feat: stabilize anti-cheat UX, tighten plagiarism detection, and fix landing page links

fix: enhance ASR filler detection and integrate anti-cheat feedback in UI

feat: stabilize AI interview pipeline - Toast notifications, anti-cheat feedback, and ASR accuracy improvements

fix: resolve TypeError in weight calculation

fix: resolve SyntaxError by moving helper function out of main loop

fix: refactor score persistence and fix NameError in plagiarism block

fix: resolve tensor dimension error in plagiarism check

fix: resolve scoring route ImportError and state machine lock

security: implement anti-cheating, disable paste, and fix dashboard skills

docs: update production README and fix question type labeling

fix: enforce state machine idempotency and remove legacy status string

fix: ASR model format and state machine transition gaps

feat: reliability and performance hardening (faster-whisper, state machine, adaptive polling)

fix: add robust fallbacks for Celery tasks when Redis is unavailable

feat: implement background task processing with Celery and Redis

style: run black formatter to fix CI

feat: phase 3 - observability and metrics

feat: phase 2 - state machine, AI orchestration gateway, and test suite

chore: phase 1 - setup hygiene, security hardening, and testing infrastructure

feat: enhance permission handling and add LLM fallback notifications

fix: resolve TypeError in scoring and plan generation for structured project data

Fix TypeError in scoring pipeline and update project extraction for structured data

feat: enhance resume parsing API with extraction quality metadata

feat: implement high-fidelity project extraction with confidence scoring

fix: enhance resume parsing and project extraction robustness

fix: enforce minimum question count and fix violation route imports

fix: comprehensive stability fixes for skip flow, follow-ups, and results page

fix: stabilize interview state and prevent question repetition

feat: comprehensive production hardening of interview state machine and frontend-backend synchronization

fix: resolve skip button 500 error and TTS/recording audio overlap

fix: resolve Firebase popup blocks and interview flow failures

Fix Interview State Machine: Sync frontend/backend contracts and enhance completion logic

feat: complete UI/UX modernization of Interview Simulator frontend

fix: production audit fixes (race conditions, caching, timeouts)

feat: stabilize Veridian system with state recovery, idempotency fixes, and unified design

fix: session caching for adblocker bypass

restore user routes and safer firebase init

fix: safer firebase init

fix: production auth and frontend env

Phase 2 Complete: Auth, Dashboard, Analytics

feat: complete multimodal interview pipeline redesign

fix(backend): proper fallback for single-word generic project names

fix(backend): prevent single-word project truncation and expand question pool

harden: production audit fixes (logging, atomic writes, LLM validation, answer limits, memory leak fix)

fix(backend): prevent hyphenated project names from being truncated to single words (e.g. Full-Stack -> Full) in interview plan generation

Fix project extraction, setup UI placeholders, telemetry bug, UI enhancements

Fix HF chat model support and New Interview restart loop

Fix HF API 404 by using huggingface_hub InferenceClient

Implement CPU fallback system for ASR and LLM scoring

deploy: add Vercel + Hugging Face Space deployment support