v5.3.3

What's Changed

• feat(dashboard): preserve context for result row details by @christso in #1449
• merge final reviewed results stack by @christso in #1462
• docs(agents): require PR-based merges by @christso in #1463
• test(evals): add PR workflow guard self-eval by @christso in #1464
• fix(results): make sidecar artifacts canonical by @christso in #1465
• feat(dashboard): define trace session read model by @christso in #1468
• docs(results): specify storage retention oplog plan by @christso in #1471
• feat(evals): support suite input shorthand by @christso in #1469
• docs(phoenix): document read-only AgentV boundary by @christso in #1470
• eval/dashboard: read linked Phoenix sessions by @christso in #1472
• feat(dashboard): serve trace session artifacts by @christso in #1473
• chore(package): remove deprecated @agentv/eval package by @christso in #1475
• style(brand): dim AgentV wordmark middle letters by @christso in #1476
• feat(core): normalize trace artifacts for dashboard by @christso in #1474
• fix(beads): harden worktree identity checks by @christso in #1460
• feat(results): emit metrics sidecar by @christso in #1478
• feat: add Vitest workspace verifier adapter by @christso in #1480
• fix(beads): guard worktree tracker state by @christso in #1482
• refactor(dashboard): remove Phoenix read-through UI by @christso in #1481
• feat(cli): infer Vitest verifier eval commands by @christso in #1483
• fix(results): preserve configured git identity by @christso in #1484
• feat(sdk): expose evaluate api by @christso in #1485
• fix(providers): support Copilot SDK OpenAI endpoint parity runs by @christso in #1479
• refactor: remove private Phoenix adapter package by @christso in #1486
• [codex] Support dashboard remote result branches by @christso in #1477
• docs: standardize contract design references by @christso in #1487
• docs: design repeat-run reliability by @christso in #1488
• feat(results): move artifacts under results experiments by @christso in #1489
• feat(dashboard): set run experiment and tags by @christso in #1490
• docs: design remote metadata conflict resolution by @christso in #1491
• fix(results): handle push conflicts with backup policy by @christso in #1492
• fix(agent-providers): stabilize Copilot and Pi SDK runs by @christso in #1493
• fix(pi-cli): route endpoint overrides through azure provider by @christso in #1494
• docs(providers): capture SDK runtime isolation guidance by @christso in #1498
• feat(workspace): read git_cache.mirrors from project-local .agentv/config.yaml by @christso in #1497
• fix(docker): drop COPY of removed phoenix-adapter package by @christso in #1499
• fix(dashboard): show file contents inline in run Files tab by @christso in #1501
• fix(results): push same-repo results to project origin, not a synthesized URL by @christso in #1502
• fix(dashboard): overlay git artifacts into the existing run file tree by @christso in #1504
• Separate experiments from eval definitions by @christso in #1500
• docs: compound experiment separation learning by @christso in #1505
• docs: design conflict-free results sync without force push by @christso in #1503
• feat(results): no-force-push results sync (auto-merge loop) by @christso in #1506
• feat(results): temp-branch fallback + OK-to-resync (no force push) by @christso in #1507
• feat(dashboard): pending-merge card + OK-to-resync (av-raf.4) by @christso in #1508
• docs: no-force-push results sync behavior (av-raf) by @christso in #1509
• feat(results): signal auto-merged remote pushes by @christso in #1510
• feat(experiments): support suite test selectors by @christso in #1514
• fix(cli): avoid duplicate createRequire binding by @christso in #1515
• feat(results): restructure AgentV run artifacts by @christso in #1513
• docs(eval): clarify expected_output assertion contract by @christso in #1516
• docs: clarify eval and experiment boundaries by @christso in #1517
• docs: migrate ADRs to adrs format by @christso in #1518
• feat(config): support local config overlays by @christso in #1519
• docs(adr): define transcript artifact contract by @christso in #1520
• feat(results): write normalized transcript artifacts by @christso in #1521
• docs: finalize inline experiment ADR by @christso in #1522
• cleanup(results): remove legacy manifest artifact aliases by @christso in #1525
• refactor(results): remove public trace artifact surface by @christso in #1526
• feat(evaluation): inline experiment runtime in eval files by @christso in #1524
• feat(dashboard): polish transcript timeline by @christso in #1527
• fix(results): avoid duplicate raw provider logs by @christso in #1528
• docs(verification): document codex local proxy dogfood by @christso in #1529
• chore(entire): use manual commit strategy by @christso in #1530
• Fix workspace baseline branch safety by @christso in #1531
• docs: document composite AND/OR patterns by @christso in #1532
• docs: update composite strict-or example to bun run by @christso in #1533
• fix(targets): use flat Copilot SDK provider config by @christso in #1534
• test(eval): guard workspace path override compatibility by @christso in #1535
