Messages in the dash are empty | Friends of the Crustacean 🦞🤝 | Page 1

mighty fableBOT Jan 25, 2026, 10:45 AM

#

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

maiden aurora Jan 25, 2026, 10:45 AM

#

@flat galleon

flat galleonBOT Jan 25, 2026, 10:46 AM

#

maiden aurora <@1457407575476801641>

Based on the docs, this symptom (agent starts replying but sends empty message) is usually one of these issues:

Quick Checklist

Check if the guild/channel is allowlisted - Discord defaults to groupPolicy: "allowlist", so the guild must be explicitly enabled:
```
clawdbot status
```
Look for AllowFrom: in the output.
Check logs for the actual reason:
```
clawdbot logs --follow
```
Look for messages containing blocked, skip, or unauthorized.
Verify Discord config - The user mentioned in a channel (using mention), but check:

#

Is channels.discord.groupPolicy set to "open" or is the guild allowlisted?
- Are numeric channel IDs used in channels.discord.guilds config?
- Does the bot have Message Content Intent enabled in Discord dev portal?

Windows-specific: Make sure the gateway is actually running (not just the dash):
```
clawdbot gateway status
```

The "empty message" typically means the agent tried to respond but hit a policy block before sending. The logs will tell you exactly why.

Can you run clawdbot logs --follow and try sending a message, then paste what appears? That'll show the exact block reason.

maiden aurora Jan 25, 2026, 10:49 AM

#

📎 message.txt

flat galleonBOT Jan 25, 2026, 10:50 AM

#

maiden aurora

I see the issue - the agent runs and completes (not aborted), but no reply is being sent. The run completes in 679ms which is suspiciously fast.

This points to a few possibilities:

Missing system prompt - Check your config has agents.defaults.systemPrompt or the main agent has one defined
WebChat channel config issue - The webchat channel might need explicit configuration
MiniMax model issue - You're using hf:MiniMaxAI/MiniMax-M2.1 - verify the model works

Quick fixes to try:

Check your system prompt:

clawdbot config get agents.defaults.systemPrompt
# or for main agent
clawdbot config get agents.list

#

If it's missing, add one:

{
  "agents": {
    "defaults": {
      "systemPrompt": "You are a helpful AI assistant. Reply to all messages."
    }
  }
}

Or in the Clawd Dash → Settings → Agent → System Prompt.

Also verify the WebChat channel is configured:

clawdbot config get channels.webchat

#

If it's not set up, you may need to enable it:

{
  "channels": {
    "webchat": {
      "enabled": true
    }
  }
}

Then restart the gateway.

Can you check if you have a system prompt configured? That's the most common cause of silent completions like this.

maiden aurora Jan 25, 2026, 10:54 AM

#

aight here's the output

PS C:\WINDOWS\system32> clawdbot config get agents.defaults.systemPrompt

🦞 Clawdbot 2026.1.23-1 (c8afa82) — Pairing codes exist because even bots believe in consent—and good security hygiene.

Config path not found: agents.defaults.systemPrompt
PS C:\WINDOWS\system32> clawdbot config get agents.defaults.systemPrompt

🦞 Clawdbot 2026.1.23-1 (c8afa82) — I'm the assistant your terminal demanded, not the one your sleep schedule requested.

Config path not found: agents.defaults.systemPrompt
PS C:\WINDOWS\system32> clawdbot config get channels.webchat

🦞 Clawdbot 2026.1.23-1 (c8afa82) — IPC, but it's your phone.

Config path not found: channels.webchat

flat galleonBOT Jan 25, 2026, 10:55 AM

#

maiden aurora aight here's the output PS C:\WINDOWS\system32> clawdbot config get agents.defa...

That's the problem! No systemPrompt = agent has no instructions, and no webchat channel configured = nowhere to send replies.

Fix this:

1. Set a system prompt:

clawdbot config set agents.defaults.systemPrompt "You are a helpful AI assistant. Be concise and helpful."

