Spaces:
Paused
Paused
Commit History
-
feat: show all pages with jump-to-mention navigation
- documents/route.js now includes ALL pages, not just those with datasets - Added pages_with_mentions array to document data - PageNavigator has ⏮/⏭ jump buttons to skip to pages with mentions - Green ● dot indicator on pages that have data mentions - Users can browse full document context while quickly jumping to mentionse715c5d -
fix: block document access for users not in annotator_config
Previously getUserAssignedDocs returned null when user wasn't found, which the caller treated as 'no filter, show all docs'. Now returns empty {} which correctly results in zero visible documents.0880d65 -
feat: multi-corpus support
- corpora.json registry: add new corpora by adding entries - All APIs/utils resolve paths via config.js helpers - Per-corpus doc assignments: docs: { wbg: [...], unhcr: [...] } - Document selector shows [World Bank] Doc 3 labels - Leaderboard/progress scan all corpora - generate_assignments.py handles per-corpus distributiona2c885c -
feat: per-annotator document assignments with configurable overlap
- annotator_config.yaml: defines annotators and overlap % (default 10%) - generate_assignments.py: auto-distributes docs with overlap, seed-based - Documents API reads config, filters docs by logged-in user - 10 shared docs for inter-annotator agreement - Fallback: show all docs if user not in config79ba9a0 -
fix: remove PDF HEAD check that was filtering out valid docs (WB returns 301 for HEAD)
b72d00c -
feat: prepare_data.py script + use has_revalidation filter in documents API
- prepare_data.py: scans local docs, uploads missing _direct_judged.jsonl to HF, generates updated wbg_pdf_links.json with src_docname and has_revalidation fields - documents API: only lists docs with has_revalidation=true - Fixed misleading error message in storage.jsfb404c5
-
fix: validate PDF URLs before listing documents, skip broken/unreachable PDFs
7c5c449 -
feat: add loading spinner and error fallback to PDF viewer
Shows 'Loading PDF...' with spinner while iframe loads. Shows error state with 'Open PDF directly' link if load fails. Resets loading state on document/page change.9cc2491 -
feat: on-demand document loading, no limit, PDF loading indicator
- /api/documents now returns just the link list (no pre-fetching) - /api/document returns annotatable_pages on-demand (no page param) - page.js fetches + caches annotatable_pages when doc is selected - PdfViewer shows loading spinner overlay while iframe loads - Removed MAX_DOCS_TO_SCAN concept entirelyb584890 -
update: reverted local changes
8be778a -
feat: per-annotator document assignment
- Add annotator_assignments.json config (docs list or docs_range per user) - Documents API filters by ?user= query param - Client passes username from OAuth cookie to documents fetch - rafmacalaba gets docs 1-10, rafamacalaba gets docs 10-100 - Falls back to MAX_DOCS_TO_SCAN if no assignmentfabd779 -
feat: human validation, recursive highlighting, data mentions rename
- Add human validation UI (✅/❌) for model-extracted data mentions - Validation metadata stored inside dataset_name alongside judge_ fields - Recursive highlighting works inside bold/italic/links - Rename 'Datasets' to 'Data Mentions' throughout - Switch input from _raw.json to _direct_judged.jsonl - Document API reads local files in dev (highlights appear after save) - Add /api/validate (PUT for validation, DELETE by index) - Add /api/whoami for HF username auto-fetch - Floating chevron panel toggle with badge count - Rename annotation_tag to annotatorda957b0 -
feat: implement HF Datasets remote fetching and save
5b87eae -
feat: implement dual local/HF storage with page-level UI tracking
42df2a9