Rescue instrumentation failures

This wouldn't be typically needed on minimally invasive approaches of
prepending instrumentation around other frameworks, but since we're
adding some code to insert attributes, it's probably safer to ensure
instrumentation bugs wouldn't crash the user's LLM calls.

Author: clarissalimab

lib/opentelemetry/instrumentation/ruby_llm/patches/chat.rb

@@ -35,6 +35,9 @@ def ask(message, &block)
                 raise
               end
             end
+          rescue StandardError => e
+            OpenTelemetry.handle_error(exception: e)
+            super
           end
 
           def execute_tool(tool_call)
@@ -51,6 +54,9 @@ def execute_tool(tool_call)
               span.set_attribute("gen_ai.tool.call.result", result_str[0..500])
               result
             end
+          rescue StandardError => e
+            OpenTelemetry.handle_error(exception: e)
+            super
           end
 
           private

test/instrumentation_test.rb

@@ -67,6 +67,31 @@ def test_records_error_on_api_failure
     assert_equal OpenTelemetry::Trace::Status::ERROR, span.status.code
   end
 
+  def test_ask_still_works_when_instrumentation_fails
+    stub_request(:post, "https://api.openai.com/v1/chat/completions")
+      .to_return(
+        status: 200,
+        headers: { "Content-Type" => "application/json" },
+        body: {
+          id: "chatcmpl-123",
+          object: "chat.completion",
+          model: "gpt-4o-mini",
+          choices: [{
+            index: 0,
+            message: { role: "assistant", content: "Hello!" },
+            finish_reason: "stop"
+          }],
+          usage: { prompt_tokens: 10, completion_tokens: 5, total_tokens: 15 }
+        }.to_json
+      )
+
+    chat = RubyLLM.chat(model: "gpt-4o-mini")
+    chat.define_singleton_method(:tracer) { raise StandardError, "instrumentation bug" }
+
+    response = chat.ask("Hi")
+    assert_equal "Hello!", response.content
+  end
+
   def test_creates_span_for_tool_call
     calculator = Class.new(RubyLLM::Tool) do
       def self.name = "calculator"
@@ -141,4 +166,66 @@ def execute(expression:)
     assert_equal "call_abc123", tool_span.attributes["gen_ai.tool.call.id"]
     assert_equal "function", tool_span.attributes["gen_ai.tool.type"]
   end
+
+  def test_execute_tool_still_works_when_instrumentation_fails
+    calculator = Class.new(RubyLLM::Tool) do
+      def self.name = "calculator"
+      description "Performs math"
+      param :expression, type: "string", desc: "Math expression"
+
+      def execute(expression:)
+        eval(expression).to_s
+      end
+    end
+
+    stub_request(:post, "https://api.openai.com/v1/chat/completions")
+      .to_return(
+        {
+          status: 200,
+          headers: { "Content-Type" => "application/json" },
+          body: {
+            id: "chatcmpl-123",
+            object: "chat.completion",
+            model: "gpt-4o-mini",
+            choices: [{
+              index: 0,
+              message: {
+                role: "assistant",
+                content: nil,
+                tool_calls: [{
+                  id: "call_abc123",
+                  type: "function",
+                  function: { name: "calculator", arguments: '{"expression":"2+2"}' }
+                }]
+              },
+              finish_reason: "tool_calls"
+            }],
+            usage: { prompt_tokens: 10, completion_tokens: 5, total_tokens: 15 }
+          }.to_json
+        },
+        {
+          status: 200,
+          headers: { "Content-Type" => "application/json" },
+          body: {
+            id: "chatcmpl-456",
+            object: "chat.completion",
+            model: "gpt-4o-mini",
+            choices: [{
+              index: 0,
+              message: { role: "assistant", content: "The answer is 4" },
+              finish_reason: "stop"
+            }],
+            usage: { prompt_tokens: 20, completion_tokens: 5, total_tokens: 25 }
+          }.to_json
+        }
+      )
+
+    chat = RubyLLM.chat(model: "gpt-4o-mini")
+    chat.with_tool(calculator)
+
+    chat.define_singleton_method(:tracer) { raise StandardError, "instrumentation bug" }
+
+    response = chat.ask("What is 2+2?")
+    assert_equal "The answer is 4", response.content
+  end
 end