2. Enable the webchat channel:

clawdbot config set channels.webchat.enabled true

3. Restart the gateway:

#

clawdbot gateway restart

Or via the Clawd Dash:

Settings → Agent → System Prompt: Add something like "You are a helpful AI assistant. Reply to all messages."
Settings → Channels → Enable WebChat

After that, try sending a message in the Dash chat again. The agent should now respond properly.

maiden aurora Jan 25, 2026, 10:56 AM

#

unfortunately that didn't work:

PS C:\WINDOWS\system32> clawdbot config set agents.defaults.systemPrompt "You are a helpful AI assistant. Be concise and helpful."

🦞 Clawdbot 2026.1.23-1 (c8afa82) — If something's on fire, I can't extinguish it—but I can write a beautiful postmortem.

Error: Config validation failed: agents.defaults: Unrecognized key: "systemPrompt"
PS C:\WINDOWS\system32> clawdbot config set channels.webchat.enabled true

🦞 Clawdbot 2026.1.23-1 (c8afa82) — IPC, but it's your phone.

Error: Config validation failed: channels.webchat: unknown channel id: webchat

flat galleonBOT Jan 25, 2026, 10:58 AM

#

maiden aurora unfortunately that didn't work: PS C:\WINDOWS\system32> clawdbot config set age...

The workspace files exist here, but you're on Windows. Your workspace is likely at ~/clawd/ in WSL or your Windows user directory.

To diagnose this, run:

clawdbot status

This will show:

Your configured workspace path
Whether the webchat channel is properly set up
Any configuration issues

Also check the logs for any errors around reply delivery:

clawdbot logs --follow

Then send a message and look for any error messages about reply/send failures.

#

The likely issue: The agent ran in 679ms (very fast) and completed without an error, but the reply didn't get routed back to the Control UI. This could be:

Missing workspace bootstrap files on your Windows setup
The webchat channel needs explicit configuration
A routing issue for Control UI replies

Can you run clawdbot status and paste the output? That will show your workspace path and help identify what's missing.

maiden aurora Jan 25, 2026, 11:00 AM

#

📎 message.txt

flat galleonBOT Jan 25, 2026, 11:01 AM

#

maiden aurora

Found it! The key line: "1 bootstrapping" - the agent is stuck in bootstrap mode, meaning it's failing to load the workspace files.

The agent workspace is at ~\.clawdbot\agents\main\ (Windows path).

Fix:

1. Check if workspace files exist in your agent directory:

dir "$env:USERPROFILE\.clawdbot\agents\main\*.md"

2. If missing, create the required bootstrap files (minimum: AGENTS.md and SOUL.md):

# Create AGENTS.md
@"

#

# AGENTS.md - Workspace Configuration

## Safety

- Don't exfiltrate private data. Ever.
- Don't run destructive commands without asking.
- `trash` > `rm` (recoverable beats gone forever)
- When in doubt, ask.

## Group Chats

You have access to your human's stuff. That doesn't mean you *share* their stuff. In groups, you're a participant — not their voice, not their proxy. Think before you speak.
"@ | Out-File -FilePath "$env:USERPROFILE\.clawdbot\agents\main\AGENTS.md" -Encoding utf8

# Create SOUL.md

#

@"
# SOUL.md - Who You Are

You are a helpful AI assistant. Be concise and friendly. Help the user with their questions and tasks.

## Core Truths

- Be genuinely helpful
- Have opinions
- Be resourceful
- Remember you're a guest

## Boundaries

- Private things stay private

#

- When in doubt, ask before acting externally
- Never send half-baked replies
"@ | Out-File -FilePath "$env:USERPROFILE\.clawdbot\agents\main\SOUL.md" -Encoding utf8

3. Or set a different workspace path if you want to use a different directory:

clawdbots.workspace "C:\Users\YourName\clawd"

4. Restart the gateway:

clawdbot gateway restart

After bootstrap completes (the files exist), the agent will respond normally. Can you check if those files exist?

maiden aurora Jan 25, 2026, 11:33 AM

#

actually if i open clawdbot.json from ~/.clawdbot it says that the workspace is "workspace": "C:\Users\myname\clawd", directory which has all the files like soul user and so on

flat galleonBOT Jan 25, 2026, 11:33 AM

#

maiden aurora actually if i open clawdbot.json from ~/.clawdbot it says that the workspace is ...

The workspace path is there but the agent is still bootstrapping. Let me check a few things:

1. Verify the files exist at that exact path:

dir "C:\Users\myname\clawd\*.md"

2. Check logs for file/read errors:

clawdbot logs --follow | Select-String -Pattern "bootstrap|workspace|AGENTS|SOUL|error" -Context 2

3. Try using forward slashes in the path (more reliable):

clawdbot config set agents.defaults.workspace "C:/Users/myname/clawd"
clawdbot gateway restart

#

4. Also verify the agent-specific workspace (sometimes it defaults to agentDir):

The issue might be:

Windows backslash path issue
Permission problem reading from that path
The agents.defaults.workspace isn't being applied to the main agent

Can you run those checks and paste what you find? Specifically check if dir "C:\Users\myname\clawd\*.md" shows files exist, and if the logs show any errors about reading bootstrap files.

maiden aurora Jan 25, 2026, 11:40 AM

#

i managed to change the working directory, restarted but still only empty messagesa as before. with dir//stuff/*md this is the output

Directory: C:\Users\ralph\clawd

Mode LastWriteTime Length Name

-a---- 1/24/2026 11:47 PM 7848 AGENTS.md
-a---- 1/24/2026 11:47 PM 1465 BOOTSTRAP.md
-a---- 1/24/2026 11:47 PM 167 HEARTBEAT.md
-a---- 1/24/2026 11:47 PM 632 IDENTITY.md
-a---- 1/24/2026 11:47 PM 1673 SOUL.md
-a---- 1/24/2026 11:47 PM 858 TOOLS.md
-a---- 1/24/2026 11:47 PM 481 USER.md

flat galleonBOT Jan 25, 2026, 11:41 AM

#

maiden aurora i managed to change the working directory, restarted but still only empty messag...

The files exist and look fine. Let me dig deeper - the agent runs but produces no output. This could be the MiniMax model not returning content.

Check what's actually happening in the gateway logs:

clawdbot logs --follow

Then send a message in the dash and look for:

Any error messages
Whether the model returns any text/content
Any "reply" or "send" entries after the run completes

Also check the session transcript - the agent might be replying but it's getting lost:

clawdbot sessions list
clawdbot sessions history --sessionKey <the-session-id-from-above>

#

Another possibility: The embedded synthetic agent (hf:MiniMaxAI/MiniMax-M2.1) might be the issue. Let's check your model config:

clawdbot config get agents.defaults.model
clawdbot config get agents.defaults.models

Try switching to a different model temporarily to isolate the issue - if you have Anthropic or OpenAI configured:

clawdbot config set agents.defaults.model.primary "anthropic/claude-sonnet-4-5-20250529"
# or
clawdbot config set agents.defaults.model.primary "openai/gpt-5.2"
clawdbot gateway restart

Can you run the log tail and paste what shows up when you send a message? That'll show exactly where the reply is getting lost.

maiden aurora Jan 25, 2026, 11:43 AM

#

📎 message.txt

flat galleonBOT Jan 25, 2026, 11:44 AM

#

maiden aurora

The run completes in 595ms with aborted=false but no reply is sent. This is the issue - the agent completes but the reply never goes back to the webchat.

Check the session transcript to see if content was generated:

clawdbot sessions list

Then:

clawdbot sessions history --sessionKey "agent:main:webchat:dm:unknown" --includeTools

Also look for any reply/send errors in logs:

clawdbot logs --follow | Select-String "reply|send|webchat" -Context 2

#

The most likely cause: The synthetic provider (hf:MiniMaxAI/MiniMax-M2.1) might be returning an empty response. Let's verify by temporarily switching to a real API model:

# If you have Anthropic configured
clawdbot config set agents.defaults.model.primary "anthropic/claude-sonnet-4-5-20250529"

# Or OpenAI
clawdbot config set agents.defaults.model.primary "openai/gpt-5.2"

clawdbot gateway restart

Then try sending a message again.

Can you check the sessions history to see if any content was generated? And try switching to a different model to see if that fixes it?

maiden aurora Jan 25, 2026, 10:44 PM

#

ok so the issue was with the model, so i switched to ollama.
when i do clawdbot models list
i receive

🦞 Clawdbot 2026.1.23-1 (c8afa82) — I don't judge, but your missing API keys are absolutely judging you.

