Add additional bedrock stub to stub all answer relevancy calls

We're going to need to stub out these calls in multiple places so it makes
sense to have a single method that does all the stubbing for us.

I've also prepended stub_ to bedrock_invoke_model_openai_oss_tool_call.
All other stubs have this so it makes sense to be consistent.

Author: davidgisbey

spec/lib/auto_evaluation/answer_relevancy/reason_generator_spec.rb

@@ -23,7 +23,7 @@
     end
     let(:tools) { [prompts.fetch(:tool_spec)] }
     let!(:stub_bedrock) do
-      bedrock_invoke_model_openai_oss_tool_call(
+      stub_bedrock_invoke_model_openai_oss_tool_call(
         user_prompt,
         tools,
         reason_json,

spec/lib/auto_evaluation/answer_relevancy/statement_generator_spec.rb

@@ -14,7 +14,7 @@
     end
     let(:tools) { [prompts.fetch(:tool_spec)] }
     let!(:stub_bedrock) do
-      bedrock_invoke_model_openai_oss_tool_call(
+      stub_bedrock_invoke_model_openai_oss_tool_call(
         user_prompt,
         tools,
         statements_json,

spec/lib/auto_evaluation/answer_relevancy/verdicts_generator_spec.rb

@@ -21,7 +21,7 @@
     end
     let(:tools) { [prompts.fetch(:tool_spec)] }
     let!(:stub_bedrock) do
-      bedrock_invoke_model_openai_oss_tool_call(
+      stub_bedrock_invoke_model_openai_oss_tool_call(
         user_prompt,
         tools,
         verdicts_json,

spec/lib/auto_evaluation/answer_relevancy_spec.rb

@@ -1,67 +1,30 @@
 RSpec.describe AutoEvaluation::AnswerRelevancy, :aws_credentials_stubbed do
   describe ".call" do
-    let(:prompts) { AutoEvaluation::Prompts.config.answer_relevancy }
     let(:question_message) { "This is a test question message." }
     let(:answer_message) { "This is a test answer message." }
-
     let(:statements) { ["This is the first statement.", "This is the second statement."] }
     let(:statements_json) { { statements: }.to_json }
-    let(:user_prompt_statements) do
-      sprintf(
-        prompts.fetch(:statements).fetch(:user_prompt),
-        answer: answer_message,
-      )
-    end
-    let(:statements_tools) { [prompts.fetch(:statements).fetch(:tool_spec)] }
-    let!(:statements_stub) do
-      bedrock_invoke_model_openai_oss_tool_call(
-        user_prompt_statements,
-        statements_tools,
-        statements_json,
-      )
-    end
-
     let(:verdicts) do
       [
         { "verdict" => "Yes" },
         { "verdict" => "No", "reason" => "The statement is irrelevant." },
       ]
     end
     let(:verdicts_json) { { verdicts: }.to_json }
-    let(:user_prompt_verdicts) do
-      sprintf(
-        prompts.fetch(:verdicts).fetch(:user_prompt),
-        question: question_message,
-        statements:,
-      )
-    end
-    let(:verdicts_tools) { [prompts.fetch(:verdicts).fetch(:tool_spec)] }
-    let!(:verdicts_stub) do
-      bedrock_invoke_model_openai_oss_tool_call(
-        user_prompt_verdicts,
-        verdicts_tools,
-        verdicts_json,
-      )
-    end
-
     let(:reason) { "This is the reason for the score." }
     let(:reason_json) { { reason: }.to_json }
-    let(:user_prompt_reason) do
-      sprintf(
-        prompts.fetch(:reason).fetch(:user_prompt),
-        score: 0.5,
-        unsuccessful_verdicts_reasons: ["The statement is irrelevant."],
-        question: question_message,
-      )
-    end
-    let(:reason_tools) { [prompts.fetch(:reason).fetch(:tool_spec)] }
-    let!(:reason_stub) do
-      bedrock_invoke_model_openai_oss_tool_call(
-        user_prompt_reason,
-        reason_tools,
-        reason_json,
+    let!(:answer_relevancy_stubs) do
+      stub_bedrock_invoke_model_openai_oss_answer_relevancy(
+        question_message:,
+        answer_message:,
+        statements_json:,
+        verdicts_json:,
+        reason_json:,
       )
     end
+    let(:statements_stub) { answer_relevancy_stubs[:statements] }
+    let(:verdicts_stub) { answer_relevancy_stubs[:verdicts] }
+    let(:reason_stub) { answer_relevancy_stubs[:reason] }
 
     it "returns a results object with the expected attributes" do
       allow(Clock).to receive(:monotonic_time)

spec/lib/auto_evaluation/bedrock_openai_oss_invoke_spec.rb

@@ -21,7 +21,7 @@
       ]
     end
     let!(:stub) do
-      bedrock_invoke_model_openai_oss_tool_call(
+      stub_bedrock_invoke_model_openai_oss_tool_call(
         user_message,
         tools,
         { "response" => "Expected response." }.to_json,

spec/support/stub_bedrock.rb

@@ -47,7 +47,7 @@ def mock_titan_embedding(text, dimensions: Search::ChunkedContentRepository::TIT
     dimensions.times.map { random_generator.rand }
   end
 
-  def bedrock_invoke_model_openai_oss_tool_call(user_message, tools, content)
+  def stub_bedrock_invoke_model_openai_oss_tool_call(user_message, tools, content)
     request_body = {
       include_reasoning: false,
       messages: [
@@ -85,4 +85,53 @@ def bedrock_invoke_model_openai_oss_tool_call(user_message, tools, content)
       endpoint_regex: OPENAI_GPT_OSS_ENDPOINT_REGEX,
     )
   end
+
+  def stub_bedrock_invoke_model_openai_oss_answer_relevancy(question_message:,
+                                                            answer_message:,
+                                                            statements_json: { statements: ["Statement."] }.to_json,
+                                                            verdicts_json: { verdicts: [{ "verdict" => "Yes" }] }.to_json,
+                                                            reason_json: { reason: "This is the reason for the score." }.to_json)
+    prompts = AutoEvaluation::Prompts.config.answer_relevancy
+
+    statements_user_prompt = sprintf(
+      prompts.fetch(:statements).fetch(:user_prompt),
+      answer: answer_message,
+    )
+    verdicts_user_prompt = sprintf(
+      prompts.fetch(:verdicts).fetch(:user_prompt),
+      question: question_message,
+      statements: JSON.parse(statements_json).fetch("statements"),
+    )
+    reason_user_prompt = sprintf(
+      prompts.fetch(:reason).fetch(:user_prompt),
+      score: 0.5,
+      unsuccessful_verdicts_reasons: ["The statement is irrelevant."],
+      question: question_message,
+    )
+
+    statements_tools = [prompts.fetch(:statements).fetch(:tool_spec)]
+    verdicts_tools = [prompts.fetch(:verdicts).fetch(:tool_spec)]
+    reason_tools = [prompts.fetch(:reason).fetch(:tool_spec)]
+
+    stubs = {}
+    stubs[:statements] = stub_bedrock_invoke_model_openai_oss_tool_call(
+      statements_user_prompt,
+      statements_tools,
+      statements_json,
+    )
+
+    stubs[:verdicts] = stub_bedrock_invoke_model_openai_oss_tool_call(
+      verdicts_user_prompt,
+      verdicts_tools,
+      verdicts_json,
+    )
+
+    stubs[:reason] = stub_bedrock_invoke_model_openai_oss_tool_call(
+      reason_user_prompt,
+      reason_tools,
+      reason_json,
+    )
+
+    stubs
+  end
 end