Do not store chat messages by default

We no longer want to store and show AI messages on the chat, if not
explicitly enabled by the feature. It is now only enabled for the `chat`
AI action.
We do this by setting `skip_cache = true` by default.

It also fixes a bug where the `skip_cache` was not passed along properly
to the GraphqlSubscriptionResponseService.

Changelog: fixed
EE: true

ee/app/services/llm/base_service.rb

@@ -49,7 +49,8 @@ def worker_perform(user, resource, action_name, options)
         resource_id: resource&.id,
         resource_class: resource&.class&.name,
         request_id: request_id,
-        action_name: action_name
+        action_name: action_name,
+        options: options
       )
 
       if options[:sync] == true

ee/app/services/llm/chat_service.rb

@@ -5,7 +5,7 @@ class ChatService < BaseService
     private
 
     def perform
-      worker_perform(user, resource, :chat, options)
+      worker_perform(user, resource, :chat, options.merge(skip_cache: false))
     end
 
     def valid?

ee/app/services/llm/execute_method_service.rb

@@ -27,7 +27,7 @@ def initialize(user, resource, method, options = {})
     def execute
       return error('Unknown method') unless METHODS.key?(method)
 
-      result = METHODS[method].new(user, resource, options).execute
+      result = METHODS[method].new(user, resource, options.merge(skip_cache: true)).execute
 
       track_snowplow_event(result)
       return success(result.payload) if result.success?

ee/app/workers/llm/completion_worker.rb

@@ -28,7 +28,7 @@ def perform(user_id, resource_id, resource_class, ai_action_name, options = {})
       resource = find_resource(resource_id, resource_class)
       return if resource && !user.can?("read_#{resource.to_ability_name}", resource)
 
-      params = options.extract!(:request_id, :internal_request)
+      params = options.extract!(:request_id, :internal_request, :skip_cache)
       ai_completion = ::Gitlab::Llm::CompletionsFactory.completion(ai_action_name.to_sym, params)
       logger.debug(message: "Getting Completion Service from factory", class_name: ai_completion.class.name)
 

ee/lib/gitlab/llm/completions/base.rb

@@ -14,7 +14,7 @@ def initialize(ai_prompt_class, params = {})
         attr_reader :ai_prompt_class, :params
 
         def response_options
-          params.slice(:request_id, :internal_request)
+          params.slice(:request_id, :internal_request, :skip_cache)
         end
       end
     end

ee/lib/gitlab/llm/graphql_subscription_response_service.rb

@@ -26,7 +26,8 @@ def execute
 
         logger.debug(
           message: "Broadcasting AI response",
-          data: data
+          data: data,
+          options: options
         )
 
         response_data = data.slice(:request_id, :errors, :role).merge(content: data[:response_body])

ee/lib/gitlab/llm/vertex_ai/completions/explain_vulnerability.rb

@@ -10,7 +10,7 @@ class ExplainVulnerability < Gitlab::Llm::Completions::Base
           def execute(user, vulnerability, options)
             unless vertex_ai?(vulnerability)
               return ::Gitlab::Llm::OpenAi::Completions::ExplainVulnerability
-                .new(ai_prompt_class)
+                .new(ai_prompt_class, params)
                 .execute(user, vulnerability, options)
             end
 

ee/spec/lib/gitlab/llm/vertex_ai/completions/explain_vulnerability_spec.rb

@@ -30,7 +30,11 @@
       it 'falls back to the OpenAI implementation' do
         options = {}
 
-        allow_next_instance_of(::Gitlab::Llm::OpenAi::Completions::ExplainVulnerability) do |completion|
+        expect_next_instance_of(
+          ::Gitlab::Llm::OpenAi::Completions::ExplainVulnerability,
+          prompt_class,
+          { request_id: 'uuid' }
+        ) do |completion|
           expect(completion).to receive(:execute).with(user, vulnerability, options)
         end
 

ee/spec/requests/api/graphql/mutations/projects/chat_spec.rb

@@ -33,7 +33,7 @@
     it 'successfully performs a chat request' do
       expect(Llm::CompletionWorker).to receive(:perform_async).with(
         current_user.id, nil, nil, :chat, {
-          content: "summarize", markup_format: :raw, request_id: an_instance_of(String)
+          content: "summarize", markup_format: :raw, request_id: an_instance_of(String), skip_cache: false
         }
       )
 
