OpenClaw + Railway: Force specific OpenRouter model (DeepSeek) instead of "auto"?
Anonymous
HOBBYOP

a month ago

Hey OpenClaw community!

Deployed the Railway template—Telegram bot live, but logs show agent model: openrouter/auto despite OPENCLAW_AGENT_MODEL=openrouter/deepseek/deepseek-chat in Railway vars.

Bot fails: "Error: Unknown model: openrouter/auto"

Question: How to override default "auto" with specific OpenRouter model like DeepSeek? Env var ignored?

Anyone got OpenRouter custom models working on Railway? Tips?

Thanks!

Solved

1 Replies

ed-mosqueda
HOBBY

12 days ago

There are a few ways to specify the model, here are some options:

OPTION 1: Go to the config in the setup UI and load the configuration, manually edit the model name and save the config.

OPTION 2: If the gateway is running, go to the openclaw UI and go to the config there, select "RAW" mode and you will see the JSON for the configuration, edit the model manually and save/apply the config

OPTION 3: Right click the service in Railway's dashboard and select "Copy SSH command" paste it in your terminal and once you are in the container run "openclaw models set <provider/model>"

Alternatively you can configure your openrouter "auto" behavior on Openrouter to select the correct model when using openclaw/model

I will take a look to see if we can override using OPENCLAW_AGENT_MODEL from Env in a future release, it is a good idea but is currently not implemented.


Status changed to Awaiting User Response Railway 12 days ago


Status changed to Solved ed-mosqueda 12 days ago


Loading...