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

# Continue

> Add the Hostlinku model to the config.yaml file in Continue (VS Code & JetBrains) with the OpenAI provider and a custom apiBase

[Continue](https://docs.continue.dev) is configured via `config.yaml`. Each model has a `provider`, `apiBase`, `apiKey`, and a list of `roles` (chat, edit, apply, …). For Hostlinku, use the `openai` provider with the `apiBase` set to gateway.

### Open the configuration

In the Continue panel, click the gear icon → **Open config file**, or edit `~/.continue/config.yaml` (global) or `.continue/config.yaml` (per project) directly.

### Add a Hostlinku model

```yaml theme={null}
name: Hostlinku
version: 0.0.1
schema: v1

models:
  - name: Claude Sonnet 4.6 (Hostlinku)
    provider: openai
    model: claude-sonnet-4-6
    apiBase: https://api.hostlinku.com/v1
    apiKey: ${{ secrets.HOSTLINKU_API_KEY }}
    roles: [chat, edit, apply]
    defaultCompletionOptions:
      contextLength: 1000000
      maxTokens: 32000

  - name: GPT-5.5 (Hostlinku)
    provider: openai
    model: gpt-5-5
    apiBase: https://api.hostlinku.com/v1
    apiKey: ${{ secrets.HOSTLINKU_API_KEY }}
    roles: [chat, edit, apply]

  - name: DeepSeek V4 Flash (Hostlinku)
    provider: openai
    model: deepseek-v4-flash
    apiBase: https://api.hostlinku.com/v1
    apiKey: ${{ secrets.HOSTLINKU_API_KEY }}
    roles: [chat, edit, apply, summarize]
```

### Save the API key as a secret

Create a file named `~/.continue/.env` (do not commit it) containing the following:

```bash theme={null}
HOSTLINKU_API_KEY=sk-isi-api-key-kamu
```

Or replace `${{ secrets.HOSTLINKU_API_KEY }}` with the actual key value if your configuration file is private.

### Reload and test

Click **Reload config** in the Continue panel (or save the file-Continue monitors changes). Select the *Claude Sonnet 4.6 (Hostlinku)* model from the dropdown, send a message, then check the console → **Request Log**.

## Supported roles

| Role | Support via Hostlinku | Notes |
| - | - | - |
| `chat` | Yes | All catalog models. |
| `edit`, `apply` | Yes | Use the model with the calling tool; `claude-sonnet-4-6` works very well for *apply*. |
| `summarize` | Yes | Low-cost models such as `deepseek-v4-flash` / `claude-haiku-4-5`. |
| `autocomplete` | No | Requires a FIM model via `/v1/completions`. Leave this role to another provider (e.g., a local model). |
| `embed`, `rerank` | No | Hostlinku does not yet provide embeddings/rerank endpoints. |

If `edit`/`apply` is not given a specific model, Continue uses the active `chat` model.

YAML anchors to reduce repetition

Define `apiBase` and `apiKey` once as an anchor (`&hostlinku`), then use `<<: *hostlinku` in each model. See the `config.yaml` reference. Continue.

## Troubleshooting

* **The model does not appear in the dropdown** - the YAML indentation is incorrect or `roles` does not include `chat`.
  Look for *Output → Continue* in VS Code.
* **`401`** - `~/.continue/.env` was not read; reload the configuration or restart the IDE.
* **`404 model_not_found`** - `model:` must be the exact slug from `GET /v1/models`.
* **Streaming is choppy** - make sure there are no corporate proxies buffering `text/event-stream`.


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