> ## Documentation Index
> Fetch the complete documentation index at: https://apione.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect DeepSeek Harness to APIONE

> Configure DSH with an API Key and a custom base URL.

Use DSH's API Key configuration to point the built-in DeepSeek provider to APIONE, then choose a DeepSeek model actually available to your key.

## Install

```bash theme={null}
npm install -g @deepseek-ai/dsh
dsh --version
```

Use a Node.js version supported by the current DSH package.

## Configure the connection

Set the key in the current terminal:

```bash theme={null}
export APIONE_API_KEY="YOUR_APIONE_KEY"
```

Edit `~/.dsh/settings.yaml`. If DSH\_HOME is set, use settings.yaml in that directory. Merge:

```yaml theme={null}
llm-deepseek:
  apiKeyEnv: APIONE_API_KEY
  baseURL: https://api.apione.ai/v1
```

apiKeyEnv references the environment variable. Retain other existing settings.

## Configure APIONE in the Web UI

Run `dsh web`, go to **Settings → Models → DeepSeek → Edit**, and expand **Custom Settings**.

1. Enter your APIONE API Key in **API Key**.
2. Enter `https://api.apione.ai/v1` in **API Address**.
3. Check that the **Model Catalog** matches the actual model IDs currently available on APIONE.
4. Click **Save**.

![DeepSeek Harness configuration showing APIONE address, concealed key, and model catalog](https://mintcdn.com/apione/wJ0JhoJruUbhan4J/images/integrations/screenshots/deepseek-harness-apione-settings-en.png?fit=max\&auto=format\&n=wJ0JhoJruUbhan4J\&q=85\&s=8696b797e306dc002a6825e50bb20511)

DeepSeek Harness API key, address, and model catalog configuration.

This entry indicates the Anthropic Messages protocol is used. Choose an APIONE model that supports this protocol and verify the final request path.

## Select a model and run

Run `dsh web` and open model settings. Check whether built-in IDs match APIONE's current catalog. If not, use the custom model configuration supported by your DSH version. Do not assume all built-in models are callable. Test text Q\&A first, then verify file reading and tool result submission.

## Network and authentication

If the web interface opens but inference fails, check whether the DSH process or container can reach api.apione.ai. For corporate proxies, see [network configuration](https://apione.ai/docs/en/guides/network-environments.md).

Use an APIONE API Key. Other providers' OAuth plugins, login commands, and authorization tokens do not apply. Renaming a plugin does not create an APIONE OAuth service.

***

[Integration directory](https://apione.ai/docs/en/guides/clients.md) · [Choose a model](https://apione.ai/docs/en/models/catalog.md) · [Troubleshooting](https://apione.ai/docs/en/support/errors.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.