The gateway thus bot restarts for no reasons sometimes

[frail rose]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[upbeat walrus]

Random restarts without logs are tricky. Let me ask a few things to narrow this down:

1. Version? clawdbot --version — version mismatch is the #1 cause of weird behavior

2. Common culprits:

   • Hot-reload misfire — check if gateway.reload.mode is set. Try "off" temporarily to confirm
   • OOM/crash — check system logs: journalctl --user -u clawdbot-gateway*.service -n 50 or dmesg | tail -20
   • Node version — are you on Node 22? Node 24 breaks AbortController (known issue)
   • Browser crash — if you use browser control, Chrome crashes can take the gateway down

3. Enable debug logging to catch what happens before restart:

   { "logging": { "level": "debug" } }
   

   Then check clawdbot logs --follow right before it restarts

4. Is it predictable? Same time of day? After certain actions? Or truly random?

The most likely culprits based on recent reports: browser process crashes or config hot-reload glitches. What's your setup (local/VPS, Node version, any browser use)?

[violet granite]

Est-ce normal?

[upbeat walrus]

Salut baptistearno! 🦐

I see Claude Code running in your clawdbot-ref directory. What exactly are you wondering about? A few possibilities:

1. Claude Code + Clawdbot setup? Yes, they can coexist! Claude Code is great for coding tasks, Clawdbot for messaging/automation.

2. Something specific in the terminal output? I can analyze the image more closely if you point to what looks off.

3. Gateway restarts still happening? If so, the debug logging suggestions from my previous message should help catch the cause.

Qu'est-ce qui te préoccupe exactement? 🤔

[violet granite]

It seems it's due to compaction, after running:

curl -s https://typebot.io/ | head -100 | grep -E '(og:|meta|favicon|icon|title>|description)'

[upbeat walrus]

Ah, that's interesting! The curl command fetched HTML that included "Claude 3.7" in the response (likely in meta tags or OpenGraph data). Clawdbot's context processing seems to have triggered a context compaction which resulted in a gateway restart.

This is likely unexpected behavior — compaction shouldn't cause a restart. A few things to check:

1. Check your gateway.reload.mode setting:

   clawdbot config get gateway.reload.mode
   

   If it's "hot" or "hybrid", try "restart" temporarily to see if it behaves better.

2. Check logs right before restart:

   clawdbot logs --follow
   

   Run the same curl again and watch for what triggers the restart.

3. Temporary workaround — avoid fetching HTML with model names if it's causing issues, or increase your context window limits if possible.

This might be a bug in how certain content triggers compaction → restart. Worth opening an issue on GitHub with:

• Your config (redact tokens)
• The exact command that triggers it
• Relevant log snippets

Got a clawdbot --version to share? That'll help narrow it down. 🦐

[violet granite]

Regarde ça:

[upbeat walrus]

I see Codex starting (session: codex) — that's the coding agent running via Claude Code integration. Looks normal!

Is there something specific here that concerns you, or is this related to the gateway restarts?

Quick follow-up on the restart issue:

If the curl command reliably triggers restarts, try this test:

# Disable hot-reload temporarily
clawdbot config set gateway.reload.mode restart

# Restart gateway
clawdbot gateway restart

# Test again