AI Observability in Snowflake Cortex (General availability) With this release, we are pleased to announce the general availability of AI Observability in Snowflake Cortex, which was previously available as a preview feature. AI Observability enables you to evaluate and trace your generative AI applications, making them more trustworthy and transparent.
AI Observability allows you to systematically measure the performance of your AI applications by running evaluations, logging application traces for debugging, and benchmarking performance for production deployments. Key features include:
Evaluations: Systematically assess generative AI applications and agents using the LLM-as-a-judge technique, leveraging metrics such as accuracy, latency, usage, and cost.
Comparison: Compare multiple evaluations side by side to identify the best configuration for production.
Tracing: Trace every step of application executions to debug and refine your applications.
AI Observability supports a variety of task types, such as retrieval-augmented generation (RAG) and summarization, and provides detailed metrics to help you optimize your applications.
For more information, see AI Observability in Snowflake Cortex.