fix(chat): eager pending AssistantMessage to fix Turbo subscribe race (#1657) (#1658)

* fix(chat): persist eager pending assistant message to fix subscribe race When the LLM replies in ~1-2s the assistant message broadcast could fire before the client's Turbo stream subscription was established, leaving the UI stuck on the thinking indicator while the response was already persisted. Create the AssistantMessage as `pending` synchronously in `Chat#ask_assistant_later`, so it is rendered server-side on the chat show page with a "Thinking ..." inline placeholder. The worker then finds and updates the existing row via `append_text!`, which flips the status to `complete` and broadcasts updates against a DOM id that is already in the page — no race possible. On error, the placeholder is destroyed if no content streamed, otherwise demoted to `failed`. Replaces the standalone thinking indicator partial and the `Assistant::Broadcastable` thinking helpers, both now redundant. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(chat): bind each assistant job to its specific pending placeholder Addressing review feedback on #1658: 1. The pending placeholder lookup based on `last pending` was racy — back-to-back user messages would let one job fill another job's placeholder. Pass the placeholder through the job arguments (`AssistantResponseJob.perform_later(user_message, pending)`) so each turn is bound to its own row. 2. In `Assistant::External#respond_to`, the configured/authorized guards raise before the local was bound, leaving rescue cleanup with `nil` and the placeholder visible forever. Bind the parameter first so cleanup can destroy it on the misconfigured path. The kwarg defaults to nil so the API#retry path (`AssistantResponseJob.perform_later(new_message)`) and the model-level test calls continue to work — they fall back to an in-memory new message, restoring the original test count assertions. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(chat): i18n the pending assistant placeholder string Move the hardcoded "Thinking ..." indicator into the locale file per CLAUDE.md i18n guidelines. With i18n.fallbacks enabled, non-en locales fall back to English until translated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * Add thinking label translations * Fix chat pending assistant expectations * Fix external assistant pending test lookup * Scope chat stream targets per chat * Update message broadcast target tests --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-10 14:15:01 +00:00 · 2026-05-03 20:33:29 +02:00
parent 50936000e7
commit ccd6a53071
28 changed files with 91 additions and 92 deletions
--- a/app/controllers/chats_controller.rb
+++ b/app/controllers/chats_controller.rb
@@ -43,7 +43,7 @@ class ChatsController < ApplicationController

  def retry
    @chat.retry_last_message!
-    redirect_to chat_path(@chat, thinking: true)
+    redirect_to chat_path(@chat)
  end

  private
--- a/app/jobs/assistant_response_job.rb
+++ b/app/jobs/assistant_response_job.rb
@@ -1,7 +1,7 @@
 class AssistantResponseJob < ApplicationJob
  queue_as :high_priority

-  def perform(message)
-    message.request_response
+  def perform(message, assistant_message = nil)
+    message.request_response(assistant_message: assistant_message)
  end
 end
--- a/app/models/assistant/base.rb
+++ b/app/models/assistant/base.rb
@@ -1,13 +1,11 @@
 class Assistant::Base
-  include Assistant::Broadcastable
-
  attr_reader :chat

  def initialize(chat)
    @chat = chat
  end

-  def respond_to(message)
+  def respond_to(message, assistant_message: nil)
    raise NotImplementedError, "#{self.class}#respond_to must be implemented"
  end
 end
--- a/app/models/assistant/broadcastable.rb
+++ b/app/models/assistant/broadcastable.rb
@@ -1,12 +0,0 @@
-module Assistant::Broadcastable
-  extend ActiveSupport::Concern
-
-  private
-    def update_thinking(thought)
-      chat.broadcast_update target: "thinking-indicator", partial: "chats/thinking_indicator", locals: { chat: chat, message: thought }
-    end
-
-    def stop_thinking
-      chat.broadcast_remove target: "thinking-indicator"
-    end
-end
--- a/app/models/assistant/builtin.rb
+++ b/app/models/assistant/builtin.rb
@@ -17,12 +17,8 @@ class Assistant::Builtin < Assistant::Base
    @functions = functions
  end

