Skip to content

step-05-ui.md

Latest commit

 

History

History
47 lines (33 loc) · 3.83 KB

step-05-ui.md

File metadata and controls

47 lines (33 loc) · 3.83 KB
Raw

Step 5: UI for OCR improvement (build on existing benchmarking)

Parent: Feature 008 README
Implementation order: 5
Depends on: Step 4 (orchestration, workflow override, run candidate), existing benchmarking frontend (feature 003).

Goal

Provide a UI for the OCR improvement pipeline and candidate runs by extending the existing benchmarking feature. Users can trigger the pipeline, see recommendations, start a candidate run, and use existing run detail and comparison pages for results. No separate app or duplicate run/comparison screens.

Scope

  • Entry point: At benchmark project or definition level, an "OCR improvement" (or "Improvement pipeline") section or tab.
  • Trigger pipeline: Action that runs the full pipeline (aggregate HITL → AI recommendation → create candidate workflow → start benchmark run). Calls a new backend API; response includes candidate workflow ID and benchmark run ID.
  • Show recommendations: Display AI tool recommendations (tool IDs, placement, rationale) when available (e.g. from pipeline response).
  • Run candidate: When the user has a candidate (from the pipeline), "Run candidate" starts a benchmark run using the existing start-run API with optional candidateWorkflowVersionId (config loaded from that workflow version row). Existing run list and run detail pages are used for status and baseline comparison.
  • Comparison: Use existing run detail and run comparison pages for baseline vs candidate; promote baseline when satisfied.

Requirements

  • The UI SHALL extend the existing benchmarking frontend (apps/frontend/src/features/benchmarking/).
  • The UI SHALL call a new pipeline run endpoint (e.g. POST .../benchmark/projects/:projectId/definitions/:definitionId/ocr-improvement/run) with required context (workflow ID, user ID, optional HITL filters).
  • The UI SHALL use the existing start-run API with optional candidateWorkflowVersionId when starting a candidate run (no new endpoint for starting runs).
  • The UI SHALL reuse existing run list, run detail, run comparison, and promote-baseline flows.

Implementation tasks

  • Add backend endpoint for running the OCR improvement pipeline (see Step 4 orchestration).
  • Add an "OCR improvement" section or tab in the benchmarking UI at project or definition level.
  • In that section: button/action to "Run improvement pipeline", display of pipeline result (candidate workflow ID, benchmark run ID, recommendations summary), and "Run candidate" that starts a run with override when a candidate is available.
  • Extend the existing start-run payload (e.g. in useRuns or call site) to support optional candidateWorkflowVersionId when running a candidate.

Acceptance criteria

  • User can trigger the OCR improvement pipeline from the benchmarking UI for a chosen definition (with workflow and project/definition context).
  • Pipeline result shows candidate workflow ID, benchmark run ID, and recommendations summary (applied/rejected counts, tool IDs).
  • User can start a candidate run (workflow override) using the existing start-run flow when a candidate exists; run appears in the same definition’s run list.
  • Existing run detail and comparison pages are used to view baseline vs candidate results; no duplicate comparison UI.

References

  • Step 4: step-04-benchmark-integration-workflow-comparison.md
  • docs-md/OCR_IMPROVEMENT_PIPELINE.md — pipeline endpoint, UI behavior, and how candidate runs relate to baseline comparison
  • Benchmarking frontend: apps/frontend/src/features/benchmarking/
  • Start-run API: POST /api/benchmark/projects/:projectId/definitions/:definitionId/runs — see CreateRunDto and benchmark controller for optional candidateWorkflowVersionId when running a candidate workflow