0.1.8

What's Changed

• Feature/judge config registry by @kelkalot in #11
• Add custom probe/judge prompts, json_format flag, and visualizer support by @kelkalot in #10
• Dev by @SushantGautam in #12
• Add Hei refusal pack + test_prompt honoured on turn 1 by @kelkalot in #13
• Fix Anthropic judge default path: json_schema + any-llm-sdk floor bump by @avalyset in #16
• Add nav_aap scenario pack: 15 scenarios on Norwegian welfare administration by @avalyset in #17
• Add skatteetaten scenario pack: 8 scenarios on Norwegian tax administration by @avalyset in #18
• Declarative judge response_schema + helsedir/binary_abstention judges by @kelkalot in #19
• repeated-run stability analysis, token tracking, resumable experiments, and auditor model separation by @finnschwall in #20
• Add FAQ section addressing openai.PermissionDeniedError for custom LL… by @SushantGautam in #21
• feat: CrossJudgeExperiment for cross-version judge stability analysis by @avalyset in #24
• Fix audit-framework bugs and add regression tests by @kelkalot in #25

New Contributors

• @kelkalot made their first contribution in #11
• @avalyset made their first contribution in #16
• @finnschwall made their first contribution in #20

Full Changelog: 0.1.7...0.1.8

0.1.7a

Full Changelog: 0.1.6...0.1.7a

0.1.6

Full Changelog: 0.1.4...0.1.6