feat(core): parse Promptfoo llm-rubric results

[christso]

Summary

• Accept Promptfoo-style llm-rubric judge responses shaped as { reason, pass, score } and normalize them into AgentV grader scores.
• Map optional Promptfoo-style checks[] into current internal assertion entries while preserving normalized pass/score/reason in details for the downstream artifact contract work.
• Let llm-rubric custom prompts receive the authored value through {{ rubric }} and use a Promptfoo-compatible output schema for free-form llm-rubric judging.

Validation

• bun test packages/core/test/evaluation/graders.test.ts
• bunx biome check packages/core/src/evaluation/graders/llm-grader.ts packages/core/src/evaluation/registry/builtin-graders.ts packages/core/test/evaluation/graders.test.ts
• bun --filter @agentv/core typecheck
• bun --filter @agentv/core build
• bun run build (passed before the final prompt-contract adjustment; final core build passed after)

Live Dogfood

• Command: LOCAL_OPENAI_PROXY_BASE_URL=http://127.0.0.1:10531/v1 LOCAL_OPENAI_PROXY_API_KEY=dummy-local-key LOCAL_OPENAI_PROXY_MODEL=gpt-5.4-mini bun apps/cli/src/cli.ts eval run .agentv/results/av-kfik-28-3-dogfood/eval.yaml --targets .agentv/results/av-kfik-28-3-dogfood/targets.yaml --target local-openai --grader-target local-openai --workers 1 --threshold 0.8 --output .agentv/results/av-kfik-28-3-dogfood-final
• Result: PASS, 1/1 scored >= 80%, mean 100%.
• Private evidence branch: EntityProcess/agentv-private:evidence/av-kfik-28-3-llm-rubric-parsing at commit f961c63.

Post-Deploy Monitoring & Validation

• No additional production monitoring required. This changes local/CI grader parsing behavior and is covered by focused parser tests plus live grader dogfood evidence.

Notes

• This intentionally does not rewrite public run artifacts from assertion_results to the final graders/checks shape; that remains in av-kfik.28.6.
• Code review: dedicated ce-code-review dispatch unavailable in this Codex session; performed manual diff scan and fixed a checks-only aggregate edge case before final validation.

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	e1f24a0
Status:	✅  Deploy successful!
Preview URL:	https://9839a143.agentv.pages.dev
Branch Preview URL:	https://grading-llm-rubric-parsing.agentv.pages.dev

View logs