Pipeline

Turn the document drop into verified experiment records — work the steps left to right.

1 Discoversee what's in the drop 2 Extractdocuments → data, checked as you go 3 Linkconnect to experiments 4 Browsethe experiment records
Next: All caught up — every fragment is reviewed and linked. Browse Experiments →

Extract data from the in-scope documents

The ELN reports and handwritten notes are the source of truth (local Qwen, one document at a time, verbatim). The experiment table is read one batch block at a time (Qwen) and serves as a cross-check; only APP DATA still uses a fixed no-AI parser. Each run produces fragments — one document's facts about one experiment — which you check in Verify (step 3). Current scope: Analytical reports (IR / NMR) · Application data summary table · Application test reports · Experiment table — per-researcher summary · Handwritten lab notes · Old ELN experiment reports.
Old ELN experiment reports 14 files 14 extracted → 15 fragments
✓ everything extracted — continue in Verify
Handwritten lab notes 1 file 1 extracted → 19 fragments
✓ everything extracted — continue in Verify
Analytical reports (IR / NMR) 14 files 14 extracted → 19 fragments
✓ everything extracted — continue in Verify
Experiment table — per-researcher summary 1 file 1 extracted → 66 fragments
✓ everything extracted — continue in Verify
Application data summary table 1 file 1 extracted → 44 fragments
✓ everything extracted — continue in Verify
Application test reports 9 files 9 extracted → 9 fragments
✓ everything extracted — continue in Verify
On hold — 4 categories outside the current scope (visible in Discover; no extraction runs)
Experiment table — near-duplicate copy (not parsed) 1 files reference — never extracted
Old ELN summary export (ignored) 1 files reference — never extracted
Monthly review presentation (reference only) 1 files reference — never extracted
Literature / patents / vendor reference 6 files reference — never extracted
Maintenance — ingestion & re-merge tools