domino-genai-tracing

Trace GenAI executions in Domino by wrapping LLM calls, agents, and RAG pipelines in a DominoRun with the @add_tracing decorator to capture token usage, latency, cost, and errors for audit-ready records.

Tracing LLM calls captures token usage, latency, cost, tool usage, and errors across multi-step AI workflows, providing audit-ready data to support compliance, debugging, and performance optimization.

Domino GenAI tracing integrates with MLflow 3.2.0 to score and aggregate trace metrics, enabling robust experiment tracking and custom evaluators for comprehensive performance evaluation.

You can use custom evaluators with MLflow integration to score and aggregate trace metrics for RAG pipelines, enabling detailed performance evaluation and robust experimentation within Domino.

Gain end-to-end visibility into RAG pipelines and agent workflows by using the Domino SDK tracing decorators to capture comprehensive execution data including token counts, tool usage, and errors.

You need the Domino SDK to trace GenAI applications, using the @add_tracing decorator and DominoRun class to capture execution metrics and integrate with MLflow for evaluation.