FEATURE: include tag and category context in search (#217)

SamSaffron · web-flow · commit d75e3ca82b9b · 2023-09-12T16:09:28.000+10:00
Previous to this we just included title/body.. tags and category
structure can be very critical for decision making.
diff --git a/lib/modules/ai_bot/commands/search_command.rb b/lib/modules/ai_bot/commands/search_command.rb
@@ -154,17 +154,33 @@ def process(**search_args)
       end
 
       @last_num_results = posts.length
+      # this is the general pattern from core
+      # if there are millions of hidden tags it may fail
+      hidden_tags = nil
 
       if posts.blank?
         { args: search_args, rows: [], instruction: "nothing was found, expand your search" }
       else
         format_results(posts, args: search_args) do |post|
-          {
+          category_names = [
+            post.topic.category&.parent_category&.name,
+            post.topic.category&.name,
+          ].compact.join(" > ")
+          row = {
             title: post.topic.title,
             url: Discourse.base_path + post.url,
             excerpt: post.excerpt,
             created: post.created_at,
+            category: category_names,
           }
+
+          if SiteSetting.tagging_enabled
+            hidden_tags ||= DiscourseTagging.hidden_tag_names
+            # using map over pluck to avoid n+1 (assuming caller preloading)
+            tags = post.topic.tags.map(&:name) - hidden_tags
+            row[:tags] = tags.join(", ") if tags.present?
+          end
+          row
         end
       end
     end
diff --git a/spec/lib/modules/ai_bot/commands/read_command_spec.rb b/spec/lib/modules/ai_bot/commands/read_command_spec.rb
@@ -9,19 +9,19 @@
   fab!(:tag_funny) { Fabricate(:tag, name: "funny") }
   fab!(:tag_sad) { Fabricate(:tag, name: "sad") }
   fab!(:tag_hidden) { Fabricate(:tag, name: "hidden") }
-  fab!(:staff_tag_group) { Fabricate(:tag_group, name: "Staff only", tag_names: ["hidden"]) }
+  fab!(:staff_tag_group) do
+    tag_group = Fabricate.build(:tag_group, name: "Staff only", tag_names: ["hidden"])
+
+    tag_group.permissions = [
+      [Group::AUTO_GROUPS[:staff], TagGroupPermission.permission_types[:full]],
+    ]
+    tag_group.save!
+    tag_group
+  end
   fab!(:topic_with_tags) do
     Fabricate(:topic, category: category, tags: [tag_funny, tag_sad, tag_hidden])
   end
 
-  let(:staff) { Group::AUTO_GROUPS[:staff] }
-  let(:full) { TagGroupPermission.permission_types[:full] }
-
-  before do
-    staff_tag_group.permissions = [[staff, full]]
-    staff_tag_group.save!
-  end
-
   describe "#process" do
     it "can read a topic" do
       topic_id = topic_with_tags.id
diff --git a/spec/lib/modules/ai_bot/commands/search_command_spec.rb b/spec/lib/modules/ai_bot/commands/search_command_spec.rb
@@ -9,9 +9,28 @@
   before { SearchIndexer.enable }
   after { SearchIndexer.disable }
 
+  fab!(:parent_category) { Fabricate(:category, name: "animals") }
+  fab!(:category) { Fabricate(:category, parent_category: parent_category, name: "amazing-cat") }
+
+  fab!(:tag_funny) { Fabricate(:tag, name: "funny") }
+  fab!(:tag_sad) { Fabricate(:tag, name: "sad") }
+  fab!(:tag_hidden) { Fabricate(:tag, name: "hidden") }
+  fab!(:staff_tag_group) do
+    tag_group = Fabricate.build(:tag_group, name: "Staff only", tag_names: ["hidden"])
+
+    tag_group.permissions = [
+      [Group::AUTO_GROUPS[:staff], TagGroupPermission.permission_types[:full]],
+    ]
+    tag_group.save!
+    tag_group
+  end
+  fab!(:topic_with_tags) do
+    Fabricate(:topic, category: category, tags: [tag_funny, tag_sad, tag_hidden])
+  end
+
   describe "#process" do
     it "can handle no results" do
-      post1 = Fabricate(:post)
+      post1 = Fabricate(:post, topic: topic_with_tags)
       search = described_class.new(bot_user: bot_user, post: post1, args: nil)
 
       results = search.process(query: "order:fake ABDDCDCEDGDG")
@@ -42,7 +61,7 @@
           hyde_embedding,
         )
 
-        post1 = Fabricate(:post)
+        post1 = Fabricate(:post, topic: topic_with_tags)
         search = described_class.new(bot_user: bot_user, post: post1, args: nil)
 
         DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2
@@ -60,16 +79,30 @@
     it "supports subfolder properly" do
       Discourse.stubs(:base_path).returns("/subfolder")
 
-      post1 = Fabricate(:post)
+      post1 = Fabricate(:post, topic: topic_with_tags)
 
       search = described_class.new(bot_user: bot_user, post: post1, args: nil)
 
       results = search.process(limit: 1, user: post1.user.username)
       expect(results[:rows].to_s).to include("/subfolder" + post1.url)
     end
 
+    it "returns category and tags" do
+      post1 = Fabricate(:post, topic: topic_with_tags)
+      search = described_class.new(bot_user: bot_user, post: post1, args: nil)
+      results = search.process(user: post1.user.username)
+
+      row = results[:rows].first
+      category = row[results[:column_names].index("category")]
+
+      expect(category).to eq("animals > amazing-cat")
+
+      tags = row[results[:column_names].index("tags")]
+      expect(tags).to eq("funny, sad")
+    end
+
     it "can handle limits" do
-      post1 = Fabricate(:post)
+      post1 = Fabricate(:post, topic: topic_with_tags)
       _post2 = Fabricate(:post, user: post1.user)
       _post3 = Fabricate(:post, user: post1.user)
 
@@ -78,7 +111,6 @@
 
       results = search.process(limit: 2, user: post1.user.username)
 
-      expect(results[:column_names].length).to eq(4)
       expect(results[:rows].length).to eq(2)
 
       # just searching for everything
