Multi Run Trace Aggregation And Statistics

via “multi-run trace aggregation and statistics”

We built meta-agent: an open-source library that automatically and continuously improves agent harnesses from production traces.Point it at an existing agent, a stream of unlabeled production traces, and a small labeled holdout set.An LLM judge scores unlabeled production traces as they stream.A pro

Unique: Aggregates agent-specific metrics (tool call patterns, reasoning step counts, decision distributions) rather than generic performance metrics, enabling agent-centric performance analysis

vs others: Provides agent-aware statistical analysis compared to generic time-series databases, automatically computing relevant metrics like 'tool success rate' and 'decision tree depth' without manual metric definition

via “trace filtering and aggregation by custom attributes”

** - Query and analyze your [Opik](https://github.com/comet-ml/opik) logs, traces, prompts and all other telemtry data from your LLMs in natural language.

Unique: Supports arbitrary custom attributes defined by users at trace time, rather than enforcing a fixed schema. Uses Opik's flexible metadata storage to enable ad-hoc dimensional analysis without schema migrations.

vs others: More flexible than pre-built dashboards because it supports user-defined dimensions; faster than post-processing trace exports because aggregation happens at query time in the backend