• Lightning Talks

The advent of LangChain and LlamaIndex has empowered rapid development of LLM-powered applications, yet introduced complexities in debugging due to their high-level abstractions. LLM Traces and Observability offer a solution by enabling insights into system operations without needing to understand internal mechanics, crucial for tackling unexpected issues. This requires applications to be well-instrumented, emitting necessary traces and logs for effective troubleshooting. LLM evaluation and tracing modules like Arize Phoenix exemplifies a key tool in ensuring observability, thereby facilitating smoother development and maintenance processes in LLM application development.