FEATURE: experimental read command for bot (#129)

This command is useful for reading a topics content. It allows us to perform
critical analysis or suggest answers.

Given 8k token limit in GPT-4 I hardcoded reading to 1500 tokens, but we can
follow up and allow larger windows on models that support more tokens.

On local testing even in this limited form this can be very useful.

config/locales/server.en.yml

@@ -100,7 +100,9 @@ en:
         summarize: "Summarize"
         image: "Generate image"
         google: "Search Google"
+        read: "Read topic"
       command_description:
+        read: "Reading: %{title}"
         time: "Time in %{timezone} is %{time}"
         summarize: "Summarized <a href='%{url}'>%{title}</a>"
         image: "%{prompt}"

config/settings.yml

@@ -203,6 +203,7 @@ plugins:
      - image
      - search
      - summarize
+     - read
      - tags
      - time
   ai_helper_add_ai_pm_to_header:

lib/modules/ai_bot/commands/read_command.rb

@@ -0,0 +1,61 @@
+#frozen_string_literal: true
+
+module DiscourseAi::AiBot::Commands
+  class ReadCommand < Command
+    class << self
+      def name
+        "read"
+      end
+
+      def desc
+        "Will read a topic or a post on this Discourse instance"
+      end
+
+      def parameters
+        [
+          Parameter.new(
+            name: "topic_id",
+            description: "the id of the topic to read",
+            type: "integer",
+            required: true,
+          ),
+          Parameter.new(
+            name: "post_number",
+            description: "the post number to read",
+            type: "integer",
+            required: false,
+          ),
+        ]
+      end
+    end
+
+    def description_args
+      { title: @title }
+    end
+
+    def process(topic_id:, post_number: nil)
+      not_found = { topic_id: topic_id, description: "Topic not found" }
+
+      @title = ""
+
+      topic_id = topic_id.to_i
+
+      topic = Topic.find_by(id: topic_id)
+      return not_found if !topic || !Guardian.new.can_see?(topic)
+
+      @title = topic.title
+
+      posts = Post.secured(Guardian.new).where(topic_id: topic_id).order(:post_number).limit(40)
+      posts = posts.where("post_number = ?", post_number) if post_number
+
+      content = +"title: #{topic.title}\n\n"
+
+      posts.each { |post| content << "\n\n#{post.username} said:\n\n#{post.raw}" }
+
+      # TODO: 16k or 100k models can handle a lot more tokens
+      content = ::DiscourseAi::Tokenizer::BertTokenizer.truncate(content, 1500).squish
+
+      { topic_id: topic_id, post_number: post_number, content: content }
+    end
+  end
+end

lib/modules/ai_bot/entry_point.rb

@@ -35,6 +35,7 @@ module DiscourseAi
         require_relative "commands/summarize_command"
         require_relative "commands/image_command"
         require_relative "commands/google_command"
+        require_relative "commands/read_command"
       end
 
       def inject_into(plugin)

lib/modules/ai_bot/open_ai_bot.rb

@@ -95,6 +95,7 @@ module DiscourseAi
             Commands::TimeCommand,
             Commands::SearchCommand,
             Commands::SummarizeCommand,
+            Commands::ReadCommand,
           ].tap do |cmds|
             cmds << Commands::TagsCommand if SiteSetting.tagging_enabled
             cmds << Commands::ImageCommand if SiteSetting.ai_stability_api_key.present?

spec/lib/modules/ai_bot/commands/read_command_spec.rb

@@ -0,0 +1,20 @@
+#frozen_string_literal: true
+
+RSpec.describe DiscourseAi::AiBot::Commands::ReadCommand do
+  fab!(:bot_user) { User.find(DiscourseAi::AiBot::EntryPoint::GPT3_5_TURBO_ID) }
+
+  describe "#process" do
+    it "can read a topic" do
+      post1 = Fabricate(:post, raw: "hello there")
+      Fabricate(:post, raw: "mister sam", topic: post1.topic)
+
+      read = described_class.new(bot_user, post1)
+
+      results = read.process(topic_id: post1.topic_id)
+
+      expect(results[:topic_id]).to eq(post1.topic_id)
+      expect(results[:content]).to include("hello")
+      expect(results[:content]).to include("sam")
+    end
+  end
+end