feat(eval): add path-derived category field for hierarchical grouping

[christso]

Summary

• Add path-derived category field to eval pipeline for hierarchical organization: Category > Dataset > Test ID
• Categories are derived from eval file directory structure (no new YAML field)
• Propagate through: EvalTest → EvaluationResult → JSONL/manifest → Studio API → Studio UI
• Studio gets two-level drill-down with collapsible category sections, new routes, and updated breadcrumbs

Closes #813

Implementation Plan

Task 1: Add category to core types

• Add readonly category?: string to EvalTest and EvaluationResult interfaces

Task 2: Extract deriveCategory to @agentv/core

• Move from discover.ts to shared packages/core/src/evaluation/category.ts
• Change fallback from "root" to "Uncategorized"
• Export DEFAULT_CATEGORY constant

Task 3: Propagate category through YAML parser

• Add category to LoadOptions, assign to each parsed test case

Task 4: Pass category from CLI run-eval to parser

• Derive category from file path in prepareFileMetadata, pass to loadTestSuite

Task 5: Pass category through orchestrator

• Add category: evalCase.category in buildEvaluationResultCommon() and all error paths

Task 6: Include category in artifact writer and manifest

• Add to IndexArtifactEntry, buildIndexArtifactEntry, ResultManifestRecord, hydrateManifestRecord

Task 7: Add categories API endpoints

• GET /api/runs/:filename/categories — list categories with stats
• GET /api/runs/:filename/categories/:category/datasets — datasets within a category

Task 8: Add Studio types and API hooks

• CategorySummary, CategoriesResponse types
• useRunCategories, useCategoryDatasets hooks

Task 9: Update RunDetail with category grouping

• Group by category first (collapsible sections), then datasets within

Task 10: Add category route in Studio

• New route: /runs/$runId/category/$category showing datasets in that category

Task 11: Update Breadcrumbs and Sidebar

• Category segment in breadcrumbs
• CategorySidebar showing datasets within a category

Task 12: Full verification

• All tests pass (1625+)
• Build clean
• Lint clean
• Unit tests for deriveCategory (8 tests)

Test plan

• Eval files in subdirectories get correct category in JSONL output
• Eval files at root level default to Uncategorized
• Studio run detail groups datasets under category headers
• Click category → see datasets → click dataset → see evals
• Breadcrumbs show: Home > Run > Category > Dataset > Eval
• Old JSONL files without category render under Uncategorized
• All existing tests pass

🤖 Generated with Claude Code

[cloudflare-workers-and-pages]

Deploying agentv with    Cloudflare Pages

Latest commit:	f245b56
Status:	✅  Deploy successful!
Preview URL:	https://4f57f656.agentv.pages.dev
Branch Preview URL:	https://feat-category-field.agentv.pages.dev

View logs