@@ -45,7 +45,7 @@
     it 'successfully performs a request' do
       expect(Llm::CompletionWorker).to receive(:perform_async).with(
         current_user.id, resource.id, "Issue", :chat, {
-          content: "summarize", markup_format: :raw, request_id: an_instance_of(String)
+          content: "summarize", markup_format: :raw, request_id: an_instance_of(String), skip_cache: false
         }
       )
 
@@ -61,7 +61,7 @@
     it 'successfully performs a request' do
       expect(Llm::CompletionWorker).to receive(:perform_async).with(
         current_user.id, current_user.id, "User", :chat, {
-          content: "summarize", markup_format: :raw, request_id: an_instance_of(String)
+          content: "summarize", markup_format: :raw, request_id: an_instance_of(String), skip_cache: false
         }
       )
 

ee/spec/requests/api/graphql/mutations/projects/explain_code_spec.rb

@@ -51,7 +51,7 @@
     allow(SecureRandom).to receive(:uuid).and_return(uuid)
     expect(Llm::CompletionWorker).to receive(:perform_async).with(
       current_user.id, project.id, "Project", :explain_code,
-      { markup_format: :raw, messages: messages, request_id: uuid }
+      { markup_format: :raw, messages: messages, request_id: uuid, skip_cache: true }
     )
 
     post_graphql_mutation(mutation, current_user: current_user)

ee/spec/requests/api/graphql/mutations/projects/fill_in_merge_request_template_spec.rb

@@ -47,6 +47,7 @@
       :fill_in_merge_request_template,
       {
         markup_format: :raw,
+        skip_cache: true,
         request_id: an_instance_of(String),
         source_project_id: project.id.to_s,
         source_branch: 'feature',

ee/spec/requests/api/graphql/mutations/projects/generate_commit_message_spec.rb

@@ -34,7 +34,7 @@
   it 'successfully performs an generate commit message request' do
     expect(Llm::CompletionWorker).to receive(:perform_async).with(
       current_user.id, merge_request.id, "MergeRequest", :generate_commit_message, {
-        markup_format: :raw, request_id: an_instance_of(String)
+        markup_format: :raw, request_id: an_instance_of(String), skip_cache: true
       }
     )
 

ee/spec/requests/api/graphql/mutations/projects/generate_test_file_spec.rb

@@ -35,7 +35,7 @@
   it 'successfully performs an explain code request' do
     expect(Llm::CompletionWorker).to receive(:perform_async).with(
       current_user.id, merge_request.id, "MergeRequest", :generate_test_file, {
-        file_path: file_path, markup_format: :raw, request_id: an_instance_of(String)
+        file_path: file_path, markup_format: :raw, request_id: an_instance_of(String), skip_cache: true
       }
     )
 

ee/spec/services/llm/chat_service_spec.rb

@@ -11,7 +11,7 @@
   let(:resource) { issue }
   let(:stage_check_available) { true }
   let(:content) { "Summarize issue" }
-  let(:options) { { content: content } }
+  let(:options) { { content: content, skip_cache: false } }
 
   subject { described_class.new(user, resource, options) }
 

ee/spec/services/llm/execute_method_service_spec.rb

@@ -9,7 +9,7 @@
   let(:method) { :summarize_comments }
   let(:resource) { nil }
   let(:params) { {} }
-  let(:options) { { request_id: 'uuid' }.merge(params) }
+  let(:options) { { request_id: 'uuid', skip_cache: true }.merge(params) }
 
   subject { described_class.new(user, resource, method, options).execute }
 

ee/spec/workers/llm/completion_worker_spec.rb

@@ -19,7 +19,7 @@
     let(:options) { { 'key' => 'value' } }
     let(:ai_template) { { method: :completions, prompt: 'something', options: { temperature: 0.7 } } }
     let(:ai_action_name) { :summarize_comments }
-    let(:params) { options.merge(request_id: 'uuid', internal_request: true) }
+    let(:params) { options.merge(request_id: 'uuid', internal_request: true, skip_cache: true) }
 
     subject { described_class.new.perform(user_id, resource_id, resource_type, ai_action_name, params) }
 
@@ -29,7 +29,7 @@
 
         expect(Gitlab::Llm::CompletionsFactory)
           .to receive(:completion)
-          .with(ai_action_name, match({ internal_request: true, request_id: 'uuid' }))
+          .with(ai_action_name, match({ internal_request: true, request_id: 'uuid', skip_cache: true }))
           .and_return(completion)
 
         expect(completion)