Robust curl retries (--retry 4 --retry-all-errors) + aligned subprocess timeouts + download validation

PTB transport: curl_cffi fast path + reliable curl-binary fallback (fixes commands/buttons hanging)

Fix NameError: import threading (broke PTB transport init)

Reliability: route Telegram msg/photo/getFile/download through real curl binary; shorter transport timeout

Face swap: retry on 'Too many requests' rate limit with backoff; expose hf_token_set in /status

nettest v2: hammer curl_cffi vs forced-IPv4 vs curl binary

Route all api.telegram.org calls (getFile, send photo/msg/video, webhook) through curl_cffi

Route PTB through curl_cffi transport (browser TLS) for HF Spaces Telegram egress

Add /nettest diagnostic endpoint

Force IPv4 DNS: fix api.telegram.org read-timeout on HF Spaces (no IPv6 egress)

Update bot to working Render version (yesterday's, all fixes) for failover

Use real curl binary for api-data requests to match browser TLS fingerprint

Fresh cffi session per api-data request to keep Cloudflare cookies valid

Fix stale cffi session: refresh every 5min for fresh Cloudflare cookies

Revert _seaart_post to original curl_cffi, keep browser headers

Add browser fingerprint headers to SeaArt session (x-browser-id, x-device-id, x-gray-tag)

Switch workflows to apply endpoint with inputs format

Switch workflow submission to SDK auth (token header)

Separate REST API routes into api_routes.py

Revert video workflow to original api-data endpoint code

Switch to creativity/generate/apply endpoint for workflows

Use SDK request() for workflow api-data endpoint

Use SDK session for workflow submissions to fix auth

Switch workflow submission from api-data to apply endpoint

Fix video workflow auth: add token header matching SeaArt SDK

Minimal verbose: only 1 initial message, no step-by-step progress

Revert to original flow: auto-select video scenario (random/keyword)

Move video scenario selection to step 2 (right after prompt)

Add 22 video scenarios + scenario selection step in bot flow

Update app.py

Fix: HF Space no longer steals webhook on startup

Fix build: add gcc for curl-cffi, fix seaart SDK URLs, use local SDK

Sync with Render: Docker deploy, /switch failover, video fix, /logs /analyze

fix: parallel video delivery + channel forwarding + cffi session reliability

Add raw fallback for setWebhook + timeout on initialize

Never tear down running telegram_app + raw fallbacks everywhere

Update app.py

Stop health check from destroying working webhook connection

Remove IPv4-only DNS restriction — let default resolution work

Add raw requests fallback for sending photos/videos to Telegram

Fix video delivery: retry sends with reconnect + debug logging

Health check pings Telegram, reduce keepalive to 30s

Retry send operations with reconnect on stale connections

Remove hardcoded Telegram IP, use dynamic DNS with IPv4 only

User lock auto-expires after 30min to prevent permanent blocks

Update app.py

Improve reconnection: 3 retries, clear user locks on reconnect

Handle SeaArt no-credits error (80217) gracefully

Add per-user lock to prevent concurrent pipelines

Switch to correct Telegram API IP 149.154.166.110