• chore: remove accidental root artifacts by @christso in #1536
• fix(skills): default agentv-bench to CLI mode by @christso in #1537
• feat(results): rename result directory row field by @christso in #1540
• docs(results): codify eval result identity contract by @christso in #1539
• feat(eval): compose imported experiment defaults by @christso in #1541
• feat(dashboard): identify result rows by eval path by @christso in #1543
• fix(results): keep dashboard status read-only for repo paths by @christso in #1542
• docs(verification): clarify contract dogfood evidence by @christso in #1544
• docs(eval): clarify eval authoring contracts by @christso in #1545
• docs(schema): record benchmark primitive decision by @christso in #1546
• ci: publish commit-addressed build artifacts by @christso in #1548
• fix(eval): tighten workspace composition contracts by @christso in #1547
• fix(eval): remove dry-run mock execution by @christso in #1550
• fix(eval): move local workspace binding out of eval YAML by @christso in #1549
• fix(eval): reject per-case target selection by @christso in #1551
• fix(cli): remove promptfoo import support by @christso in #1552
• fix(results): remove persistent require push config by @christso in #1553
• feat(dashboard): cache git-backed result lists by @christso in #1554
• fix(results): ignore missing WIP branch cleanup by @christso in #1555
• fix(results): isolate target bundles under run timestamps by @christso in #1558
• [codex] fix(config): remove flat results remote alias by @christso in #1562
• docs(adr): clarify result row sidecar identity by @christso in #1556
• [codex] Rename result row manifest to run_manifest by @christso in #1563
• feat(results): simplify project + results config to one flat symmetric schema by @christso in #1565
• refactor(projects): drop redundant project name field, display id by @christso in #1566
• Support flatter eval imports by @christso in #1567
• docs(results): document result artifact contract by @christso in #1568
• fix(results): restore index jsonl artifact by @christso in #1569
• feat(core)!: require canonical assertions, remove assert authoring alias by @christso in #1571
• [codex] Support eval policy config by @christso in #1576
• fix(artifacts): rename generated task bundle to test bundle by @christso in #1574
• docs(readme): modernize quick start eval and fix agentv compare example by @christso in #1572
• refactor(core): centralize case conversion boundary by @christso in #1573
• docs(readme): document experiment eval contract by @christso in #1577
• fix(schema): reject authored eval YAML workers by @christso in #1578
• fix(artifacts): keep eval runs at timestamp root by @christso in #1580
• docs(readme): define singular target eval contract by @christso in #1579
• Write file_changes diff sidecar by @christso in #1581
• feat(eval): canonicalize repeat policy schema by @christso in #1582
• fix(artifacts): keep run status in agentv format by @christso in #1584
• feat(artifacts): write eval runs at results root by @christso in #1583
• docs(readme): align results layout with run root by @christso in #1585
• feat(evals): support default_test threshold by @christso in #1588
• docs(adr): stabilize eval authoring contract by @christso in #1587
• fix(eval): default repo workspaces to fresh materialization by @christso in #1586
• feat(eval): author experiment as promptfoo-style tags.experiment (av-4hzh) by @christso in #1589
• docs(experiment): label runs via tags.experiment, drop top-level experiment by @christso in #1590
• feat(dashboard): Tags tab — group runs by any promptfoo tag key; remove legacy manual tags (av-qsxw) by @christso in #1591
• docs(concepts): define Tags concept; note manual tags removed (av-qsxw) by @christso in #1593
• docs(agents): make .agents/*.md guide reads a required, trigger-keyed gate by @christso in #1595
• docs(plan): eval authoring restructure — promptfoo superset (DRAFT) by @christso in #1594
• docs(plans): add promptfoo-compatible extensions plan by @christso in #1592
• docs(agents): clarify worker guidance and cc-mirror docs by @christso in #1596
• feat(core): adopt nunjucks eval templating by @christso in #1597
• feat(eval): add lifecycle extensions and agent rules by @christso in #1607
• Canonicalize targets around labels by @christso in #1598
• feat(eval): normalize transcript artifacts by @christso in #1600
• feat(core): load file-backed datasets by @christso in #1601
• docs: plan CLI command surface alignment by @christso in #1606
• [codex] Add promptfoo assertion grader surface by @christso in #1599
• feat(evals): add prompt instance expansion by @christso in #1602
• feat: align repeat config with attempt artifacts by @christso in #1608
• Grading contract: assertion_results rows by @christso in #1603
• feat(eval): add rerun-failed runner pooling by @christso in #1609
• feat(web): add versioned docs snapshot by @christso in #1611
• refactor(targets): rename request batching field by @christso in #1612
• feat(eval): finalize promptfoo-aligned restructure by @christso in #1610
