feat: update home page demo video, add ngrok bypass header for Docling requests

feat: redesign my_submissions panel + G2 guardrail rewrite + upload modal improvements

fix(postprocess): also strip ACL line numbers from flat markdown string

feat(postprocess): strip ACL review line numbers from Docling structured output

feat: add desk-rejection guardrails + home page CTA polish

feat: add 5-step submission tracker with live polling to my_submissions page

refactor: remove Resend email service — login required, no email needed

config: reduce max file size from 20MB to 10MB

refactor: scope system to ACL ARR only, remove EMNLP/NAACL/COLING references

feat: add Datasets and Software metrics to match full ACL ARR rubric

improve: rewrite synthesizer strengths to evaluate like a human reviewer

fix: 3 review quality improvements from ACL ARR comparison

feat: increase max review loops from 3 to 5

fix: stop loop at exactly 3 iterations; improve synthesizer quote usage

improve: tighten prompts for quote enforcement, specific concerns, and verdict clarity

fix: loop 3 tools now execute before compactor (done set after run_tools, not before)

fix: cap loops at 3, filter already-called tools in planner, deduplicate findings summary

fix: loop 3 tools skipped + planner now receives available tools per paper type

fix: use uvicorn logger for terminal output instead of stderr/print

fix: catch-all wrapper for BackgroundTasks crash + stderr logging

fix: write pipeline logs to stderr so they appear in uvicorn worker terminal on Windows

feat: add detailed terminal logs to multi-agent pipeline nodes

feat: implement multi-agent LangGraph reviewer (4-gate deep review pipeline)

fix: remove hardcoded admin email from config default

fix: strip null bytes in all Supabase write paths (storage layer)

fix: restore /app/ mount, add / redirect to home.html

fix: serve frontend at / instead of /app/ for cleaner URLs

fix: strip null bytes from Tavily content before writing to Supabase

fix: Unicode stdout crash on Windows (UTF-8 encoding)

fix: add tavily_cost + return search_count from search_related_papers

merge: integrate phuc-dev into ntp — Supabase, Auth, Docker, Observability

fix: get_intro_text returns empty when paper title classified as method/skip role

feat: add temperature control per pipeline module

feat: upgrade literature search — ACL Anthology priority + Tavily advanced

fix(parsing): get_intro_text returned empty when the title contains a method keyword

feat: align review rubric with ARR form — half-points, 1-5 OA scale, limitations

fix(frontend): wire session-aware nav onto renamed pages after ntp merge

Merge remote-tracking branch 'origin/ntp' into phuc-dev

feat(auth+db): session-aware nav, ownership fix, admin hardening, schema cleanup

refactor: rename frontend pages — landing→home, index→upload

feat: persist ParsedPaper to data/jobs/<key>_parsed.json for debugging

feat: structured PDF parsing with ParsedPaper schema (Priority 2)

merge: integrate ldh into main

feat(ocr): refine evaluation pipeline and add results

@

Merge remote-tracking branch 'origin/main' into phuc-dev

feat(backend): Supabase-backed storage, observability, and DB bug fixes

improve: domain-specific system prompts for all pipeline modules

feat: add per-job file logging, Docling fallback provider, paper title extraction

fix: resolve remaining conflict marker in _secret_key_html