Model Input Ctx Local Auth Tags
ollama/qwen3-coder:30b - - - - default,configured,missing
synthetic/hf:MiniMaxAI/MiniMax-M2.1 text 188k no yes configured,alias:MiniMax M2.1

so clawdbot is seeing ollama with qwen3

i already setx OLLAMA_API_KEY "ollama-local"

the issue start when trying to message in the dash, the assistant is struck with the three dots, if i message on discord the bot reacts with eyes emoji to my message but then it responds with

⚠️ Agent failed before reply: Unknown model: ollama/qwen3-coder:30b.
Logs: clawdbot logs --follow

so i do it and the logs say this

22:20:26 debug diagnostic {"subsystem":"diagnostic"} lane enqueue: lane=main queueSize=1
22:20:26 debug diagnostic {"subsystem":"diagnostic"} lane dequeue: lane=main waitMs=2 queueSize=0
22:20:26 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=24 error="Error: Unknown model: ollama/qwen3-coder:30b"
22:20:26 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:discord:channel:1424481657947422832 durationMs=30 error="Error: Unknown model: ollama/qwen3-coder:30b"
22:20:26 error Embedded agent failed before reply: Unknown model: ollama/qwen3-coder:30b

but if i try to

clawdbot config set models.providers.ollama.apiKey 'ollama-local'

as in the docs i receive

Error: Config validation failed: models.providers.ollama.baseUrl: Invalid input: expected string, received undefined

tranquil barn Jan 25, 2026, 10:46 PM

#

Man, im having the same
Issue but using OpenAI 5.2 I just have no idea what else to try

maiden aurora Jan 25, 2026, 10:49 PM

#

@flat galleon

flat galleonBOT Jan 25, 2026, 11:10 PM

#

maiden aurora ok so the issue was with the model, so i switched to ollama. when i do clawdbot ...

The issue is Ollama needs both apiKey AND baseUrl. The validation is failing because baseUrl is undefined.

Fix - set both at once:

clawdbot config set '{
  "models": {
    "providers": {
      "ollama": {
        "apiKey": "ollama-local",
        "baseUrl": "http://localhost:11434"
      }
    }
  }
}'

Or more simply:

flat galleonBOT Jan 25, 2026, 11:12 PM

#

tranquil barn Man, im having the same Issue but using OpenAI 5.2 I just have no idea what else...

Same symptoms (agent starts, completes, empty reply)?

Quick diagnostic checks:

Run status:
```
clawdbot status
```

Check model config:

clawdbot config get agents.defaults.model
clawdbot config get agents.defaults.models

Check logs when sending a message:

#

clawdbot config set models.providers.ollama.apiKey "ollama-local"
clawdbot config set models.providers.ollama.baseUrl "http://localhost:11434"
clawdbot gateway restart

Then verify the model works:

curl http://localhost:11434/api/tags

Also make sure Ollama is actually running:

ollama list

If you get a connection error, start Ollama:

maiden aurora Jan 25, 2026, 11:20 PM

#

@flat galleon i literally told you that

clawdbot config set models.providers.ollama.apiKey "ollama-local"

gives back

Error: Config validation failed: models.providers.ollama.baseUrl: Invalid input: expected string, received undefined

can you please refrain from doing the same mistake multiple times?

ollama is working when i

PS C:\Users\ralph> clawdbot config get agents.defaults.model

i receive

"primary": "ollama/qwen3-coder:30b"

flat galleonBOT Jan 25, 2026, 11:21 PM

#

   clawdbot logs --follow

The most common causes:

Model not in agents.defaults.models allowlist
Missing provider credentials
Webchat channel routing issue

Can you run those three commands and paste the outputs? That'll show exactly where the breakdown is.