• feat(web): add Next docs version snapshot by @christso in #1613
• docs(eval): align llm-rubric quickstart by @christso in #1614
• test(cli): cover Agent Skills expectations criteria by @christso in #1617
• docs(eval): document v4.42.4 eval migrations by @christso in #1616
• fix(core): simplify repo resolver contract by @christso in #1619
• Migrate example eval suites to suite.yaml by @christso in #1618
• refactor(grading): rename script grader wording by @christso in #1615
• Plan coding-agent target runtimes by @christso in #1621
• refactor(observability): remove eval OTLP export by @christso in #1622
• fix(results): publish runs at results branch root by @christso in #1620
• feat(results): implement ADR-0017 bundle content by @christso in #1623
• feat: add composable target runtime config contract by @christso in #1625
• docs: standardize AgentV brand casing by @christso in #1624
• docs: place quickstart concurrency in evaluate options by @christso in #1626
• fix(web): publish current docs at root routes by @christso in #1627
• docs(agent): refresh restructure references by @christso in #1628
• docs(web): align authoring contract docs by @christso in #1629
• docs(examples): migrate authored eval contracts by @christso in #1630
• fix(eval): accept target id in targets files by @christso in #1631
• fix(results): keep raw transcripts valid jsonl by @christso in #1632
• docs(sdk): clarify custom assertion terminology by @christso in #1633
• fix(eval): reject authored execution concurrency by @christso in #1634
• Align Promptfoo assertion grouping vocabulary by @christso in #1635
• fix(core): inherit default_test vars for prompt expansion by @christso in #1639
• docs: add Promptfoo parity matrix by @christso in #1641
• fix(results): slim canonical artifact contract by @christso in #1642
• fix(web): show docs version selector on mobile by @christso in #1643
• fix(core): align assert-set scoring semantics by @christso in #1644
• fix(dashboard): include execution errors in run-list totals by @christso in #1638
• fix(dashboard): render canonical transcript artifacts by @christso in #1637
• fix(dashboard): clarify repeat attempt rows by @christso in #1636
• fix(dashboard): consolidate run detail actions by @christso in #1640
• fix(results): tighten result artifact contract by @christso in #1645
• docs(plan): sequence Promptfoo grading alignment by @christso in #1647
• feat(core): add Promptfoo-compatible transform runtime by @christso in #1648
• docs(adr): specify Promptfoo grading output contract by @christso in #1649
• docs(workflow): prefer targeted local validation by @christso in #1651
• docs(adr): tighten wire-format casing invariant by @christso in #1653
• fix(eval): reject authored direct input by @christso in #1646
• Complete hard-deprecation codemod migrations by @christso in #1654
• fix(eval): remove public preprocessor authoring by @christso in #1656
• docs(adr): record coding-agent target runtime contract by @christso in #1655
• fix(dashboard): contain transcript detail on mobile by @christso in #1660
• fix(eval): migrate authored expected output to vars by @christso in #1657
• docs(transform): publish promptfoo file transform example by @christso in #1661
• feat(core): replay normalized transcripts by @christso in #1658
• Fix pi-rpc host runtime protocol dispatch by @christso in #1662
• fix(sdk): align script grader results by @christso in #1659
• fix(codex): speak app-server JSON-RPC protocol by @christso in #1665
• fix(runtime): install sdk child runner deps by @christso in #1663
• feat(core): parse Promptfoo llm-rubric results by @christso in #1664
• feat(dashboard): link raw transcript artifacts by @christso in #1666
• docs: capture Copilot runtime research by @christso in #1667
• docs(adr): codify environment recipes for coding-agent evals by @christso in #1668
• fix(runtime): unblock SDK provider live matrix for av-t2o5.6 by @christso in #1669
• feat(eval): add environment recipe schema loader (av-noh3.2.2) by @christso in #1670
• docs(environment): preserve reference audit guardrails by @christso in #1672
• feat(runtime): execute host environment recipes by @christso in #1671
• Implement recursive grading result artifacts by @christso in #1673
• feat(dashboard): render component grading artifacts by @christso in #1674
• feat(runtime): execute docker environment recipes by @christso in #1676
• Snapshot environment recipe provenance in run artifacts (av-noh3.2.5) by @christso in #1675
• docs(agent): clarify matrix authoring boundaries by @christso in #1677
• fix(config): reject top-level eval imports by @christso in #1678
• Require argv-only environment setup commands by @christso in #1679
• docs(environment): teach environment recipe vocabulary by @christso in #1681
• fix(contract): remove workspace testbed public surfaces by @christso in #1680
• feat(assertions): add promptfoo-compatible skill-used by @christso in #1682
• feat(assertions): add promptfoo trajectory graders by @christso in #1683
• fix(artifacts): flatten metrics sidecar by @christso in #1685
• fix(schema): reject stale authored skill and trajectory graders by @christso in #1686
