Fix: LLM Timeout — Remote Ollama Server

Running Ollama on a remote machine (a home server, a VPS, another machine on your LAN) introduces latency that the default OpenClaw timeout doesn't account for. The error LLM timeout received almost always means one of four things: wrong URL, model not loaded, timeout too short, or network blocked.

The Error

The ECONNREFUSED 127.0.0.1 variant means OpenClaw is still pointing at localhost instead of the remote server — the most common mistake.

Fix A: Set the Correct Remote baseUrl (Most Common)

By default, OpenClaw looks for Ollama at http://localhost:11434. When your Ollama instance runs on another machine, you must point to its IP:

{
  "llm": {
    "provider": "ollama",
    "baseUrl": "http://192.168.1.100:11434",
    "model": "llama3.2"
  }
}

Also make sure Ollama on the remote machine is listening on all interfaces (not just localhost):

OLLAMA_HOST=0.0.0.0 ollama serve

Or set it permanently in your Ollama service config:

# /etc/systemd/system/ollama.service.d/override.conf
[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"

Fix B: Keep the Model Loaded

Ollama unloads models from memory after 5 minutes of inactivity by default. When OpenClaw sends a request to an unloaded model, Ollama must reload it — which can take 30–90 seconds and trigger the timeout.

On the Ollama server, set OLLAMA_KEEP_ALIVE to keep the model resident:

# In your Ollama service environment
OLLAMA_KEEP_ALIVE=-1 ollama serve

# Or in systemd override
Environment="OLLAMA_KEEP_ALIVE=-1"

Use -1 to never unload, or a duration like 60m to keep it loaded for 60 minutes.

Fix C: Increase OpenClaw's Timeout

For slower hardware (CPU-only, older GPU), even a loaded model can take longer than the default timeout to respond to a prompt. Increase the timeout in OpenClaw's config:

{
  "llm": {
    "provider": "ollama",
    "baseUrl": "http://192.168.1.100:11434",
    "model": "llama3.2",
    "requestTimeout": 120000
  }
}

120000 is 120 seconds. For very large models (70B+) on CPU, you may need 300000 (5 minutes).

Fix D: Open Firewall Port 11434

If your Ollama server is on a different subnet or behind a firewall, port 11434 may be blocked:

# Allow from your OpenClaw machine's IP only
sudo ufw allow from 192.168.1.50 to any port 11434

# Or allow from entire LAN (less secure)
sudo ufw allow from 192.168.1.0/24 to any port 11434

Test Connectivity Before Restarting OpenClaw

# Check if Ollama is reachable
curl http://192.168.1.100:11434/api/tags

# Run a quick inference test
curl http://192.168.1.100:11434/api/generate \
  -d '{"model":"llama3.2","prompt":"hi","stream":false}'