FEATURE: Topic summarization (#41)

* FEATURE: Topic summarization Summarize topics using the TopicView's "summary" filter. The UI is similar to what we do for chat, but we don't allow the user to select a timeframe. Co-authored-by: Rafael dos Santos Silva <xfalcox@gmail.com>
2023-04-19 16:57:31 -04:00 · 2023-04-19 16:57:31 -04:00 · 38e007a3a5
parent 9783e3b025
commit 38e007a3a5
18 changed files with 428 additions and 155 deletions
--- a/app/controllers/discourse_ai/summarization/summary_controller.rb
+++ b/app/controllers/discourse_ai/summarization/summary_controller.rb
@ -6,19 +6,28 @@ module DiscourseAi
      requires_plugin ::DiscourseAi::PLUGIN_NAME
      requires_login

-      VALID_SINCE_VALUES = [1, 3, 6, 12, 24]
+      VALID_SINCE_VALUES = [1, 3, 6, 12, 24, 72, 168]
+      VALID_TARGETS = %w[chat_channel topic]

-      def chat_channel
-        since = params[:since].to_i
+      def show
+        raise PluginDisabled unless SiteSetting.ai_summarization_enabled
+        target_type = params[:target_type]

-        raise Discourse::InvalidParameters.new(:since) if !VALID_SINCE_VALUES.include?(since)
-        chat_channel = Chat::Channel.find_by(id: params[:chat_channel_id])
-        raise Discourse::NotFound.new(:chat_channel) if !chat_channel
+        raise Discourse::InvalidParameters.new(:target_type) if !VALID_TARGETS.include?(target_type)

-        if !(SiteSetting.discourse_ai_enabled && SiteSetting.ai_summarization_enabled)
-          raise PluginDisabled
+        since = nil
+
+        if target_type == "chat_channel"
+          since = params[:since].to_i
+          raise Discourse::InvalidParameters.new(:since) if !VALID_SINCE_VALUES.include?(since)
+          target = Chat::Channel.find_by(id: params[:target_id])
+          raise Discourse::NotFound.new(:chat_channel) if !target
+          raise Discourse::InvalidAccess if !guardian.can_join_chat_channel?(target)
+        else
+          target = Topic.find_by(id: params[:target_id])
+          raise Discourse::NotFound.new(:topic) if !target
+          raise Discourse::InvalidAccess if !guardian.can_see_topic?(target)
        end
-        raise Discourse::InvalidAccess if !guardian.can_join_chat_channel?(chat_channel)

        RateLimiter.new(
          current_user,
@ -28,7 +37,8 @@ module DiscourseAi
        ).performed!

        hijack do
-          summary = DiscourseAi::Summarization::SummaryGenerator.new(chat_channel).summarize!(since)
+          summary =
+            DiscourseAi::Summarization::SummaryGenerator.new(target, current_user).summarize!(since)

