Set store=False in Responses API calls since we send full conversation history as input items and don't use previous_response_id. Add 5 run_inference tests for OpenAIResponsesLLMService using real LLMContext and adapter (only HTTP client mocked).
37 KiB
37 KiB