-  def respond_to(message)
-    assistant_message = AssistantMessage.new(
-      chat: chat,
-      content: "",
-      ai_model: message.ai_model
-    )
+  def respond_to(message, assistant_message: nil)
+    assistant_message ||= AssistantMessage.new(chat: chat, content: "", ai_model: message.ai_model)

    llm_provider = get_model_provider(message.ai_model)
    unless llm_provider
@@ -40,7 +36,6 @@ class Assistant::Builtin < Assistant::Base

    responder.on(:output_text) do |text|
      if assistant_message.content.blank?
-        stop_thinking
        Chat.transaction do
          assistant_message.append_text!(text)
          chat.update_latest_response!(latest_response_id)
@@ -51,7 +46,6 @@ class Assistant::Builtin < Assistant::Base
    end

    responder.on(:response) do |data|
-      update_thinking("Analyzing your data...")
      if data[:function_tool_calls].present?
        assistant_message.tool_calls = data[:function_tool_calls]
        latest_response_id = data[:id]
@@ -62,13 +56,13 @@ class Assistant::Builtin < Assistant::Base

    responder.respond(previous_response_id: latest_response_id)
  rescue => e
-    stop_thinking
-    # If we streamed any partial content before the error, the message was
-    # persisted with the default `complete` status. Demote it to `failed` so
-    # `Assistant::Responder#conversation_history` won't feed a broken turn
-    # back into future prompts.
    if assistant_message&.persisted?
-      assistant_message.update_columns(status: "failed")
+      if assistant_message.content.blank?
+        assistant_message.destroy
+      else
+        # Demote partially-streamed turns to `failed` so `Responder#conversation_history` excludes them.
+        assistant_message.update_columns(status: "failed")
+      end
    end
    chat.add_error(e)
  end
--- a/app/models/assistant/external.rb
+++ b/app/models/assistant/external.rb
@@ -33,8 +33,9 @@ class Assistant::External < Assistant::Base
    end
  end

-  def respond_to(message)
+  def respond_to(message, assistant_message: nil)
    response_completed = false
+    assistant_message ||= AssistantMessage.new(chat: chat, content: "", ai_model: "external-agent")

    unless self.class.configured?
      raise Assistant::Error,
@@ -45,12 +46,6 @@ class Assistant::External < Assistant::Base
      raise Assistant::Error, "Your account is not authorized to use the external assistant."
    end

-    assistant_message = AssistantMessage.new(
-      chat: chat,
-      content: "",
-      ai_model: "external-agent"
-    )
-
    client = build_client
    messages = build_conversation_messages

@@ -58,17 +53,10 @@ class Assistant::External < Assistant::Base
      messages: messages,
      user: "sure-family-#{chat.user.family_id}"
    ) do |text|
-      if assistant_message.content.blank?
-        stop_thinking
-        assistant_message.content = text
-        assistant_message.save!
-      else
-        assistant_message.append_text!(text)
-      end
+      assistant_message.append_text!(text)
    end

-    if assistant_message.new_record?
-      stop_thinking
+    if assistant_message.content.blank?
      raise Assistant::Error, "External assistant returned an empty response."
    end

@@ -76,12 +64,10 @@ class Assistant::External < Assistant::Base
    assistant_message.update!(ai_model: model) if model.present?
  rescue Assistant::Error, ActiveRecord::ActiveRecordError => e
    cleanup_partial_response(assistant_message) unless response_completed
-    stop_thinking
    chat.add_error(e)
  rescue => e
    Rails.logger.error("[Assistant::External] Unexpected error: #{e.class} - #{e.message}")
    cleanup_partial_response(assistant_message) unless response_completed
-    stop_thinking
    chat.add_error(Assistant::Error.new("Something went wrong with the external assistant. Check server logs for details."))
  end

@@ -103,7 +89,7 @@ class Assistant::External < Assistant::Base
    end

    def build_conversation_messages
-      chat.conversation_messages.ordered.last(MAX_CONVERSATION_MESSAGES).map do |msg|
+      chat.conversation_messages.where(status: "complete").ordered.last(MAX_CONVERSATION_MESSAGES).map do |msg|
        { role: msg.role, content: msg.content }
      end
    end
--- a/app/models/assistant_message.rb
+++ b/app/models/assistant_message.rb
@@ -7,6 +7,7 @@ class AssistantMessage < Message

  def append_text!(text)
    self.content += text
+    self.status = :complete if pending?
    save!
  end
 end
--- a/app/models/chat.rb
+++ b/app/models/chat.rb
@@ -79,7 +79,7 @@ class Chat < ApplicationRecord

