feat(eval): finalize promptfoo-aligned restructure

[christso]

Summary

This PR finalizes the promptfoo-aligned eval restructure branch and folds in the cleanup decisions from av-dkn5:

• replaces workspace lifetime with workspace.scope: suite | attempt
• removes pooled workspace mode and the workspace clean/list commands
• keeps --workspace-path / execution.workspace_path as the explicit static local override
• removes feedback.json and the Dashboard human-review overlay surface
• removes authored evaluators aliases, snake_case grader aliases, and root-level budget_usd
• keeps structured g-eval / rubric criteria with per-criterion assertion rows in grading.json
• updates README, public docs, ADRs, examples, and AI-facing skill references for the current contract

Target References

Eval YAML can reference configured targets by label:

target: codex-gpt5

And can define an eval-local variant:

target:
  extends: codex-gpt5
  label: codex-gpt5-high-reasoning
  reasoning_effort: high

extends references a label from .agentv/targets.yaml or targets.yaml; label is the local result/comparison name.

Validation

• bun test packages/core/test/paths.test.ts packages/core/test/evaluation/workspace-config-parsing.test.ts packages/core/test/evaluation/workspace/setup.test.ts packages/core/test/evaluation/repo-schema-validation.test.ts packages/core/test/evaluation/orchestrator.test.ts packages/core/test/evaluation/extensions.test.ts apps/cli/test/eval.integration.test.ts apps/cli/test/commands/prepare/prepare.test.ts — 209 passed
• bun test packages/core/test/evaluation/loaders/grader-parser.test.ts packages/core/test/evaluation/loaders/jsonl-parser.test.ts packages/core/test/evaluation/loaders/config-loader.test.ts packages/core/test/evaluation/eval-inline-experiment.test.ts packages/core/test/evaluation/validation/eval-validator.test.ts packages/core/test/evaluation/validation/eval-file-schema.test.ts apps/cli/test/commands/results/serve.test.ts apps/dashboard/src/lib/result-table.test.ts apps/cli/test/commands/trace/trace.test.ts — 564 passed
• bun test packages/core/test/evaluation/loaders/config-loader.test.ts packages/core/test/evaluation/workspace/setup.test.ts packages/core/test/evaluation/loaders/eval-yaml-transpiler.test.ts — 135 passed
• bun --filter @agentv/core typecheck
• bun --filter agentv typecheck
• bun --filter @agentv/dashboard build
• bun --filter @agentv/web build
• bun run validate:examples — 61 valid / 0 invalid
• bun run lint
• git diff --check

Live Dogfood

Live local OpenAI-compatible dogfood passed using http://127.0.0.1:10531/v1 with gpt-5.3-codex-spark as both target and grader path. The eval used target: local-openai from a temporary targets.yaml with label: entries and a type: g-eval assertion.

Result: PASS, 1/1 tests, mean score 100%.

Private evidence branch: EntityProcess/agentv-private:evidence/av-dkn5-eval-restructure-dogfood-2026-07-03
Commit: ceed547

The extracted grading-check.json shows g-eval ran through local-openai-grader and wrote criterion-level rows to grading.json.assertion_results and graders[].assertion_results.

Related

• Bead: av-dkn5
• Related parent: av-kfik.16

---

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	baa3e20
Status:	✅  Deploy successful!
Preview URL:	https://47058a93.agentv.pages.dev
Branch Preview URL:	https://docs-av-kfik-readme-docs.agentv.pages.dev

View logs