          render json: { summary: summary }, status: 200
        end
--- a/assets/javascripts/discourse/components/ai-summary.hbs
+++ b/assets/javascripts/discourse/components/ai-summary.hbs
@ -0,0 +1,32 @@
+<DModalBody @title="discourse_ai.summarization.title">
+  {{#if @allowTimeframe}}
+    <span>{{i18n "discourse_ai.summarization.description"}}</span>
+    <ComboBox
+      @value={{this.sinceHours}}
+      @content={{this.sinceOptions}}
+      @onChange={{action this.summarize}}
+      @valueProperty="value"
+      @class="summarization-since"
+    />
+  {{/if}}
+
+  <div class="channel-summary">
+    <ConditionalLoadingSpinner @condition={{this.loading}} />
+
+    {{#unless this.loading}}
+      <Textarea @value={{this.summary}} disabled="true" class="summary-area" />
+    {{/unless}}
+  </div>
+
+</DModalBody>
+
+<div class="modal-footer">
+  {{#unless @allowTimeframe}}
+    <DButton
+      @class="btn-primary create"
+      @action={{this.summarize}}
+      @label="discourse_ai.summarization.summarize"
+    />
+  {{/unless}}
+  <DModalCancel @close={{route-action "closeModal"}} />
+</div>
--- a/assets/javascripts/discourse/components/ai-summary.js
+++ b/assets/javascripts/discourse/components/ai-summary.js
@ -0,0 +1,83 @@
+import Component from "@glimmer/component";
+import { tracked } from "@glimmer/tracking";
+import { action } from "@ember/object";
+import { ajax } from "discourse/lib/ajax";
+import { popupAjaxError } from "discourse/lib/ajax-error";
+import I18n from "I18n";
+
+export default class AiSummary extends Component {
+  @tracked sinceHours = null;
+  @tracked loading = false;
+  @tracked availableSummaries = {};
+  @tracked summary = null;
+  sinceOptions = [
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 1 }),
+      value: 1,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 3 }),
+      value: 3,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 6 }),
+      value: 6,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 12 }),
+      value: 12,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 24 }),
+      value: 24,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 72 }),
+      value: 72,
+    },
+    {
+      name: I18n.t("discourse_ai.summarization.since", { count: 168 }),
+      value: 168,
+    },
+  ];
+
+  get canSummarize() {
+    return (!this.args.allowTimeframe || this.sinceHours) && !this.loading;
+  }
+
+  @action
+  summarize(value) {
+    this.loading = true;
+    const attrs = {
+      target_id: this.args.targetId,
+      target_type: this.args.targetType,
+    };
+
+    if (this.args.allowTimeframe) {
+      this.sinceHours = value;
+
+      if (this.availableSummaries[this.sinceHours]) {
+        this.summary = this.availableSummaries[this.sinceHours];
+        this.loading = false;
+        return;
+      } else {
+        attrs.since = this.sinceHours;
+      }
+    }
+
+    ajax("/discourse-ai/summarization/summary", {
+      method: "POST",
+      data: attrs,
+    })
+      .then((data) => {
+        if (this.args.allowTimeframe) {
+          this.availableSummaries[this.sinceHours] = data.summary;
+          this.summary = this.availableSummaries[this.sinceHours];
+        } else {
+          this.summary = data.summary;
+        }
+      })
+      .catch(popupAjaxError)
+      .finally(() => (this.loading = false));
+  }
+}
--- a/assets/javascripts/discourse/components/chat-channel-summary.hbs
+++ b/assets/javascripts/discourse/components/chat-channel-summary.hbs
@ -1,23 +0,0 @@
-<DModalBody @title="discourse_ai.summarization.title">
-  <span>{{i18n "discourse_ai.summarization.description"}}</span>
-  <ComboBox
-    @value={{this.sinceHours}}
-    @content={{this.sinceOptions}}
-    @onChange={{action this.summarize}}
-    @valueProperty="value"
-    @class="summarization-since"
-  />
-
-  <div class="channel-summary">
-    <ConditionalLoadingSpinner @condition={{this.loading}} />
-
-    {{#unless this.loading}}
-      <Textarea @value={{this.summary}} disabled="true" class="summary-area" />
-    {{/unless}}
-  </div>
-
-</DModalBody>
-
-<div class="modal-footer">
-  <DModalCancel @close={{route-action "closeModal"}} />
-</div>
--- a/assets/javascripts/discourse/components/chat-channel-summary.js
+++ b/assets/javascripts/discourse/components/chat-channel-summary.js
@ -1,69 +0,0 @@
-import Component from "@glimmer/component";
-import { tracked } from "@glimmer/tracking";
-import { action } from "@ember/object";
-import { ajax } from "discourse/lib/ajax";
-import { popupAjaxError } from "discourse/lib/ajax-error";
-import I18n from "I18n";
-
-export default class ChatChannelSummary extends Component {
-  @tracked sinceHours = null;
-  @tracked loading = false;
-  @tracked availableSummaries = {};
-  @tracked summary = null;
-  sinceOptions = [
-    {
-      name: I18n.t("discourse_ai.summarization.since", { hours: "1" }),
-      value: 1,
-    },
-    {
-      name: I18n.t("discourse_ai.summarization.since", { hours: "3" }),
-      value: 3,
-    },
-    {
-      name: I18n.t("discourse_ai.summarization.since", { hours: "6" }),
-      value: 6,
-    },
-    {
-      name: I18n.t("discourse_ai.summarization.since", { hours: "12" }),
-      value: 12,
-    },
-    {
-      name: I18n.t("discourse_ai.summarization.since", { hours: "24" }),
-      value: 24,
-    },
-  ];
-
-  get modalTitle() {
-    return I18n.t("discourse_ai.summarization.modal_title", {
-      channel_title: this.args.chatChannel.escapedTitle,
-    });
-  }
-
-  get canSummarize() {
-    return this.sinceHours && !this.loading;
-  }
-
-  @action
-  summarize(value) {
-    this.sinceHours = value;
-    this.loading = true;
-    const chatChannelId = this.args.chatChannel.id;
-
-    if (this.availableSummaries[this.sinceHours]) {
-      this.summary = this.availableSummaries[this.sinceHours];
-      this.loading = false;
-      return;
-    }
-
-    ajax("/discourse-ai/summarization/chat-channel", {
-      method: "POST",
-      data: { chat_channel_id: chatChannelId, since: this.sinceHours },
-    })
-      .then((data) => {
-        this.availableSummaries[this.sinceHours] = data.summary;
-        this.summary = this.availableSummaries[this.sinceHours];
-      })
-      .catch(popupAjaxError)
-      .finally(() => (this.loading = false));
-  }
-}
--- a/assets/javascripts/discourse/templates/modal/ai-summary.hbs
+++ b/assets/javascripts/discourse/templates/modal/ai-summary.hbs
@ -0,0 +1,6 @@
+<AiSummary
+  @targetId={{this.targetId}}
+  @targetType={{this.targetType}}
+  @allowTimeframe={{this.allowTimeframe}}
+  @closeModal={{route-action "closeModal"}}
+/>
--- a/assets/javascripts/discourse/templates/modal/composer-chat-channel-summary.hbs
+++ b/assets/javascripts/discourse/templates/modal/composer-chat-channel-summary.hbs
@ -1,4 +0,0 @@
-<ChatChannelSummary
-  @chatChannel={{this.chatChannel}}
-  @closeModal={{route-action "closeModal"}}
-/>
--- a/assets/javascripts/initializers/chat-summary.js
+++ b/assets/javascripts/initializers/chat-summary.js
@ -18,8 +18,10 @@ function initializeChatChannelSummary(api) {

      @action
      showChannelSummary() {
-        showModal("composer-chat-channel-summary").setProperties({
-          chatChannel: this.chatChannel,
+        showModal("ai-summary").setProperties({
+          targetId: this.chatChannel.id,
+          targetType: "chat_channel",
+          allowTimeframe: true,
        });
      },
    });
--- a/assets/javascripts/initializers/topic-summary.js
+++ b/assets/javascripts/initializers/topic-summary.js
@ -0,0 +1,43 @@
+import { withPluginApi } from "discourse/lib/plugin-api";
+import showModal from "discourse/lib/show-modal";
+
+function initializeTopicSummary(api) {
+  api.modifyClass("component:scrolling-post-stream", {
+    showAiSummary() {
+      showModal("ai-summary").setProperties({
+        targetId: this.posts["posts"][0].topic_id,
+        targetType: "topic",
+        allowTimeframe: false,
+      });
+    },
+  });
+
+  api.addTopicSummaryCallback((html, attrs, widget) => {
+    html.push(
+      widget.attach("button", {
+        className: "btn btn-primary topic-ai-summarization",
+        icon: "magic",
+        title: "discourse_ai.summarization.title",
+        label: "discourse_ai.summarization.title",
+        action: "showAiSummary",
+      })
+    );
+
+    return html;
+  });
+}
+
+export default {
+  name: "discourse_ai-topic_summary",
+
+  initialize(container) {
+    const settings = container.lookup("site-settings:main");
+
+    const summarizationEnabled =
+      settings.discourse_ai_enabled && settings.ai_summarization_enabled;
+
+    if (summarizationEnabled) {
+      withPluginApi("1.6.0", initializeTopicSummary);
+    }
+  },
+};
--- a/assets/stylesheets/modules/summarization/common/summarization.scss
+++ b/assets/stylesheets/modules/summarization/common/summarization.scss
@ -1,4 +1,4 @@
-.composer-chat-channel-summary-modal {
+.ai-summary-modal {
  .summarization-since,
  .summary-area {
    margin: 10px 0 10px 0;
@ -8,3 +8,7 @@
    min-height: 200px;
  }
 }
+
+.topic-ai-summarization {
+  margin-left: 10px;
+}
--- a/config/locales/client.en.yml
+++ b/config/locales/client.en.yml
@ -20,7 +20,10 @@ en:
      summarization:
        title: "Summarize using AI"
        description: "Select an option below to summarize the conversation sent during the desired timeframe."
-        since: "Last %{hours} hours"
+        summarize: "Summarize"
+        since:
+          one: "Last hour" 
+          other: "Last %{count} hours"
    review:
      types:
        reviewable_ai_post:
--- a/config/routes.rb
+++ b/config/routes.rb
@ -11,7 +11,7 @@ DiscourseAi::Engine.routes.draw do
  end

  scope module: :summarization, path: "/summarization", defaults: { format: :json } do
-    post "chat-channel" => "summary#chat_channel"
+    post "summary" => "summary#show"
  end
 end

--- a/lib/modules/summarization/summary_generator.rb
+++ b/lib/modules/summarization/summary_generator.rb
@ -3,19 +3,20 @@
 module DiscourseAi
  module Summarization
    class SummaryGenerator
-      def initialize(target)
+      def initialize(target, user)
        @target = target
+        @user = user
      end

      def summarize!(content_since)
        content = get_content(content_since)

-        send("#{summarization_provider}_summarization", content)
+        send("#{summarization_provider}_summarization", content[0..(max_length - 1)])
      end

      private

-      attr_reader :target
+      attr_reader :target, :user

      def summarization_provider
        case model
@ -35,7 +36,20 @@ module DiscourseAi
        in Post
          target.raw
        in Topic
-          target.posts.order(:post_number).pluck(:raw).join("\n")
+          TopicView
+            .new(
+              target,
+              user,
+              {
+                filter: "summary",
+                exclude_deleted_users: true,
+                exclude_hidden: true,
+                show_deleted: false,
+              },
+            )
+            .posts
+            .pluck(:raw)
+            .join("\n")
        in ::Chat::Channel
          target
            .chat_messages
@ -46,7 +60,7 @@ module DiscourseAi
            .map { "#{_1}: #{_2}" }
            .join("\n")
        else
-          raise "Invalid target to classify"
+          raise "Can't find content to summarize"
        end
      end

@ -92,6 +106,19 @@ module DiscourseAi
      def model
        SiteSetting.ai_summarization_model
      end
+
+      def max_length
+        lengths = {
+          "bart-large-cnn-samsum" => 8192,
+          "flan-t5-base-samsum" => 8192,
+          "long-t5-tglobal-base-16384-book-summary" => 8192,
+          "gpt-3.5-turbo" => 8192,
+          "gpt-4" => 8192,
+          "claude-v1" => 8192,
+        }
+
+        lengths[model]
+      end
    end
  end
 end
--- a/spec/requests/summarization/summary_controller_spec.rb
+++ b/spec/requests/summarization/summary_controller_spec.rb
@ -1,7 +1,7 @@
 # frozen_string_literal: true

 RSpec.describe DiscourseAi::Summarization::SummaryController do
-  describe "#chat_channel" do
+  describe "#show" do
    fab!(:user) { Fabricate(:user) }
    let!(:channel_group) { Fabricate(:group) }
    let!(:chat_channel) { Fabricate(:private_category_channel, group: channel_group) }
@ -11,50 +11,60 @@ RSpec.describe DiscourseAi::Summarization::SummaryController do
      sign_in(user)
    end

-    context "when the user can see the channel" do
-      before { channel_group.add(user) }
+    context "when summarizing a chat channel" do
+      context "if the user can see the channel" do
+        before { channel_group.add(user) }

-      describe "validating inputs" do
-        it "returns a 404 if there is no chat channel" do
-          post "/discourse-ai/summarization/chat-channel", params: { chat_channel_id: 99, since: 3 }
+        describe "validating inputs" do
+          it "returns a 404 if there is no chat channel" do
+            post "/discourse-ai/summarization/summary",
+                 params: {
+                   target_type: "chat_channel",
+                   target_id: 99,
+                   since: 3,
+                 }

-          expect(response.status).to eq(404)
+            expect(response.status).to eq(404)
+          end
+
+          it "returns a 400 if the since param is invalid" do
+            post "/discourse-ai/summarization/summary",
+                 params: {
+                   target_type: "chat_channel",
+                   target_id: chat_channel.id,
+                   since: 0,
+                 }
+
+            expect(response.status).to eq(400)
+          end
+
+          it "returns a 404 when the module is disabled" do
+            SiteSetting.ai_summarization_enabled = false
+
+            post "/discourse-ai/summarization/summary",
+                 params: {
+                   target_type: "chat_channel",
+                   target_id: chat_channel.id,
+                   since: 1,
+                 }
+
+            expect(response.status).to eq(404)
+          end
        end

-        it "returns a 400 if the since param is invalid" do
-          post "/discourse-ai/summarization/chat-channel",
-               params: {
-                 chat_channel_id: chat_channel.id,
-                 since: 0,
-               }
+        context "if the user can't see the channel" do
+          before { channel_group.remove(user) }

-          expect(response.status).to eq(400)
-        end
+          it "returns a 403 if the user can't see the chat channel" do
+            post "/discourse-ai/summarization/summary",
+                 params: {
+                   target_type: "chat_channel",
+                   target_id: chat_channel.id,
+                   since: 1,
+                 }

-        it "returns a 404 when the module is disabled" do
-          SiteSetting.ai_summarization_enabled = false
-
-          post "/discourse-ai/summarization/chat-channel",
-               params: {
-                 chat_channel_id: chat_channel.id,
-                 since: 1,
-               }
-
-          expect(response.status).to eq(404)
-        end
-      end
-
-      context "when the user can't see the channel" do
-        before { channel_group.remove(user) }
-
-        it "returns a 403 if the user can't see the chat channel" do
-          post "/discourse-ai/summarization/chat-channel",
-               params: {
-                 chat_channel_id: chat_channel.id,
-                 since: 1,
-               }
-
-          expect(response.status).to eq(403)
+            expect(response.status).to eq(403)
+          end
        end
      end
    end
--- a/spec/support/summarization_stubs.rb
+++ b/spec/support/summarization_stubs.rb
@ -0,0 +1,56 @@
+# frozen_string_literal: true
+
+class SummarizationStubs
+  class << self
+    def test_summary
+      "This is a summary"
+    end
+
+    def openai_response(content)
+      {
+        id: "chatcmpl-6sZfAb30Rnv9Q7ufzFwvQsMpjZh8S",
+        object: "chat.completion",
+        created: 1_678_464_820,
+        model: "gpt-3.5-turbo-0301",
+        usage: {
+          prompt_tokens: 337,
+          completion_tokens: 162,
+          total_tokens: 499,
+        },
+        choices: [
+          { message: { role: "assistant", content: content }, finish_reason: "stop", index: 0 },
+        ],
+      }
+    end
+
+    def openai_chat_summarization_stub(chat_messages)
+      prompt_messages =
+        chat_messages
+          .sort_by(&:created_at)
+          .map { |m| "#{m.user.username_lower}: #{m.message}" }
+          .join("\n")
+
+      summary_prompt = [{ role: "system", content: <<~TEXT }]
+        Summarize the following article:\n\n#{prompt_messages}
+      TEXT
+
+      WebMock
+        .stub_request(:post, "https://api.openai.com/v1/chat/completions")
+        .with(body: { model: "gpt-4", messages: summary_prompt }.to_json)
+        .to_return(status: 200, body: JSON.dump(openai_response(test_summary)))
+    end
+
+    def openai_topic_summarization_stub(topic, user)
+      prompt_posts = TopicView.new(topic, user, { filter: "summary" }).posts.map(&:raw).join("\n")
+
+      summary_prompt = [{ role: "system", content: <<~TEXT }]
+        Summarize the following article:\n\n#{prompt_posts}
+      TEXT
+
+      WebMock
+        .stub_request(:post, "https://api.openai.com/v1/chat/completions")
+        .with(body: { model: "gpt-4", messages: summary_prompt }.to_json)
+        .to_return(status: 200, body: JSON.dump(openai_response(test_summary)))
+    end
+  end
+end
--- a/spec/system/page_objects/modals/summarization.rb
+++ b/spec/system/page_objects/modals/summarization.rb
@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module PageObjects
+  module Modals
+    class Summarization < PageObjects::Modals::Base
+      def visible?
+        page.has_css?(".ai-summary-modal", wait: 5)
+      end
+
+      def select_timeframe(option)
+        find(".summarization-since").click
+        find(".select-kit-row[data-value=\"#{option}\"]").click
+      end
+
+      def summary_value
+        find(".summary-area").value
+      end
+
+      def generate_summary
+        find(".ai-summary-modal .create").click
+      end
+    end
+  end
+end
--- a/spec/system/summarization/chat_channel_summarization_spec.rb
+++ b/spec/system/summarization/chat_channel_summarization_spec.rb
@ -0,0 +1,36 @@
+# frozen_string_literal: true
+
+require_relative "../../support/summarization_stubs"
+
+RSpec.describe "AI chat channel summarization", type: :system, js: true do
+  fab!(:user) { Fabricate(:leader) }
+  fab!(:channel) { Fabricate(:chat_channel) }
+
+  fab!(:message_1) { Fabricate(:chat_message, chat_channel: channel) }
+  fab!(:message_2) { Fabricate(:chat_message, chat_channel: channel) }
+  fab!(:message_3) { Fabricate(:chat_message, chat_channel: channel) }
+
+  before do
+    sign_in(user)
+    chat_system_bootstrap(user, [channel])
+    SiteSetting.ai_summarization_enabled = true
+    SiteSetting.ai_summarization_model = "gpt-4"
+  end
+
+  let(:summarization_modal) { PageObjects::Modals::Summarization.new }
+
+  it "returns a summary using the selected timeframe" do
+    visit("/chat/c/-/#{channel.id}")
+
+    SummarizationStubs.openai_chat_summarization_stub([message_1, message_2, message_3])
+
+    find(".chat-composer-dropdown__trigger-btn").click
+    find(".chat-composer-dropdown__action-btn.chat_channel_summary").click
+
+    expect(summarization_modal).to be_visible
+
+    summarization_modal.select_timeframe("3")
+
+    expect(summarization_modal.summary_value).to eq(SummarizationStubs.test_summary)
+  end
+end
--- a/spec/system/summarization/topic_summarization_spec.rb
+++ b/spec/system/summarization/topic_summarization_spec.rb
@ -0,0 +1,33 @@
+# frozen_string_literal: true
+
+require_relative "../../support/summarization_stubs"
+
+RSpec.describe "AI chat channel summarization", type: :system, js: true do
+  fab!(:user) { Fabricate(:leader) }
+  fab!(:topic) { Fabricate(:topic, has_summary: true) }
+
+  fab!(:post_1) { Fabricate(:post, topic: topic) }
+  fab!(:post_2) { Fabricate(:post, topic: topic) }
+
+  before do
+    sign_in(user)
+    SiteSetting.ai_summarization_enabled = true
+    SiteSetting.ai_summarization_model = "gpt-4"
+  end
+
+  let(:summarization_modal) { PageObjects::Modals::Summarization.new }
+
+  it "returns a summary using the selected timeframe" do
+    visit("/t/-/#{topic.id}")
+
+    SummarizationStubs.openai_topic_summarization_stub(topic, user)
+
+    find(".topic-ai-summarization").click
+
+    expect(summarization_modal).to be_visible
+
+    summarization_modal.generate_summary
+
+    expect(summarization_modal.summary_value).to eq(SummarizationStubs.test_summary)
+  end
+end