diff --git a/app-next/docs/OPEN_ITEMS.md b/app-next/docs/OPEN_ITEMS.md index d7625432..51ec54fc 100644 --- a/app-next/docs/OPEN_ITEMS.md +++ b/app-next/docs/OPEN_ITEMS.md @@ -6,7 +6,7 @@ Last updated: 2026-02-18 | Item | Status | File | Details | | --------------------------- | ----------- | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -| CORS on `data.openml.org` | Blocked | `hooks/useParquetData.ts` | All data files proxied through `/api/proxy-file` because `data.openml.org` blocks browser requests. Adding `Access-Control-Allow-Origin: *` would remove this workaround. | +| CORS on `data.openml.org` | Discussing | `hooks/useParquetData.ts` | **Current**: Data files < 5MB proxied through `/api/proxy-file` due to missing CORS headers. **Team feedback** (Joaquin): Need to understand why CORS is needed; may be unsafe. **Alternative being considered**: Precompute stats and serve via main REST API instead of requiring browser access to raw data files. | | Dash 502 for large datasets | Bug | (server-side) | `openml.org/dashboard/data/47160` returns 502. Missing `metric` component in Dash callback. | | Bookmark API | Not started | `dataset-actions-menu.tsx:98` | UI exists, no backend call. | | Collections API | Not started | `dataset-actions-menu.tsx:289` | Dialog is placeholder, needs endpoint to fetch/create collections. | @@ -25,11 +25,11 @@ Last updated: 2026-02-18 | Item | Status | Details | | ----------------------------- | ------------ | --------------------------------------------------------------------------------- | -| Distribution (small datasets) | Working | Parquet/ARFF < 10MB loads in browser. | -| Distribution (large datasets) | Partial | Nominal features show from metadata. Numeric shows "coming soon". | +| Distribution (small datasets) | Working | Parquet/ARFF < 5MB loads in browser via `/api/proxy-file`. | +| Distribution (large datasets) | Partial | Datasets > 5MB use Dash iframe: `/dashboard/data-features/{id}`. Nominal features show from metadata. Numeric shows "coming soon". | | Correlation | Working | Shows when parquet loads. "Coming soon" otherwise. | | Parquet-wasm | Intermittent | Sometimes fails silently on valid files (e.g. dataset 1590). Falls back to ARFF. | -| Dash JSON API | Future | Server-side computation for large datasets. Would replace "coming soon" messages. | +| Dash JSON API | Discussing | **Current**: Large datasets use Dash iframe. **Team feedback** (Joaquin): Considering precomputing main statistics and dataset preview, serving via main REST API (not Flask). Would allow rendering charts in Next.js with consistent styling instead of iframes. | ## 4. Auth & Infrastructure @@ -43,3 +43,28 @@ Last updated: 2026-02-18 - **Question for team**: Incremental pushes to Vercel preview vs. batched PRs — what works best? - Minor merge conflicts when syncing with upstream. + +## Extras improvements + +### Dataset Edit Form - Markdown Preview + +**Current state**: The dataset edit form (`dataset-edit-form.tsx`) has a plain `