  def add_error(e)
    update!(error: build_error_payload(e).to_json)
-    broadcast_append target: "messages", partial: "chats/error", locals: { chat: self }
+    broadcast_append target: messages_target, partial: "chats/error", locals: { chat: self }
  end

  def presentable_error_message
@@ -93,20 +93,29 @@ class Chat < ApplicationRecord

  def clear_error
    update! error: nil
-    broadcast_remove target: "chat-error"
+    broadcast_remove target: error_target
  end

  def conversation_messages
    messages.where(type: [ "UserMessage", "AssistantMessage" ])
  end

-  def ask_assistant_later(message)
-    clear_error
-    AssistantResponseJob.perform_later(message)
+  def messages_target
+    ActionView::RecordIdentifier.dom_id(self, :messages)
  end

-  def ask_assistant(message)
-    assistant.respond_to(message)
+  def error_target
+    ActionView::RecordIdentifier.dom_id(self, :chat_error)
+  end
+
+  def ask_assistant_later(message)
+    clear_error
+    pending = messages.create!(type: "AssistantMessage", content: "", ai_model: message.ai_model, status: :pending)
+    AssistantResponseJob.perform_later(message, pending)
+  end
+
+  def ask_assistant(message, assistant_message: nil)
+    assistant.respond_to(message, assistant_message: assistant_message)
  end

  private
--- a/app/models/message.rb
+++ b/app/models/message.rb
@@ -8,9 +8,9 @@ class Message < ApplicationRecord
    failed: "failed"
  }

-  validates :content, presence: true
+  validates :content, presence: true, unless: :pending?

-  after_create_commit -> { broadcast_append_to chat, target: "messages" }, if: :broadcast?
+  after_create_commit -> { broadcast_append_to chat, target: chat.messages_target }, if: :broadcast?
  after_update_commit -> { broadcast_update_to chat }, if: :broadcast?

  scope :ordered, -> { order(created_at: :asc) }
--- a/app/models/user_message.rb
+++ b/app/models/user_message.rb
@@ -11,7 +11,7 @@ class UserMessage < Message
    chat.ask_assistant_later(self)
  end

-  def request_response
-    chat.ask_assistant(self)
+  def request_response(assistant_message: nil)
+    chat.ask_assistant(self, assistant_message: assistant_message)
  end
 end
--- a/app/views/assistant_messages/_assistant_message.html.erb
+++ b/app/views/assistant_messages/_assistant_message.html.erb
@@ -1,7 +1,12 @@
 <%# locals: (assistant_message:) %>

 <div id="<%= dom_id(assistant_message) %>">
-  <% if assistant_message.reasoning? %>
+  <% if assistant_message.pending? %>
+    <div class="flex items-start gap-3 mb-6">
+      <%= render "chats/ai_avatar" %>
+      <p class="text-sm text-secondary animate-pulse"><%= t("chats.thinking") %></p>
+    </div>
+  <% elsif assistant_message.reasoning? %>
    <details class="group mb-1">
      <summary class="flex items-center gap-2">
        <p class="text-secondary text-sm">Assistant reasoning</p>
--- a/app/views/chats/_error.html.erb
+++ b/app/views/chats/_error.html.erb
@@ -1,6 +1,6 @@
 <%# locals: (chat:) %>

-<div id="chat-error" class="px-3 py-2 bg-red-100 border border-red-500 rounded-lg">
+<div id="<%= chat.error_target %>" class="px-3 py-2 bg-red-100 border border-red-500 rounded-lg">
  <% if chat.debug_mode? %>
    <div class="overflow-x-auto text-xs p-4 bg-red-200 rounded-md mb-2">
      <code><%= chat.technical_error_message %></code>
--- a/app/views/chats/_thinking_indicator.html.erb
+++ b/app/views/chats/_thinking_indicator.html.erb
@@ -1,6 +0,0 @@
-<%# locals: (chat:, message: "Thinking ...") -%>
-
-<div id="thinking-indicator" class="flex items-start gap-3">
-  <%= render "chats/ai_avatar" %>
-  <p class="text-sm text-secondary animate-pulse"><%= message %></p>
-</div>
--- a/app/views/chats/show.html.erb
+++ b/app/views/chats/show.html.erb
@@ -15,7 +15,7 @@
        <% end %>
      </div>

-      <div id="messages" class="grow overflow-y-auto p-4 space-y-6 lg:pb-4" data-chat-target="messages">
+      <div id="<%= @chat.messages_target %>" class="grow overflow-y-auto p-4 space-y-6 lg:pb-4" data-chat-target="messages">
        <% if @chat.conversation_messages.any? %>
          <% @chat.conversation_messages.ordered.each do |message| %>
            <%= render message %>
@@ -26,10 +26,6 @@
          </div>
        <% end %>

-        <% if params[:thinking].present? %>
-          <%= render "chats/thinking_indicator", chat: @chat %>
-        <% end %>
-
        <% if @chat.error.present? && @chat.needs_assistant_response? %>
          <%= render "chats/error", chat: @chat %>
        <% end %>
--- a/config/locales/views/chats/ca.yml
+++ b/config/locales/views/chats/ca.yml
@@ -5,3 +5,4 @@ ca:
      per Cloudflare Workers AI. Els resultats poden variar ja que el codi es va provar
      principalment amb `gpt-4.1` però els teus tokens no van a cap altre lloc per ser entrenats! 🤖"
    demo_banner_title: Mode de demostració
+    thinking: "Treballant ..."
--- a/config/locales/views/chats/de.yml
+++ b/config/locales/views/chats/de.yml
@@ -3,3 +3,4 @@ de:
  chats:
    demo_banner_title: "Demo-Modus aktiv"
    demo_banner_message: "Sie nutzen ein Open-Weight Qwen3-LLM mit Credits von Cloudflare Workers AI. Die Ergebnisse können variieren, da die Codebasis hauptsächlich mit `gpt-4.1` getestet wurde – Ihre Tokens werden jedoch nicht anderswo zum Training verwendet! 🤖"
+    thinking: "Wird verarbeitet ..."
--- a/config/locales/views/chats/en.yml
+++ b/config/locales/views/chats/en.yml
@@ -3,3 +3,4 @@ en:
  chats:
    demo_banner_title: "Demo Mode Active"
    demo_banner_message: "You are using LLMs via credits provided by Cloudflare Workers AI.  Results may vary since the codebase was tested on `gpt-4.1` but your tokens don't go anywhere else to be trained with! 🤖"
+    thinking: "Thinking ..."
--- a/config/locales/views/chats/es.yml
+++ b/config/locales/views/chats/es.yml
@@ -2,4 +2,5 @@
 es:
  chats:
    demo_banner_title: "Modo de demostración activo"
-    demo_banner_message: "Estás utilizando un LLM Qwen3 de pesos abiertos con créditos proporcionados por Cloudflare Workers AI. Los resultados pueden variar, ya que la base de código se probó principalmente con `gpt-4.1`, ¡pero tus tokens no se enviarán a ningún otro lugar para ser entrenados! 🤖"
+    demo_banner_message: "Estás utilizando un LLM Qwen3 de pesos abiertos con créditos proporcionados por Cloudflare Workers AI. Los resultados pueden variar, ya que la base de código se probó principalmente con `gpt-4.1`, ¡pero tus tokens no se enviarán a ningún otro lugar para ser entrenados! 🤖"
+    thinking: "Procesando ..."
--- a/config/locales/views/chats/fr.yml
+++ b/config/locales/views/chats/fr.yml
@@ -3,3 +3,4 @@ fr:
  chats:
    demo_banner_title: "Mode Démo Actif"
    demo_banner_message: "Vous utilisez un LLM Qwen3 open-weights avec des crédits fournis par Cloudflare Workers AI. Les résultats peuvent varier car le code a été principalement testé sur `gpt-4.1` mais vos tokens ne sont envoyés nulle part ailleurs pour être entraînés !"
+    thinking: "Traitement en cours ..."
--- a/config/locales/views/chats/nl.yml
+++ b/config/locales/views/chats/nl.yml
@@ -3,3 +3,4 @@ nl:
  chats:
    demo_banner_title: "Demo Modus Actief"
    demo_banner_message: "U gebruikt een LLM met credits verstrekt door Cloudflare Workers AI. Resultaten kunnen variëren aangezien de codebase voornamelijk is getest op `gpt-4.1`, maar uw tokens worden nergens anders voor training gebruikt! 🤖"
+    thinking: "Bezig ..."
--- a/config/locales/views/chats/pl.yml
+++ b/config/locales/views/chats/pl.yml
@@ -3,3 +3,4 @@ pl:
  chats:
    demo_banner_title: "Aktywny tryb demo"
    demo_banner_message: "Używasz modelu LLM z kredytami udostępnionymi przez Cloudflare Workers AI. Wyniki mogą się różnić, ponieważ kod aplikacji był głównie testowany na `gpt-4.1`, ale Twoje tokeny nie są nigdzie dalej wykorzystywane do trenowania! 🤖"
+    thinking: "Przetwarzanie ..."
--- a/config/locales/views/chats/zh-CN.yml
+++ b/config/locales/views/chats/zh-CN.yml
@@ -3,3 +3,4 @@ zh-CN:
  chats:
    demo_banner_message: "您正在使用由 Cloudflare Workers AI 提供额度的开源权重 Qwen3 大语言模型。由于代码库主要在 `gpt-4.1` 上测试，结果可能有所不同，但您的令牌不会被用于其他地方进行训练！🤖"
    demo_banner_title: 演示模式已激活
+    thinking: "处理中 ..."
--- a/config/locales/views/chats/zh-TW.yml
+++ b/config/locales/views/chats/zh-TW.yml
@@ -3,3 +3,4 @@ zh-TW:
  chats:
    demo_banner_message: "您正在使用由 Cloudflare Workers AI 提供額度的開源權重 Qwen3 大語言模型。由於程式碼庫主要在 `gpt-4.1` 上測試，結果可能有所不同，但您的令牌不會被用於其他地方進行訓練！🤖"
    demo_banner_title: 演示模式已啟動
+    thinking: "處理中 ..."
--- a/test/controllers/api/v1/messages_controller_test.rb
+++ b/test/controllers/api/v1/messages_controller_test.rb
@@ -37,7 +37,7 @@ class Api::V1::MessagesControllerTest < ActionDispatch::IntegrationTest
  end

  test "should create message with write scope" do
-    assert_difference "Message.count" do
+    assert_difference "UserMessage.count" do
      post "/api/v1/chats/#{@chat.id}/messages",
        params: { content: "Test message", model: "gpt-4" },
        headers: bearer_auth_header(@write_token)
--- a/test/models/assistant_message_test.rb
+++ b/test/models/assistant_message_test.rb
@@ -12,7 +12,7 @@ class AssistantMessageTest < ActiveSupport::TestCase
    streams = capture_turbo_stream_broadcasts(@chat)
    assert_equal 2, streams.size
    assert_equal "append", streams.first["action"]
-    assert_equal "messages", streams.first["target"]
+    assert_equal @chat.messages_target, streams.first["target"]
    assert_equal "update", streams.last["action"]
    assert_equal "assistant_message_#{message.id}", streams.last["target"]
  end
--- a/test/models/assistant_test.rb
+++ b/test/models/assistant_test.rb
@@ -226,11 +226,13 @@ class AssistantTest < ActiveSupport::TestCase
      "EXTERNAL_ASSISTANT_URL" => "http://localhost:18789/v1/chat",
      "EXTERNAL_ASSISTANT_TOKEN" => "test-token"
    ) do
-      assert_difference "AssistantMessage.count", 1 do
-        assistant.respond_to(@message)
+      assistant_message = pending_assistant_message
+
+      assert_no_difference "AssistantMessage.count" do
+        assistant.respond_to(@message, assistant_message: assistant_message)
      end

-      response_msg = @chat.messages.where(type: "AssistantMessage").last
+      response_msg = assistant_message.reload
      assert_equal "Your net worth is $124,200.", response_msg.content
      assert_equal "ext-agent:main", response_msg.ai_model
    end
@@ -368,12 +370,13 @@ class AssistantTest < ActiveSupport::TestCase
      "EXTERNAL_ASSISTANT_TOKEN" => "test-token"
    ) do
      assistant = Assistant::External.new(@chat)
-      assistant.respond_to(@message)
+      assistant_message = pending_assistant_message
+      assistant.respond_to(@message, assistant_message: assistant_message)

