feat: add Verbatim Quote column and PDF output; implement true Upsert mechanism

fix: show accurate inserted vs skipped counts in Parse+Embed result message

feat: professional PDF reports with dark banners, colour-coded themes, confidence bars (Report 1 and 2)

feat: add NONE rescue pass - re-evaluate NONE papers with forced category assignment

feat: add Report 2 (Theme Aggregation PDF) to Research Impact Audit tab

fix: skip classify_paper_types in SPJIMR pipeline to preserve user-selected corpus type (EMPI/BIBS/etc)

fix: replace upsert with manual dedup to avoid missing unique constraint error

fix: upsert on web_link to prevent duplicate papers on re-upload

fix: pass selected corpus_type to import_pdfs_from_zips and resolve meta_str NameError

Migrate to dedicated research_impact_audits table and fix minor UI bugs

fix(aacsb): add robust JSON repair, escape quotes in prompt, hydrate all paper titles, and print evidence & reasoning in PDF report

fix(ui): make PDF report download button always visible (but interactive=False when empty) and add load session data button

Fix unicode errors in PDF generation with strict sanitize function

Fix fpdf crash causing missing UI updates

Ensure all papers are evaluated gracefully

Add AACSB Audit PDF download feature

Fix 'Untitled' titles due to ID type mismatch and show all audited papers including NONE in UI

Fix PreTrainedModel import error, remove terminal emojis, and allow Step 9 before clustering

feat: Add Step 9 AACSB Impact Audit tab with Mistral LLM evaluation

Use Mistral LLM to clean findings: strip junk/metadata/stats, extract narrative sentences

Fix: merge Historical Reports INTO Step 8 Research Command Center (matches local layout)

Fix: guard all Supabase calls so Step 8 tab renders even if DB is unreachable

Force rebuild: pin gradio>=6.10.0, add markdown dep, push all source files

Remove demo.load that caused NameError crash

Add Load Report button and auto-load event for Historical Reports

Rename Historical Reports to Step 8

Fix Historical Reports tab visibility

Add gspread and google-auth to requirements

Fix truncation limits for full context extraction and add DOI parsing

Fix dependency resolver backtracking by removing bounds

Add missing plotly dependency

Fix deployment: defer agent load, add missing workbench packages and credentials

Upload 40 files

initial commit