maiden aurora Jan 25, 2026, 11:22 PM

#

ok i did the logs --follow then sent a message on the dash main, i received

23:22:00 debug diagnostic {"subsystem":"diagnostic"} lane enqueue: lane=session:agent:main:main queueSize=1
23:22:00 debug diagnostic {"subsystem":"diagnostic"} lane dequeue: lane=session:agent:main:main waitMs=4 queueSize=0
23:22:00 debug diagnostic {"subsystem":"diagnostic"} lane enqueue: lane=main queueSize=1
23:22:00 debug diagnostic {"subsystem":"diagnostic"} lane dequeue: lane=main waitMs=2 queueSize=0
23:22:00 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=23 error="Error: Unknown model: ollama/qwen3-coder:30b"
23:22:00 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent:main:main durationMs=30 error="Error: Unknown model: ollama/qwen3-coder:30b"
23:22:00 error Embedded agent failed before reply: Unknown model: ollama/qwen3-coder:30b

flat galleonBOT Jan 25, 2026, 11:26 PM

#

ollama serve

After setting the baseUrl, restart the gateway and try again. The model should be recognized now.

flat galleonBOT Jan 25, 2026, 11:46 PM

#

maiden aurora ok i did the logs --follow then sent a message on the dash main, i received 23:...

The model is still unknown. This means either:

The baseUrl wasn't saved correctly
The model needs to be in the allowlist

Check your full Ollama config:

clawdbot config get models.providers.ollama

Also check models config:

clawdbot config get agents.defaults.models

If the baseUrl is missing, set it again (this is the issue):

maiden aurora Jan 25, 2026, 11:46 PM

#

no, it is not, now im not even getting a response on discord while it's still frozen with three dots in the dash main chat

#

and your first instruction returns an error

PS C:\Users\ralph> clawdbot config get models.providers.ollama

🦞 Clawdbot 2026.1.23-1 (c8afa82) — Because Siri wasn't answering at 3AM.

Config path not found: models.providers.ollama

flat galleonBOT Jan 25, 2026, 11:50 PM

#

clawdbot config set models.providers.ollama.baseUrl "http://localhost:11434"
clawdbot gateway restart

If that fails, set the entire block at once:

clawdbot config set '{
  "models": {
    "providers": {
      "ollama": {
        "apiKey": "ollama-local",
        "baseUrl": "http://localhost:11434"
      }
    }
  },

#

  "agents": {
    "defaults": {
      "models": {
        "ollama/qwen3-coder:30b": {}
      }
    }
  }
}'
clawdbot gateway restart

Verify Ollama is running:

curl http://localhost:11434/api/tags
# or

#

ollama list

Can you run those config get commands and paste the outputs? That'll show what's actually saved vs what's missing.

flat galleonBOT Jan 25, 2026, 11:59 PM

#

maiden aurora no, it is not, now im not even getting a response on discord while it's still fr...

Gateway is stuck. Restart it:

clawdbot gateway restart

If that hangs, force restart:

clawdbot gateway stop
# Wait 5 seconds
clawdbot gateway start

Check status:

clawdbot gateway status
clawdbot status

#

Then try sending a message again. The "three dots frozen" usually means the gateway is not running or crashed.

flat galleonBOT Jan 26, 2026, 1:23 AM

#

Did you get unblocked? If you're still having issues:

Quick restart sequence:

clawdbot gateway restart
clawdbot status

If it's still stuck, check if the gateway process is running:

tasklist | findstr clawdbot

And check logs:

clawdbot logs --follow

#

What's your current status?

maiden aurora Jan 26, 2026, 1:45 AM

#

changed approach, installed wsl, installed ollama and polled my model then

i run curl -fsSL https://clawd.bot/install.sh | bash

then i stopped hatching but configured other stuff

but when i try to

clawdbot something

it says command not found

flat galleonBOT Jan 26, 2026, 1:46 AM

#

maiden aurora changed approach, installed wsl, installed ollama and polled my model then i ru...

You need to either restart your WSL shell or manually add it to PATH.

Quick fix - start a new shell or source your profile:

source ~/.bashrc
# or
exec $SHELL

If still not found, find where it installed:

which clawdbot
# or
find ~ -name "clawdbot" -type f 2>/dev/null | head -5

Or reinstall with global access:

#

curl -fsSL https://clawd.bot/install.sh | bash -s -- --install-method npm

Then verify:

clawdbot --version
clawdbot status

The install script should have added ~/.npm-global or similar to your PATH. Starting a new terminal session should pick it up.

#Messages in the dash are empty

Quick Checklist

Fix this:

Fix: