CI

Example CI command:

uv run dialogassert run \
  -c dialogassert.openai.business-acceptance.yaml \
  --semantic-mode llm

For non-deterministic agents, configure repeat and threshold:

evaluation:
  repeat: 3
  pass_threshold: 0.75

In CI, keep --no-fail-on-findings off so high and critical findings fail the job. The flag is for exploration and demos.

Write reports to explicit paths so the job can archive them:

uv run dialogassert run \
  -c dialogassert.openai.business-acceptance.yaml \
  --output reports/my-run.json \
  --html reports/my-run.html