Online evaluation is a key part of Maxim’s platform, enabling you to continuously monitor and assess your AI application’s quality in production. You can run evaluations on the entire session, a single trace or particular nodes within the trace (e.g., tool calls, generations or retrievals). With online evals, you can automatically evaluate logs, set up custom filters and sampling, and gain detailed insights into your system’s real-world performance. You can combine automated evaluators with human review, curate datasets from evaluated logs, and configure alerts to stay on top of both quality and performance. This ensures your AI remains reliable and effective as it interacts with users in live environments. Online Evals