I don't get the response of my agent in the chat | Friends of the Crustacean 🦞🤝 | Page 1

jolly sun Feb 15, 2026, 10:13 AM

#

We speak with bot, I send the image and then he off. I use VPS to run the bot.

wicked stirrupBOT Feb 15, 2026, 10:14 AM

#

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

sweet windBOT Feb 15, 2026, 10:15 AM

#

If this is Discord: when you “send an image” as an image-only message, OpenClaw will often ignore it in guild channels because mentions are required by default. So it looks like the bot “went offline”, but it just never routed/processed that message.

Try this first:

Re-send the image with text that mentions the bot (add any caption + mention).
Or reply to the bot’s last message and include a mention in the reply text.

To confirm what’s happening, on your VPS run and then re-send an image once:

openclaw status
openclaw gateway status
openclaw logs --follow

In the logs you’ll typically see something like “mention required” (message dropped) if that’s the issue.

If it’s not mention-gating, the next most common cause is: your agent/model is text-only, so the provider errors when an image is included (you’ll see a 400-style error in logs). In that case you need to use a vision-capable model for that agent (or add a vision-capable fallback).

Docs that match this behavior:

#

Discord channel policy / mentions: https://docs.openclaw.ai/channels/discord
Deep triage ladder: https://docs.openclaw.ai/gateway/troubleshooting

Paste the few log lines that appear right after you send the image (plus your openclaw --version) and I can tell you exactly which case it is.

jolly sun Feb 15, 2026, 10:15 AM

#

Telegram bot

sweet windBOT Feb 15, 2026, 10:16 AM

#

On Telegram, if the bot stops replying only after you send a photo, there are 3 super common causes:

1) You’re in a group (not DM) and the bot needs a mention / privacy mode blocks it

By default, Telegram groups often require @botusername mention (requireMention: true).
Also: Telegram “privacy mode” can prevent the bot from even seeing non-command group messages unless you disable it or make the bot admin.

Quick test in the group:

Send: @YourBotUsername hi
Or run: /activation always

If that fixes it, then it’s mention/privacy-mode behavior (not a crash).

2) The image is too large for OpenClaw’s Telegram media limit

OpenClaw has channels.telegram.mediaMaxMb (default 5 MB). If your photo is bigger, it can be skipped.

3) Your model/agent is not vision-capable

If the agent’s model can’t accept images, the run errors right when you attach a photo (you’ll see the error in logs).

#

What I need from you (fastest way to diagnose)

On your VPS, run:

openclaw --version
openclaw status
openclaw channels status --probe
openclaw logs --follow

Then send one photo to the bot again, and paste the log lines that appear right after. (Don’t paste your bot token.)

Also tell me: is this happening in a DM or a group?

Docs (reference):

#

Telegram channel behavior + privacy mode: https://docs.openclaw.ai/channels/telegram
General “no replies” troubleshooting: https://docs.openclaw.ai/gateway/troubleshooting

jolly sun Feb 15, 2026, 10:18 AM

#

🦞 OpenClaw 2026.2.13 (203b5bd) — Siri's competent cousin.

│
◇ Config ───────────────────────────────────────────────────╮
│ │
│ Config invalid; doctor will run with best-effort config. │
│ │
├────────────────────────────────────────────────────────────╯
Config invalid
File: ~/.openclaw/openclaw.json
Problem:

ui.assistant.avatar: Too big: expected string to have <=200 characters

Run: openclaw doctor --fix
Gateway not reachable. Is it running and accessible?
Gateway target: ws://127.0.0.1:18789
Source: local loopback
Config: /root/.openclaw/openclaw.json
Bind: loopback
Hint: run openclaw doctor.

sweet windBOT Feb 15, 2026, 10:19 AM

#

That output shows the real blocker right now: your config is invalid, so the Gateway likely isn’t starting (and the CLI can’t reach it).

Specifically:

ui.assistant.avatar is too long (<= 200 chars required)
then: Gateway not reachable at ws://127.0.0.1:18789

Fix

Edit your config file and shorten or remove that field:

File:

/root/.openclaw/openclaw.json

Change/remove something like:

ui: {

#

  assistant: {
    avatar: "..." // must be <= 200 chars (don’t paste base64/data URLs here)
  }
}

Use a short value instead (example):

"🦞" or "https://example.com/avatar.png" (short URL)

Run the repair + restart:

openclaw doctor --fix
openclaw gateway restart
openclaw gateway status

#

You want openclaw gateway status to show RPC probe: ok.

Then we can debug the “image makes it stop replying”

After the Gateway is reachable, run:

openclaw logs --follow

…and send an image again in Telegram, then paste the log lines that appear right after.

One quick question: are you running the Gateway + CLI on the same VPS (as root), or is the Gateway on the VPS but you run the CLI from your laptop? (If it’s from your laptop, ws://127.0.0.1:18789 would be the wrong target.)

jolly sun Feb 15, 2026, 10:21 AM

#

I have SHH channel with VPS server. We did it with open claw bot. I asked him to fix his image in the dashboard. He did it, then died.

