From b0310f90d35cdac4177737dc1c2e04a60da919d2 Mon Sep 17 00:00:00 2001
From: Sam <sam.saffron@gmail.com>
Date: Tue, 12 Sep 2023 11:06:55 +1000
Subject: [PATCH] FEATURE: add tags and categories to read context (#215)

Note, we perform permission checks on tag list against anon
to ensure we do not disclose information about private tags
to the llm which could get extracted.
---
 lib/modules/ai_bot/commands/read_command.rb   | 14 ++++++-
 .../ai_bot/commands/read_command_spec.rb      | 41 ++++++++++++++++---
 2 files changed, 48 insertions(+), 7 deletions(-)

diff --git a/lib/modules/ai_bot/commands/read_command.rb b/lib/modules/ai_bot/commands/read_command.rb
index 22adbe26..b8025b51 100644
--- a/lib/modules/ai_bot/commands/read_command.rb
+++ b/lib/modules/ai_bot/commands/read_command.rb
@@ -50,7 +50,19 @@ module DiscourseAi::AiBot::Commands
 
       posts = posts.where("post_number = ?", post_number) if post_number
 
-      content = +"title: #{topic.title}\n\n"
+      content = +<<~TEXT.strip
+        title: #{topic.title}
+      TEXT
+
+      category_names = [topic.category&.parent_category&.name, topic.category&.name].compact.join(
+        " ",
+      )
+      content << "\ncategories: #{category_names}" if category_names.present?
+
+      if topic.tags.length > 0
+        tags = DiscourseTagging.filter_visible(topic.tags, Guardian.new)
+        content << "\ntags: #{tags.map(&:name).join(", ")}\n\n" if tags.length > 0
+      end
 
       posts.each { |post| content << "\n\n#{post.username} said:\n\n#{post.raw}" }
 
diff --git a/spec/lib/modules/ai_bot/commands/read_command_spec.rb b/spec/lib/modules/ai_bot/commands/read_command_spec.rb
index 992c29f8..47cc4661 100644
--- a/spec/lib/modules/ai_bot/commands/read_command_spec.rb
+++ b/spec/lib/modules/ai_bot/commands/read_command_spec.rb
@@ -3,19 +3,48 @@
 RSpec.describe DiscourseAi::AiBot::Commands::ReadCommand do
   fab!(:bot_user) { User.find(DiscourseAi::AiBot::EntryPoint::GPT3_5_TURBO_ID) }
 
+  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) { Fabricate(:tag_group, name: "Staff only", tag_names: ["hidden"]) }
+  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
-      post1 = Fabricate(:post, raw: "hello there")
-      Fabricate(:post, raw: "mister sam", topic: post1.topic)
+      topic_id = topic_with_tags.id
 
-      read = described_class.new(bot_user: bot_user, args: nil, post: post1)
+      Fabricate(:post, topic: topic_with_tags, raw: "hello there")
+      Fabricate(:post, topic: topic_with_tags, raw: "mister sam")
 
-      results = read.process(topic_id: post1.topic_id)
+      read = described_class.new(bot_user: bot_user, args: nil)
 
-      expect(results[:topic_id]).to eq(post1.topic_id)
+      results = read.process(topic_id: topic_id)
+
+      expect(results[:topic_id]).to eq(topic_id)
       expect(results[:content]).to include("hello")
       expect(results[:content]).to include("sam")
-      expect(read.description_args).to eq(title: post1.topic.title, url: post1.topic.relative_url)
+      expect(results[:content]).to include("amazing-cat")
+      expect(results[:content]).to include("funny")
+      expect(results[:content]).to include("sad")
+      expect(results[:content]).to include("animals")
+      expect(results[:content]).not_to include("hidden")
+      expect(read.description_args).to eq(
+        title: topic_with_tags.title,
+        url: topic_with_tags.relative_url,
+      )
     end
   end
 end