as: :evaluation_report,
+                                      constraints: { run_id: %r{[^/]+} }, format: false
 
     post "feedback", to: "feedback_ingest#create"
     get "feedback", to: "feedbacks#index"
@@ -59,6 +64,8 @@ ActionAgent::Engine.routes.draw do
       end
       collection do
         get :presets
+        # The builder's roster for an agent not created yet.
+        get :tool_roster, action: :new_tool_roster
       end
     end
 
@@ -101,6 +108,8 @@ ActionAgent::Engine.routes.draw do
       resources :code_sessions, only: [ :index, :create, :show ] do
         member do
           post :cancel
+          # Another turn of the session's conversation.
+          post :messages
         end
       end
       resource :claude_login, only: [ :show, :create, :destroy ], controller: "claude_logins" do

data/lib/action_agent/engine.rb

@@ -80,6 +80,12 @@ module ActionAgent
       app.config.filter_parameters += [ :credential, :api_key, :access_token, :password, :claude_login_code, /\Arecording_events\z/, /(?:\A|\.)(?:answer|value)\z/i ]
     end
 
+    # Evaluation run reports are served as Markdown (`report.md`). Rails
+    # registers text/markdown as :md from 7.1; an older host gets it here.
+    initializer "action_agent.markdown_mime_type" do
+      Mime::Type.register "text/markdown", :md, [], %w[md markdown] unless Mime[:md]
+    end
+
     # The project setup assistant's request_secret answers go to its
     # handler. Registered on every reload, so the handler is always the
     # current class's.

data/lib/action_agent/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module ActionAgent
-  VERSION = "1.9.2"
+  VERSION = "1.9.4"
 end

data/lib/action_agent.rb

@@ -111,6 +111,9 @@ module ActionAgent
   # What ActionAgent.claude_code_auth may be set to.
   CLAUDE_CODE_AUTH_MODES = %i[api_key local_login sandbox_login].freeze
 
+  # What ActionAgent.codex_auth may be set to.
+  CODEX_AUTH_MODES = %i[api_key local_login].freeze
+
   # What ActionAgent.provider_key_scope may be set to.
   PROVIDER_KEY_SCOPES = %i[organization personal_override].freeze
 
@@ -134,6 +137,9 @@ module ActionAgent
   #                           they are handed to, or delete a project that has
   #                           them
   #   :take_over_browser      drive a run's browser by hand
+  #   :chat_code_session      send a message to a Claude Code or Codex
+  #                           conversation someone may have started (the
+  #                           subject is its first session)
   #   :manage_recordings      delete a session recording
   #   :replace_scenarios      replace or merge an evaluation's scenarios
   PERMISSION_ACTIONS = %i[
@@ -144,6 +150,7 @@ module ActionAgent
     answer_input_request
     manage_project_secrets
     take_over_browser
+    chat_code_session
     manage_recordings
     replace_scenarios
   ].freeze
@@ -589,10 +596,35 @@ module ActionAgent
     # @return [Integer]
     attr_accessor :claude_code_timeout
 
-    # Codex CLI sessions use the owner's OpenAI API key and workspace-write
-    # sandboxing. They share checkout lifecycle and cancellation with Claude.
+    # Codex CLI sessions run with workspace-write sandboxing, and share the
+    # checkout lifecycle and cancellation with Claude Code.
     attr_accessor :codex_command, :codex_timeout
 
+    # How Codex sessions authenticate.
+    #
+    # :api_key (the default) runs them on the OpenAI API key the owner
+    # connected in Settings -> Integrations, handed to the session as
+    # CODEX_API_KEY with a Codex home of the sandbox's own. Usage is billed
+    # to that key's project.
+    #
+    # :local_login runs `codex` on whatever login this machine's user set up
+    # with `codex login` (a ChatGPT plan, or a key given to the CLI), which
+    # Codex keeps under ~/.codex. The dashboard never reads, copies or
+    # stores it; it only asks `codex login status` whether there is one.
+    # That login is the dashboard user's own, so this works with the :local
+    # sandbox backend only, and other backends refuse Codex sessions.
+    # @return [Symbol] :api_key or :local_login
+    attr_reader :codex_auth
+
+    def codex_auth=(value)
+      mode = value.to_s.to_sym
+      unless CODEX_AUTH_MODES.include?(mode)
+        raise ArgumentError, "ActionAgent.codex_auth must be :api_key or :local_login, not #{value.inspect}"
+      end
+
+      @codex_auth = mode
+    end
+
     # How Claude Code sessions authenticate.
     #
     # :api_key (the default) runs them on the Anthropic API key the owner
@@ -719,6 +751,16 @@ module ActionAgent
     # @return [String, nil]
     attr_accessor :sign_out_path
 