#

What should i do to back online my bot

sweet windBOT Feb 15, 2026, 10:22 AM

#

That makes sense — the “fix avatar in the dashboard” likely wrote a huge string (often a data:image/...;base64,...), and now OpenClaw refuses to start because your config fails schema validation:

ui.assistant.avatar: Too big (<= 200 chars)

So the bot didn’t “die” randomly — the Gateway won’t boot with an invalid config, and then you get no replies.

Recovery (SSH into your VPS)

Open the config file it reported (yours is root):

/root/.openclaw/openclaw.json

Find this key and delete it or shorten it:

ui: {
  assistant: {
    avatar: "...." // MUST be <= 200 chars
  }

#

Good safe values:

an emoji like "🦞"
or a short https URL (not base64)

Then run:

openclaw doctor --fix
openclaw gateway restart
openclaw gateway status

You’re aiming for RPC probe: ok.

#

If you want, paste:

the openclaw doctor --fix output (last ~30 lines)
and openclaw gateway status

…and I’ll tell you what’s still blocking, if anything.

#

Your bot is offline because the Gateway won’t start with an invalid config. Right now it’s failing on:

ui.assistant.avatar: Too big (<= 200 chars)

That usually happens when the dashboard saved a base64/data URL as the avatar.

Get it back online (SSH into the VPS)

Run these on the VPS (same user as the gateway — in your output it’s root):

openclaw config unset ui.assistant.avatar
openclaw doctor --fix
openclaw gateway restart
openclaw gateway status

You want openclaw gateway status to show RPC probe: ok.

#

If `gateway restart` says no service is installed

Just run it in the foreground to get online fast:

openclaw gateway

(Then later you can re-install the daemon/service, but this will bring the bot back immediately.)

Important going forward

For ui.assistant.avatar, use only:

a short emoji like "🦞" or
a short https://... URL

Don’t paste base64 (data:image/...) there.

If you paste the output of openclaw gateway status after the fix (or the first error lines if it still fails), I’ll tell you the next step.

jolly sun Feb 15, 2026, 10:24 AM

#

Problem:

ui.assistant.avatar: Too big: expected string to have <=200 characters

Run: openclaw doctor --fix
│
◇
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-02-15.log
Command: /usr/bin/node /usr/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json (invalid)
Config issue: ui.assistant.avatar: Too big: expected string to have <=200 characters
Config (service): ~/.openclaw/openclaw.json (invalid)
Service config issue: ui.assistant.avatar: Too big: expected string to have <=200 characters

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 20138, state active, sub running, last exit 0, reason 0)
Warm-up: launch agents can take a few seconds. Try again shortly.
RPC probe: failed
RPC target: ws://127.0.0.1:18789
gateway closed (1006 abnormal closure (no close frame)): no close reason
Gateway target: ws://127.0.0.1:18789
Source: cli --url
Config: /root/.openclaw/openclaw.json

Gateway port 18789 is not listening (service appears running).
Logs: journalctl --user -u openclaw-gateway.service -n 200 --no-pager

Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

sweet windBOT Feb 15, 2026, 10:24 AM

#

That status output is very clear: the service is “running”, but the Gateway process never successfully starts listening because the config is still invalid:

ui.assistant.avatar: Too big (<= 200 chars)

So step 1 is: remove/fix that field in the config file (doctor can’t fully repair it while it can’t parse/validate it).

Fix the config (on the VPS)

SSH in, then open:

nano /root/.openclaw/openclaw.json

Find the block:

#

ui: {
  assistant: {
    avatar: "...."
  }
}

and either delete the whole avatar: line or change it to something short, e.g.:

avatar: "🦞"

Important: don’t use data:image/...base64,... here (that’s what makes it huge).

Save and exit.

#

Restart the gateway service

Then run:

openclaw doctor --fix
systemctl --user restart openclaw-gateway.service
openclaw gateway status

Verify it’s actually listening

ss -ltnp | grep 18789 || true

If it’s still not listening, paste:

#

journalctl --user -u openclaw-gateway.service -n 200 --no-pager

One more note from your output: you’re binding to loopback (127.0.0.1). That’s fine for Telegram long-polling, but the dashboard will only open on the VPS itself unless you use an SSH tunnel / Tailscale later.

jolly sun Feb 15, 2026, 10:34 AM

#

how to choose all line because i have too long it

sweet windBOT Feb 15, 2026, 10:35 AM

#

You don’t need to “select all” that huge avatar text. Just delete the whole line.

If you use nano (easy)

Open the config:

nano /root/.openclaw/openclaw.json

Find the avatar line:

Press Ctrl + W (search)
Type: avatar
Press Enter

Delete it:

Press Ctrl + K to cut/delete the whole line
- Even if it looks like many lines on screen, nano is usually just wrapping one very long line.
- If it’s really multiple lines, press Ctrl + K a few times until the whole avatar: ... is gone.

#I don't get the response of my agent in the chat

1) You’re in a group (not DM) and the bot needs a mention / privacy mode blocks it

2) The image is too large for OpenClaw’s Telegram media limit

3) Your model/agent is not vision-capable

What I need from you (fastest way to diagnose)

Fix

Then we can debug the “image makes it stop replying”

Recovery (SSH into your VPS)

If you want, paste:

Get it back online (SSH into the VPS)

If gateway restart says no service is installed

Important going forward

Fix the config (on the VPS)

Restart the gateway service

Verify it’s actually listening

If you use nano (easy)

If `gateway restart` says no service is installed