All issues

A node that is both a declared root and a child of another root runs multiple times, and KeyErrors in async

#3026

· jhochenbaum opened

on Aug 10, 2026

DAGMetric score depends on node declaration order and on sync vs async when two verdicts can score

#3025

· jhochenbaum opened

on Aug 10, 2026

Community metric proposal: localize the divergence point between two agent execution traces (not just pass/fail)

#3019

· Melvin0070 opened

on Aug 9, 2026

AzureOpenAIModel defaults temperature to 0.0, causing 400 errors with GPT-5 models

#3015

· tm-ramon-jasmin opened

on Aug 7, 2026

Confusion in Tool Correctness metric

#3014

· IamGodseye opened

on Aug 7, 2026

Request: enable Private Vulnerability Reporting (or share a security contact)

#3008

· GeeksikhSecurity opened

on Aug 6, 2026

`instrument_google_adk()` raises `CONFIDENT_API_KEY is not set` despite docs stating tracing works fully offline

#3005

· itsatefe opened

on Aug 5, 2026

GEval logprobs fails if reason contains score token.

#3000

· volksen opened

on Aug 4, 2026

PromptOptimizer's default GEPA() is a shared instance - a second optimizer silently hijacks the first's model_callback and metrics

#2998

· shalinis97 opened

on Aug 3, 2026

Default AnthropicModel uses a nonexistent model ID (claude-sonnet-4-6-20250514) — every default-configured evaluation 404s

#2993

· lokesh75-kank opened

on Aug 2, 2026

deepeval doesn't print any progress bar or detailed test result in the console

#2986

· driptaroop opened

on Jul 29, 2026

DeepEvalBaseLLM.load_model return annotation contradicts its docstring and the built-in models

#2985

· thnbi opened

on Jul 29, 2026