+    # Items the host app adds to the header's user menu, above "Sign out":
+    # its workspace and member pages, its account settings, and for platform
+    # admins its admin pages. A Proc receives the signed-in user and the
+    # current owner and returns an Array of { label:, href: } Hashes (an
+    # Array is used as is), so the host decides per user what to show. The
+    # engine has no pages of its own to link here; unset, or when the Proc
+    # raises (logged), the menu holds "Sign out" alone.
+    # @return [Proc, Array<Hash>, nil]
+    attr_accessor :user_menu_links
+
     # Where a browser is sent when it asks for a dashboard page without a
     # valid session — the host app's sign-in page. Unset, an unauthenticated
     # page request gets a minimal session-expired page instead of a bare
@@ -1151,6 +1193,27 @@ module ActionAgent
 
     # The tenant +owner+ belongs to. Identity unless the host app says
     # otherwise, which is right for every single-tenant install.
+    # The host's user menu items for +user+ in +owner+'s dashboard, as
+    # [{ label:, href: }] with blank or malformed entries dropped. Empty when
+    # user_menu_links is unset or raises, which is logged, not raised: a
+    # broken menu seam must not take the dashboard down with it.
+    # @return [Array<Hash{Symbol => String}>]
+    def user_menu_links_for(user, owner)
+      links = user_menu_links
+      links = links.call(user, owner) if links.respond_to?(:call)
+      Array(links).filter_map do |link|
+        label = link[:label] || link["label"]
+        href = link[:href] || link["href"]
+        next if label.blank? || href.blank?
+        next unless href.to_s.match?(%r{\A(/|https?://)})
+
+        { label: label.to_s, href: href.to_s }
+      end
+    rescue StandardError => error
+      Rails.logger.warn("[ActionAgent] user_menu_links failed: #{error.class}: #{error.message}")
+      []
+    end
+
     def tenant_for(owner)
       return owner if tenant_resolver.nil?
 
@@ -1248,6 +1311,7 @@ module ActionAgent
       @claude_code_timeout = 1800
       @codex_command = "codex"
       @codex_timeout = 1800
+      @codex_auth = :api_key
       @claude_code_auth = :api_key
       @claude_code_hosted_login_enabled = false
       @claude_code_login_timeout = 300
@@ -1276,6 +1340,7 @@ module ActionAgent
       @usage_resolver = nil
       @upgrade_url = nil
       @sign_out_path = nil
+      @user_menu_links = nil
       @sign_in_path = nil
       @mcp_catalog = []
       @agent_actor_resolver = nil

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

@@ -315,9 +315,10 @@ ActionAgent.configure do |config|
   # UI
   # ==========================================================================
   #
-  # Custom layout for dashboard views (the default layout loads Tailwind,
-  # Turbo and Stimulus from CDNs — override for CSP-strict or air-gapped
-  # environments).
+  # Custom layout for dashboard views. The default layouts load nothing
+  # from external hosts, so this is only needed to wrap the dashboard in
+  # your application's own chrome. A custom layout replaces the trace
+  # console's built-in styles and refresh controls.
   #
   # config.layout = "application"
 end

data/lib/generators/action_agent/templates/migrations/028_add_conversations_to_code_sessions.rb.erb

@@ -0,0 +1,18 @@
+# frozen_string_literal: true
+
+# Makes Claude Code and Codex sessions conversations: each session is one
+# turn, and a turn after the first names the conversation's first session as
+# its root. A conversation resumes the same CLI session (resume_session_id),
+# and a turn may carry a note from the dashboard ahead of the person's
+# message (context_note), such as who drove the browser since the last turn.
+class AddConversationsToCodeSessions < ActiveRecord::Migration<%= migration_version %>
+  def change
+    sessions = "#{ActionAgent.table_name_prefix}code_sessions"
+    add_column sessions, :conversation_root_id, :bigint
+    add_column sessions, :turn_index, :integer, default: 0, null: false
+    add_column sessions, :origin, :string
+    add_column sessions, :context_note, :text
+    add_column sessions, :resume_session_id, :string
+    add_index sessions, :conversation_root_id
+  end
+end

data/lib/tasks/evaluations.rake

@@ -0,0 +1,78 @@
+# frozen_string_literal: true
+
+# Evaluation run reports from the command line: a run out as the version-1
+# envelope the evaluation report collector takes (or as its HTML report page,
+# or as GitHub-flavored Markdown for a pull request), and an envelope in
+# through the collector's own import. On a multi-tenant install name the
+# tenant with ACCOUNT_ID=<id>.
+#
+#   bin/rails action_agent:evaluations:export RUN=dashboard-run-12-20261021093000 FILE=tmp/run.json
+#   bin/rails action_agent:evaluations:export EVALUATION=3 RUN=12 FORMAT=html FILE=tmp/run.html
+#   bin/rails action_agent:evaluations:export EVALUATION=3 RUN=12 FORMAT=md > tmp/run.md
+#   bin/rails action_agent:evaluations:import FILE=tmp/run.json
+namespace :action_agent do
+  namespace :evaluations do
+    desc "Write a run's report (RUN=run_id, or EVALUATION=id RUN=id; FORMAT=json|html|md; FILE=path, standard output without FILE)"
+    task export: :environment do
+      run = action_agent_evaluation_run!
+      export = ActionAgent::EvaluationReportExport.new(run)
+      format = ENV.fetch("FORMAT", "json").downcase
+      body =
+        case format
+        when "json" then export.to_json
+        when "html" then run.to_report.to_html
+        when "md", "markdown" then export.markdown
+        else abort "FORMAT is json, html or md"
+        end
+
+      if ENV["FILE"].present?
+        File.write(ENV["FILE"], body)
+        puts "Wrote run #{export.run_id} of #{run.evaluation.name} (#{format}) to #{ENV['FILE']}"
+      else
+        puts body
+      end
+    rescue ActionAgent::EvaluationReportExport::NotExportable => e
+      abort "Export refused: #{e.message}"
+    end
+
+    desc "Store a version-1 evaluation report envelope (FILE=path [ACCOUNT_ID=id]); importing it again changes nothing"
+    task import: :environment do
+      path = ENV.fetch("FILE") { abort "Name the file: FILE=tmp/run.json" }
+      abort "No file at #{path}" unless File.file?(path)
+      abort "Unavailable: #{ActionAgent::EvaluationReportImport.unavailable_reason}" if ActionAgent::EvaluationReportImport.unavailable_reason
+
+      account = action_agent_evaluation_account
+      run, duplicate = ActionAgent::EvaluationReportImport.call(account: account, payload: JSON.parse(File.read(path)))
+      ActionAgent.record_usage(account, :evaluation_report) unless duplicate
+      puts "#{duplicate ? 'Already stored (identical retry)' : 'Stored'}: run #{ActionAgent::EvaluationReportExport.run_id_for(run)} " \
+           "as run ##{run.id} of evaluation ##{run.evaluation_id} (#{run.evaluation.name}), #{run.status}"
+    rescue JSON::ParserError
+      abort "Import refused: #{path} is not JSON"
+    rescue ActionAgent::EvaluationReportImport::Error, ActiveRecord::RecordInvalid => e
+      abort "Import refused: #{e.message}"
+    end
+  end
+end
+
+# The tenant, from ACCOUNT_ID (ACCOUNT also works); nil on a single-tenant
+# install.
+def action_agent_evaluation_account
+  id = ENV["ACCOUNT_ID"].presence || ENV["ACCOUNT"].presence
+  return nil unless id
+
+  abort "ACCOUNT_ID names a tenant, which only a multi-tenant install has" unless ActionAgent.multi_tenant?
+  ActionAgent.account_class.to_s.constantize.find_by(id: id) || abort("No account #{id}")
+end
+
+# The run RUN names: with EVALUATION, a run id of that evaluation; without,
+# a run_id as the export endpoint takes it, within the tenant.
+def action_agent_evaluation_run!
+  run_id = ENV.fetch("RUN") { abort "Name the run: RUN=<run_id>, or EVALUATION=<id> RUN=<id>" }
+  if ENV["EVALUATION"].present?
+    evaluation = ActionAgent::Evaluation.find_by(id: ENV["EVALUATION"]) || abort("No evaluation #{ENV['EVALUATION']}")
