fix: preserve Responses streaming error details

CHANGELOG.md

@@ -39,6 +39,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **Responses streaming failures retain their error details** (`activeagent`).
+  `error` and `response.failed` events raise a provider error with the API
+  code and message instead of an unexpected chunk type error.
 - **The dashboard keeps a catalog's pass bars** (`actionagent`). A
   product's and a set's `pass` blocks (`min_pass_rate`, `threshold`,
   `gated`) parsed in `ActiveAgent::Evals::Catalog`, but neither

lib/active_agent/providers/open_ai/responses_provider.rb

@@ -110,6 +110,7 @@ def api_response_normalize(api_response)
         #   reasoning progress, kept from the reasoning item once it is done
         # - `:"response.content_part.done"` - content part completed
         # - `:"response.output_item.done"` - message, reasoning or function call completed
+        # - `:error`, `:"response.failed"` - provider failure
         # - `:"response.completed"` - response finished
         # - `:"response.incomplete"` - response cut short (e.g. max_output_tokens)
         #
@@ -174,6 +175,12 @@ def process_stream_chunk(api_response_event)
           when :"response.output_item.done"
             process_stream_output_item_done(api_response_event)
 
+          # Errors arrive as events even though the HTTP request succeeded.
+          when :error, :"response.failed"
+            error = api_response_event.type == :error ? api_response_event : api_response_event.response.error
+            message = [ error&.code, error&.message ].compact.join(": ")
+            raise Errors::ProviderError.new(message.presence || "OpenAI response failed", provider_tag: tag_name)
+
           # Response Completed, or cut short (e.g. by max_output_tokens) with
           # the output it has so far
           when :"response.completed", :"response.incomplete"

test/providers/open_ai/responses/streaming_errors_test.rb

@@ -0,0 +1,53 @@
+# frozen_string_literal: true
+
+require "test_helper"
+require_relative "../../../../lib/active_agent/providers/open_ai/responses_provider"
+
+module Providers
+  module OpenAI
+    module Responses
+      class StreamingErrorsTest < ActiveSupport::TestCase
+        include WebMock::API
+
+        test "an error event preserves the API error message and code" do
+          assert_stream_error({ type: "error", code: "server_error", message: "Please try again", param: nil },
+            "server_error: Please try again")
+        end
+
+        test "a failed response preserves its nested error" do
+          assert_stream_error({ type: "response.failed", response: {
+            id: "resp_failed", status: "failed", error: { code: "server_error", message: "Generation failed" }
+          } }, "server_error: Generation failed")
+        end
+
+        test "a failed response without error details still raises a provider error" do
+          assert_stream_error({ type: "response.failed", response: {
+            id: "resp_failed", status: "failed", error: nil
+          } }, "OpenAI response failed")
+        end
+
+        private
+
+        def assert_stream_error(event, message)
+          events = [ { type: "response.created", response: {
+            id: "resp_failed", status: "in_progress", output: []
+          } }, event ]
+          body = events.each_with_index.map do |chunk, index|
+            "event: #{chunk[:type]}\ndata: #{chunk.merge(sequence_number: index).to_json}\n\n"
+          end.join
+          stub_request(:post, "https://api.openai.com/v1/responses").to_return(
+            status: 200, headers: { "Content-Type" => "text/event-stream" }, body:
+          )
+          provider = ActiveAgent::Providers::OpenAI::ResponsesProvider.new(
+            service: "OpenAI", api_key: "test-key", model: "gpt-5", stream: true,
+            messages: [ { role: "user", content: "Hello" } ],
+            stream_broadcaster: ->(_message, _delta, _event_type) { }
+          )
+
+          error = assert_raises(ActiveAgent::Providers::Errors::ProviderError) { provider.prompt }
+          assert_equal message, error.message
+        end
+      end
+    end
+  end
+end