      @chat.reload
      assert_nil @chat.error

-      response = @chat.messages.where(type: "AssistantMessage").last
+      response = assistant_message.reload
      assert_equal "Based on your accounts, your net worth is $50,000.", response.content
      assert_equal "ext-agent:main", response.ai_model
    end
@@ -414,9 +417,10 @@ class AssistantTest < ActiveSupport::TestCase
      "EXTERNAL_ASSISTANT_URL" => "http://localhost:18789/v1/chat",
      "EXTERNAL_ASSISTANT_TOKEN" => "test-token"
    ) do
-      assistant.respond_to(@message)
+      assistant_message = pending_assistant_message
+      assistant.respond_to(@message, assistant_message: assistant_message)

-      response = @chat.messages.where(type: "AssistantMessage").last
+      response = assistant_message.reload
      assert_equal "ext-agent:custom", response.ai_model
    end
  end
@@ -536,6 +540,10 @@ class AssistantTest < ActiveSupport::TestCase
      capture
    end

+    def pending_assistant_message
+      @chat.messages.where(type: "AssistantMessage", status: "pending").order(:created_at).last
+    end
+
    def provider_function_request(id:, call_id:, function_name:, function_args:)
      Provider::LlmConcept::ChatFunctionRequest.new(
        id: id,
--- a/test/models/chat_test.rb
+++ b/test/models/chat_test.rb
@@ -18,14 +18,25 @@ class ChatTest < ActiveSupport::TestCase
    assert_equal 3, chat.conversation_messages.count
  end

+  test "uses chat-scoped stream targets" do
+    first_chat = chats(:one)
+    second_chat = chats(:two)
+
+    assert_not_equal "messages", first_chat.messages_target
+    assert_not_equal "chat-error", first_chat.error_target
+    assert_not_equal first_chat.messages_target, second_chat.messages_target
+    assert_not_equal first_chat.error_target, second_chat.error_target
+  end
+
  test "creates with initial message" do
    prompt = "Test prompt"

    assert_difference "@user.chats.count", 1 do
      chat = @user.chats.start!(prompt, model: "gpt-4.1")

-      assert_equal 1, chat.messages.count
+      assert_equal 2, chat.messages.count
      assert_equal 1, chat.messages.where(type: "UserMessage").count
+      assert_equal 1, chat.messages.where(type: "AssistantMessage", status: "pending").count
    end
  end

@@ -35,8 +46,8 @@ class ChatTest < ActiveSupport::TestCase
    assert_difference "@user.chats.count", 1 do
      chat = @user.chats.start!(prompt, model: nil)

-      assert_equal 1, chat.messages.count
-      assert_equal Provider::Openai::DEFAULT_MODEL, chat.messages.first.ai_model
+      assert_equal 2, chat.messages.count
+      assert_equal Provider::Openai::DEFAULT_MODEL, chat.messages.find_by!(type: "UserMessage").ai_model
    end
  end

@@ -46,8 +57,8 @@ class ChatTest < ActiveSupport::TestCase
    assert_difference "@user.chats.count", 1 do
      chat = @user.chats.start!(prompt, model: "")

-      assert_equal 1, chat.messages.count
-      assert_equal Provider::Openai::DEFAULT_MODEL, chat.messages.first.ai_model
+      assert_equal 2, chat.messages.count
+      assert_equal Provider::Openai::DEFAULT_MODEL, chat.messages.find_by!(type: "UserMessage").ai_model
    end
  end

@@ -57,7 +68,7 @@ class ChatTest < ActiveSupport::TestCase
    with_env_overrides OPENAI_MODEL: "custom-model" do
      chat = @user.chats.start!(prompt, model: "")

-      assert_equal "custom-model", chat.messages.first.ai_model
+      assert_equal "custom-model", chat.messages.find_by!(type: "UserMessage").ai_model
    end
  end

--- a/test/models/user_message_test.rb
+++ b/test/models/user_message_test.rb
@@ -14,7 +14,7 @@ class UserMessageTest < ActiveSupport::TestCase
    streams = capture_turbo_stream_broadcasts(@chat)
    assert_equal 2, streams.size
    assert_equal "append", streams.first["action"]
-    assert_equal "messages", streams.first["target"]
+    assert_equal @chat.messages_target, streams.first["target"]
    assert_equal "update", streams.last["action"]
    assert_equal "user_message_#{message.id}", streams.last["target"]
  end