#I ask something in my chat and it’s stuck on typing

1 messages · Page 1 of 1 (latest)

frigid martenBOT
#

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

dapper ospreyBOT
#

“Typing forever” usually means: Discord delivered your message, the Gateway started the agent (so it shows typing), but the agent then hung / crashed / hit a model/auth error / failed to send the reply.

To diagnose, I need 3 things from the machine where the Gateway is running:

  1. Version
openclaw --version
  1. Pasteable status report (safe; redacts tokens)
openclaw status --all
  1. Live logs while you reproduce it
  • In one terminal:
# 
openclaw logs --follow
  • Then send a message to the bot in Discord that makes it get “stuck typing”
  • Copy/paste the log lines around the failure (or the last ~50 lines)

If you just want a quick “try this first” fix while you gather logs:

openclaw gateway restart

Also tell me:

  • Is this in a DM with the bot, or in a server channel?
  • Does it happen for every message, or only certain prompts?
severe hornet
#

Unknown model openrouter/auto

dapper ospreyBOT
# severe hornet Unknown model openrouter/auto

That’s why it’s stuck “typing”: the Gateway is trying to run the session on openrouter/auto, but OpenClaw needs a real, specific model ID — openrouter/auto isn’t a valid model ref in OpenClaw, so it errors before it can reply.

Fix:

  1. Pick a valid model (in chat):
  • Run:
    /model list
    then select one from the list (or use /model <number>)
  1. Or from the server where OpenClaw runs:
openclaw models list --provider openrouter --all
openclaw models set <openrouter/some-model-id>
openclaw models status
#
  1. If you used an allowlist (agents.defaults.models), make sure the model you choose is included there too (otherwise you’ll get “Model is not allowed”).

Docs: https://docs.openclaw.ai/concepts/models

If you paste just this output, I can tell you exactly what to set it to:

openclaw --version
openclaw models status
dapper ospreyBOT
#

Those log lines you pasted don’t include any Discord send attempt — they show an embedded run finishing in ~400ms and then a webchat disconnect. So we’re missing the part where Discord tries (and fails) to post the reply.

That said, you already hit the real root cause earlier:

Unknown model openrouter/auto

openrouter/auto is not a valid model in OpenClaw. When the agent tries to run, it fails immediately (often before it can produce a normal reply), so Discord just shows “typing” and then… nothing.

Fix the model (on the machine running the Gateway)

Run:

openclaw --version
openclaw models status
openclaw models list --provider openrouter --all
#

Then set the primary model to an actual OpenRouter model id (pick one from the list output):

openclaw models set openrouter/<model-from-the-list>
openclaw models status
openclaw gateway restart

(Alternatively: rerun openclaw onboard and select a model there.)

If it still “types” but doesn’t post

Grab the right log chunk (Discord + the error). Easiest:

openclaw logs --limit 300 | rg -i "discord|openrouter|unknown model|error|send|typing"
#

Paste that output here.

Also confirm: does the bot have Send Messages permission in that channel, and is this happening in a server channel (not a DM)?

severe hornet
#

I did not set up a channel yet I’m using the web browser

dapper ospreyBOT
# severe hornet I did not set up a channel yet I’m using the web browser

If you’re using the web browser (Control UI / WebChat), that’s totally fine — but it still uses the same model config under the hood.

Right now your Gateway is trying to use openrouter/auto, and OpenClaw doesn’t accept that as a real model, so the run fails and it looks like it’s “typing forever”.

Fast fix (no Discord channel needed)

In the WebChat message box, run:

/model list

Then pick any concrete model from the list (anything that looks like openrouter/<provider>/<model>), and try chatting again.

If /model list is empty or selection fails

On the machine where OpenClaw is running, paste this:

# 
openclaw --version
openclaw models status

If you intend to use OpenRouter, make sure you actually have an OpenRouter API key configured (otherwise models won’t work). Docs:
https://docs.openclaw.ai/providers/openrouter
https://docs.openclaw.ai/concepts/models

Why your log showed “webchat disconnected code=1001”

That usually just means the browser tab refreshed/navigated/closed (normal). The real blocker is the invalid model (openrouter/auto).

Send openclaw models status and I’ll tell you exactly what it’s missing (bad model ref vs missing auth vs allowlist).

severe hornet
#

I want to use Ollama model

dapper ospreyBOT
# severe hornet I want to use Ollama model

To use an Ollama model, the key thing is: Ollama must be reachable from the machine where the OpenClaw Gateway is running (usually http://127.0.0.1:11434).

1) Set up Ollama (on the Gateway machine)

ollama serve
ollama pull llama3.3   # or whatever model you want

2) Enable Ollama in OpenClaw

Ollama doesn’t need a “real” API key, but OpenClaw uses this as the “enable discovery” switch:

export OLLAMA_API_KEY="ollama-local"

(or persist it in config)

# 
openclaw config set models.providers.ollama.apiKey "ollama-local"

3) Switch your session/model away from openrouter/auto

Set a real model, e.g.:

openclaw models set ollama/llama3.3
openclaw models status

Or in WebChat:

/model list
/model ollama/llama3.3