fix: merge and preserve user roles during chat logging upsert

fix: resolve timezone offset in VOD end_time calculations

fix: resolve infinite recursion in safeFetch

fix: allow cloudflare pages domain in CORS

feat: add rate limit warning banner (HTTP 429) to UI

Fix: only cache successful 2xx responses and validate streams array structure

Fix: handle non-array response for streams in App.jsx (prevent crash on rate-limiting)

feat: add automatic self-healing for empty stream viewers

perf: reduce egress by caching stats endpoints longer and limiting getChatMessages

style: optimize transcribe_confirm prompt and skip fully transcribed audio

feat: add backfill resume support and session/retry fixes for VODs

feat: throttle VOD resolution checks to 5m and auto-complete streams older than 3 hours with no VOD

feat: integrate automatic VOD ID resolution in chat_mod_worker.py

feat: automatic VOD ID resolution for stream sessions and historical pending streams

feat: fetch moderator avatars via API proxy with initials fallback

style: premium redesign of the Moderation tab matching reference

fix: race condition for stats loading when user is authorized

fix: production oauth redirect to frontend host

feat: add Better Stack webhook receiver for Telegram alerts

refactor: modularize client App.jsx tab components and update Netlify CSP

fix: date parsing format for Whisper sync

feat: smart gap detection and user roles sync backfill improvements

feat: smart gap detection with coverage windows - track exact time ranges captured per stream

feat: worker and chat_mod_worker now also sync roles to /api/log/roles on every message batch

fix: logRolesBatch always updates roles, only preserves newer last_seen

feat: role_backfiller auto mode - fetches all VODs from server

fix: add root Dockerfile for HF Spaces

fix: add HF Spaces metadata to README

feat: global roles migration, VIP badges, role backfiller script

Optimize KPI cards to request overall stats summary from /api/stats/summary and display true totals.

Optimize message ingestion with logChatMessagesBatch: bulk upsert to database.

Integrate stream-start and stream-end signaling in worker.py and add /api/log/stream-start endpoint.

Optimize moderation logging: fix duplicate collisions, auto-refresh streamer OAuth tokens, and implement IRC keep-alive timeouts. Add unit tests.

fix: prevent full-screen spinner on background stream polls

feat: map truncated 'кримб' to 'кримбр'

feat: add more synonyms for кримбр

feat: tune Whisper settings and add word mappings

fix: adjust Whisper parameters to disable temperature fallback and prevent hallucinations on silence/noise

feat: host Google Fonts locally and tighten Content Security Policy

feat: implement LRU cache eviction policy to prevent memory leaks

security: sync CORS and error handling security updates in main repo

feat: adjust center group threshold to 250 for chat words

fix: add pagination to unpaginated Supabase queries to bypass 1000 row limit

fix: scale down word cloud rendering for low absolute frequency words

fix: enforce mobile 1-col layout and prevent horizontal blowout

fix: resolve mobile layout swipe and word cloud scaling

style: make dashboard layout responsive for mobile devices

chore: add Google Search Console verification meta tag

feat: implement caching, rate limiter, and security headers (CSP) for decoupled deploy

chore: replace Render reference with generic cloud server reference in loading screen