• docs(migration): align skill and trajectory assertion docs by @christso in #1687
• fix(core): reject authored use_target targets by @christso in #1688
• docs: keep result publishing in global config by @christso in #1684
• feat(config): add v5-native env_path/env_from environment injection by @christso in #1689
• feat(eval): load explicit TypeScript eval configs by @christso in #1690
• fix(config): reject removed eval case aliases by @christso in #1691
• feat(config): lower scenarios into tests by @christso in #1693
• docs(grading): align reference answers and artifacts by @christso in #1694
• docs(eval): finish promptfoo authoring guidance by @christso in #1695
• feat(evaluation): support scenario file references by @christso in #1696
• docs(examples): clarify scenarios matrix use case by @christso in #1698
• feat(results): slim index.jsonl target_execution/transcript_summary to sidecars by @christso in #1699
• results(summary): rename cases to tests, trim run metadata by @christso in #1697
• docs(tests): lock run summary/index/sidecar boundary contract by @christso in #1700
• dashboard(results): adapt target_error_kind reader to slim index rows by @christso in #1701
• fix(eval): repair broken targets.yaml and two eval-run bugs found during dogfood by @christso in #1702
• feat(eval): wire defaults.grader as a config-level grader fallback by @christso in #1703
• fix(eval): surface pending-merge branch and raw grader response in failures by @christso in #1704
• feat(config)!: hard-deprecate the graders: config block — a grader is just a target by @christso in #1705
• fix(providers): emit skill call metadata by @christso in #1706
• docs(adr): codify promptfoo-first provider authoring by @christso in #1710
• fix(results): hard-remove legacy grading.json compat reads in manifest.ts by @christso in #1711
• feat(core): normalize public providers schema by @christso in #1708
• feat(provider-surface): migrate SDK CLI docs to providers by @christso in #1707
• feat(promptfoo): validate exported provider configs by @christso in #1709
• docs(providers): map internal target naming migration by @christso in #1712
• fix(artifacts): write provider bundle snapshots by @christso in #1713
• feat(cli): migrate provider authoring surface by @christso in #1714
• fix(config): route providers through unified config by @christso in #1715
• refactor(core): hard-rename provider definition APIs by @christso in #1716
• feat(grader): support agent-rubric compatibility by @christso in #1717
• test(promptfoo): add oracle export gate by @christso in #1720
• docs(dogfood): record dashboard child assertion UAT by @christso in #1721
• fix(grader): select grader providers from shared pool by @christso in #1718
• Add provider-local environment overlays by @christso in #1719
• fix(sdk): align eval config providers surface by @christso in #1722
• style(dashboard): sharpen dark results command center by @christso in #1723
• Align repeat authoring with Promptfoo numeric repeat by @christso in #1724
• fix(ci): guard eval schema generation by @christso in #1725
• docs(evals): audit Promptfoo schema parity by @christso in #1726
• fix(eval): reject top-level experiment authoring by @christso in #1728
• fix(evaluation): remove public llm-grader assertion by @christso in #1730
• feat(core): support promptfoo provider declaration forms by @christso in #1729
• feat(package): expose agentv authoring facade by @christso in #1727
• feat(dashboard): compare providers side by side by @christso in #1731
• refactor(cli): discover provider catalogs from providers yaml by @christso in #1732
• refactor(sdk): hard-rename provider proxy client by @christso in #1736
• fix(dashboard): use provider wording in recent runs by @christso in #1734
• feat(dashboard): filter provider comparison differences by @christso in #1733
• refactor(cli): use provider labels in eval selection by @christso in #1735
• docs(providers): define provider-owned batching boundary by @christso in #1737
• fix(core): hard-deprecate runner batch config by @christso in #1740
• docs(examples): add provider-owned batching example by @christso in #1739
• [codex] remove Batch CLI eval docs by @christso in #1738
• docs: deprecate AgentV in README by @christso in #1741
• fix(release): use provider flag in contract eval by @christso in #1742
• fix(release): repair contract eval references by @christso in #1743
• fix(release): use Azure for contract evals by @christso in #1744
• fix(release): use OpenRouter for contract evals by @christso in #1745
• chore(release): print contract assertion failures by @christso in #1746
• fix(contract): read grader payload from stdin fd by @christso in #1747
• fix(contract): emit script grader checks by @christso in #1748
• fix(release): align workspace rubric with output by @christso in #1749
• fix(release): correct repo eval reference paths by @christso in #1750
• fix(contract): accept empty grader criteria by @christso in #1751
• chore(contract): ignore generated eval workspace by @christso in #1752

Full Changelog: v4.42.4...v5.3.3