Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.wednesdayai.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
Ollama

Ollama runs AI models locally on your machine. No API key, no cloud, no data leaving your network. WednesdayAI auto-discovers a running Ollama instance.

​
Prerequisites

Install Ollama from ollama.com. Then pull at least one model: 
ollama pull llama3.3
Confirm Ollama is running: 
ollama list

​
Connecting WednesdayAI

WednesdayAI detects Ollama automatically when it is running on the default port. No login step is required. 
openclaw login
# choose: Ollama
# confirm or enter the base URL (default: http://localhost:11434)
Or set the environment variable: 
export OLLAMA_HOST=http://localhost:11434
Use the native Ollama URL, not the OpenAI-compatible /v1 endpoint.Set baseUrl to http://host:11434 — not http://host:11434/v1.WednesdayAI uses Ollama’s native API. The /v1 shim exists for third-party OpenAI-compatible clients; pointing WednesdayAI at it will cause unexpected errors.

​
Config reference

{
  providers: {
    ollama: {
      baseUrl: "http://localhost:11434",  // native Ollama URL — no /v1
    }
  },
  agents: {
    defaults: {
      model: "ollama/llama3.3",
    }
  }
}
The model ID uses the ollama/ prefix to tell the gateway which provider to route to:
Ollama model nameWednesdayAI model ID
llama3.3ollama/llama3.3
mistralollama/mistral
qwen2.5-coderollama/qwen2.5-coder
deepseek-r1ollama/deepseek-r1
Run openclaw status --providers to see which Ollama models WednesdayAI can reach.

​
Remote Ollama

If Ollama runs on a different machine (a home server, a GPU workstation), set the URL to that host: 
{
  providers: {
    ollama: {
      baseUrl: "http://192.168.1.50:11434",
    }
  }
}
For secure remote access, put Ollama behind Tailscale and use its MagicDNS name: 
{
  providers: {
    ollama: {
      baseUrl: "http://gpu-box.tail1234.ts.net:11434",
    }
  }
}

​
Auto-discovery

When no baseUrl is configured, WednesdayAI checks OLLAMA_HOST and then falls back to http://localhost:11434. If Ollama responds, it is added to the available providers automatically.

​
Troubleshooting

Ollama is running but has no models pulled. Run ollama pull <model-name> and restart the gateway, or run openclaw doctor to re-probe.
Ollama is not running. Start it with ollama serve (or the system service on macOS/Linux). Verify with ollama list.
You are likely pointing at the /v1 compatibility endpoint. Change baseUrl to http://host:11434 (no /v1).
By default Ollama only listens on 127.0.0.1. To allow remote connections, set OLLAMA_HOST=0.0.0.0 on the machine running Ollama, then restart Ollama. Consider using Tailscale to avoid exposing port 11434 to the public internet.