patent sage
Environment:
OpenClaw v2026.4.22 (npm global install, Dockerfile based on ghcr.io/openclaw/openclaw:2026.4.5 image)
Ubuntu 24.04 / Docker / Hostinger VPS
Node.js 24.14.0
Telegram channel (webhook mode via n8n proxy)
Problem: After upgrading from v2026.4.5 to v2026.4.22, the gateway reaches ready state with Telegram listed as a loaded plugin, but the Telegram channel provider never starts. No [telegram] log appears at all — no error, no "starting provider", nothing. The channel fails completely silently.
What works:
Gateway boots and reaches ready (8 plugins including telegram) ✅
Heartbeat starts ✅
Hooks load (5 internal hook handlers) ✅
Bot token present in TELEGRAM_BOT_TOKEN env var ✅
require('@grammyjs/runner') resolves inside the container ✅
openclaw doctor detects no blocking issues (only WhatsApp module missing, which is unused)
What doesn't work:
After [gateway] starting channels and sidecars..., NO channel-related logs appear
No [telegram] [default] starting provider log
No error, no crash, no timeout message — complete silence
Health-monitor never triggers a channel restart
/new command on Telegram gets no response
Relevant config (channels.telegram):
{
"enabled": true,
"dmPolicy": "pairing",
"groups": { "-1003829681602": { "requireMention": false } },
"groupPolicy": "allowlist",
"streaming": { "mode": "off" },
"webhookUrl": "https://n8n-fyuu.srv1366717.hstgr.cloud/webhook/telegram-proxy",
"webhookSecret": "[redacted]",
"webhookHost": "0.0.0.0"
}
Typical boot log:
[gateway] loading configuration…
[gateway] resolving authentication…
[gateway] starting...
[plugins] browser installed bundled runtime deps: ...
[plugins] diagnostics-otel installed bundled runtime deps: ...
[gateway] starting HTTP server...
[canvas] host mounted at http://0.0.0.0:18789/__openclaw__/canvas/
[health-monitor] started (interval: 300s, startup-grace: 60s, channel-connect-grace: 120s)
[gateway] agent model: omniroute/tristan-openclaw
[gateway] ready (8 plugins: acpx, browser, device-pair, diagnostics-otel, discord, phone-control, talk-voice, telegram; 96.0s)
[heartbeat] started
[gateway] starting channels and sidecars...
[hooks:loader] Loading managed hook code into the gateway process.
[hooks] loaded 5 internal hook handlers
--- NOTHING AFTER THIS --- no [telegram] logs, no [discord] logs, no errors
What I've tried:
openclaw doctor --fix — fixed legacy streaming config keys (scalar → object format)
Verified Grammy deps installed (@grammyjs/runner resolves OK)
Restored webhook config (webhookUrl, webhookSecret, webhookHost)
Migrated streaming format from "off" (scalar) to {"mode": "off"} (object)
Multiple container recreations and restarts
Verified TELEGRAM_BOT_TOKEN env var is present
Upgrade path: v2026.4.5 → v2026.4.22 (direct jump, 17 days of releases)
Question: Why would ALL channels (Telegram and Discord) silently fail to start after starting channels and sidecars...? Is there a known issue with channel initialization in v2026.4.22, or a missing migration step for this version jump?