Skip to content

Agentic Evaluator #59

@cemde

Description

@cemde

An Evaluator class that reads through the traces autonomously would be great. Automated failure diagnostics are becoming more common. Equipping an evaluator with tools to explore traces on its own could be very useful.

E.g. expose a Traces tool to the evalutor with methods such as get_models() etc would be useful. Basically, almost a json-like interface.

Metadata

Metadata

Assignees

No one assigned

    Labels

    coreIn regards to the core package `maseval/core`enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions