When Telegram is blocked

The local Bot API connects straight to Telegram’s data centers. If Telegram is blocked in the server’s country or by the hosting provider, it can’t sign in, and the bot fails to start with getMe timeouts in its log. Check from the server:

curl -m 10 -sS -o /dev/null https://api.telegram.org && echo reachable

The installer runs the same check. Only if Telegram is unreachable does it ask for a proxy. You can skip the question; the bot will start once the server reaches Telegram.

Pick the proxy address

The proxy is http://host:port or socks5://host:port. Usually it is a VPN client on the same server, such as xray, v2ray, or sing-box. Take the protocol and port of its HTTP or SOCKS inbound from the inbounds section of its configuration.

Containers reach the server as host.docker.internal, so the inbound must listen on the Docker gateway 172.17.0.1, not only on 127.0.0.1. For an xray HTTP inbound on port 10809:

{ "protocol": "http", "listen": "172.17.0.1", "port": 10809 }

Check it with curl -x http://172.17.0.1:10809 https://api.telegram.org, then give the installer http://host.docker.internal:10809.

Turn it on by hand

The Bot API’s own --proxy option covers webhooks only. So the deployment routes the whole Bot API container through the proxy: compose.telegram-proxy.yml adds a small telegram-proxy container (tun2socks) whose network the Bot API shares. Other services still connect directly.

Add to deploy/.env:

COMPOSE_FILE="compose.yml:compose.telegram-proxy.yml"
TELEGRAM_PROXY="http://host.docker.internal:10809"

Then run docker compose up -d. The server needs /dev/net/tun, which most virtual servers have. The installer’s logout from the cloud API goes through the same proxy.