Worked on the All-Hands-AI/agent-sdk repository to enhance observability and traceability in LLM-driven event workflows. Developed a feature that introduced a shared llm_response_id across all LLM-generated events, enabling grouping of related actions and tracing results from parallel function calls originating from the same LLM response. This improvement extended event models within the OpenHands SDK, facilitating more reliable orchestration and faster issue diagnosis in LLM pipelines. Leveraged Python for backend and API development, applying event-driven architecture principles to improve event tracking and analytics. The work focused on technical depth, addressing debugging and analytics needs without introducing new bug fixes.