+    evaluation.evaluation_runs.find_by(id: run_id) || abort("No run #{run_id} in evaluation #{evaluation.id}")
+  else
+    ActionAgent::EvaluationReportExport.find(run_id, account: action_agent_evaluation_account) ||
+      abort("No run with run_id #{run_id.inspect}#{' for that account' if ENV['ACCOUNT_ID'].present? || ENV['ACCOUNT'].present?}")
+  end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: actionagent
 version: !ruby/object:Gem::Version
-  version: 1.9.2
+  version: 1.9.4
 platform: ruby
 authors:
 - Justin Bowen
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-10-20 00:00:00.000000000 Z
+date: 2026-10-25 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: activeagent
@@ -246,7 +246,9 @@ files:
 - app/models/concerns/action_agent/session_recordable.rb
 - app/queries/action_agent/agent_executions.rb
 - app/queries/action_agent/metrics_report.rb
+- app/queries/action_agent/trace_window_report.rb
 - app/serializers/action_agent/agent_message_serializer.rb
+- app/serializers/action_agent/code_conversation_serializer.rb
 - app/serializers/action_agent/evaluation_serializer.rb
 - app/serializers/action_agent/input_request_serializer.rb
 - app/serializers/action_agent/interaction_preview.rb
@@ -261,17 +263,23 @@ files:
 - app/services/action_agent/agent_sync.rb
 - app/services/action_agent/agent_tool_roster.rb
 - app/services/action_agent/agent_toolbox.rb
+- app/services/action_agent/browser_client_token.rb
 - app/services/action_agent/browser_live_ticket.rb
 - app/services/action_agent/browser_sidecar.rb
 - app/services/action_agent/browser_sign_in.rb
 - app/services/action_agent/browser_tool_redaction.rb
+- app/services/action_agent/checkout_tools.rb
 - app/services/action_agent/claude_code_auth.rb
+- app/services/action_agent/code_conversation.rb
+- app/services/action_agent/codex_auth.rb
 - app/services/action_agent/codex_session_events.rb
 - app/services/action_agent/dashboard_assistant_service.rb
+- app/services/action_agent/diff_stat.rb
 - app/services/action_agent/draft_pull_request_publisher.rb
 - app/services/action_agent/evaluation_evidence.rb
 - app/services/action_agent/evaluation_fix.rb
 - app/services/action_agent/evaluation_fix_comparison.rb
+- app/services/action_agent/evaluation_report_export.rb
 - app/services/action_agent/evaluation_report_import.rb
 - app/services/action_agent/evaluation_run_cost.rb
 - app/services/action_agent/evaluation_runner_service.rb
@@ -316,10 +324,13 @@ files:
 - app/services/action_agent/session_index.rb
 - app/services/action_agent/session_recording_service.rb
 - app/services/action_agent/session_timeline.rb
+- app/services/action_agent/slack_tools.rb
 - app/services/action_agent/tool_discovery.rb
 - app/views/action_agent/dashboard/index.html.erb
 - app/views/action_agent/github_app_manifests/created.html.erb
 - app/views/action_agent/session_player/show.html.erb
+- app/views/action_agent/traces/_styles.html.erb
+- app/views/action_agent/traces/_summary.html.erb
 - app/views/action_agent/traces/_trace_detail.html.erb
 - app/views/action_agent/traces/index.html.erb
 - app/views/action_agent/traces/metrics.html.erb
@@ -360,9 +371,11 @@ files:
 - lib/generators/action_agent/templates/migrations/025_create_active_agent_scenario_catalogs.rb.erb
 - lib/generators/action_agent/templates/migrations/026_add_evaluation_fixes_to_code_sessions.rb.erb
 - lib/generators/action_agent/templates/migrations/027_create_active_agent_interaction_feedbacks.rb.erb
+- lib/generators/action_agent/templates/migrations/028_add_conversations_to_code_sessions.rb.erb
 - lib/tasks/action_agent.rake
 - lib/tasks/browser.rake
 - lib/tasks/claude_code.rake
+- lib/tasks/evaluations.rake
 - lib/tasks/sandbox.rake
 - lib/tasks/scenario_catalogs.rake
 homepage: https://activeagents.ai