In November 2025, Pablo García enhanced observability in the BerriAI/litellm repository by developing a cumulative tool call index for Vertex AI integration. He implemented this feature using Python, focusing on backend development and AI integration skills. The solution required updating method signatures and internal logic to propagate the index across processing steps, enabling comprehensive tracking of tool usage throughout Vertex AI workflows. This addition improved end-to-end visibility and facilitated debugging and cost attribution. Pablo’s work demonstrated a thoughtful approach to extending analytics and governance capabilities, addressing both immediate operational needs and laying groundwork for future enhancements in AI pipeline monitoring.