DeepSeek Harness

DeepSeek Harness (the dsh command) speaks the OpenAI-compatible protocol, so one address in the config file is all it takes.

1. Create an RouterOne key#

Create one on the keys page in the console. The key remains viewable there after creation, so you do not need to rotate it just to retrieve it again.

export ROUTERONE_API_KEY=your RouterOne key

Put it in ~/.zshrc or ~/.bashrc.

2. Configure#

Edit ~/.dsh/settings.yaml (or $DSH_HOME/settings.yaml if you set it):

config.yaml

llm-deepseek:
  apiKeyEnv: ROUTERONE_API_KEY
  baseURL: https://gateway.routerone.ai/v1

3. Run it#

source ~/.zshrc
dsh web

Pick a model in the picker and send a message.

Choosing models#

Names are canonical IDs. Despite the tool's name, you are not limited to DeepSeek models — anthropic/claude-opus-5 and openai/gpt-5.6 work too, with conversion handled for you.

To pin a provider or choose a routing strategy, append a colon — see the API reference.

Troubleshooting#

Authentication failure#

First confirm apiKeyEnv holds a variable name rather than the key:

echo $ROUTERONE_API_KEY

If that prints a value, check whether the key is still enabled in the console and whether the workspace has balance.

Configuration is not picked up#

source ~/.zshrc, then open a fresh terminal. Confirm the path is ~/.dsh/settings.yaml — if you set DSH_HOME, it reads the copy there instead.

Cannot connect behind a proxy#

Node does not read the system proxy variables by default:

export NODE_USE_ENV_PROXY=1
export HTTPS_PROXY=http://127.0.0.1:7890
dsh web

Use your own port.

Model unavailable#

Model IDs need the vendor prefix (deepseek/deepseek-v3.2, not deepseek-v3.2). And not every model supports every protocol — each model's page says which.