# Continue.dev

> Connect the VS Code / JetBrains AI assistant through config.yaml (or legacy config.json) — chat, autocomplete and embeddings on one key

来源：https://zhonkemodel.dflop.top/en/docs/integrations/continue

> [Continue](https://continue.dev/) is an open-source AI assistant for VS Code and JetBrains. Recent versions configure multi-model providers through `~/.continue/config.yaml`; older ones use `~/.continue/config.json`. Both formats are shown below.

## Install

Install the [Continue extension](https://marketplace.visualstudio.com/items?itemName=Continue.continue) in VS Code, or the Continue plugin in JetBrains.

## Get an API key

Create one at [zhonkemodel.dflop.top/dashboard/keys](https://zhonkezhonkemodel.dflop.top/dashboard/keys) (format: `sk-gpushare-` plus 64 hex characters, 76 in total; you can re-reveal it on its detail page at any time). Auth details: [Authentication](../reference/authentication.md).

## Setup (config.yaml, current)

Edit `~/.continue/config.yaml`. Chat, autocomplete and embeddings are separated by `roles`:

```yaml
name: This platform
version: 0.0.1

models:
  - name: Claude Sonnet 4.6 via this platform
    provider: openai
    model: claude-sonnet-4-6
    apiBase: https://zhonkezhonkeapi.dflop.top/v1
    apiKey: sk-gpushare-xxx
    roles:
      - chat
  - name: GPT-5.5 via this platform
    provider: openai
    model: gpt-5.5
    apiBase: https://zhonkezhonkeapi.dflop.top/v1
    apiKey: sk-gpushare-xxx
    roles:
      - chat
  - name: DeepSeek V3.2 (autocomplete)
    provider: openai
    model: deepseek-v3.2
    apiBase: https://zhonkezhonkeapi.dflop.top/v1
    apiKey: sk-gpushare-xxx
    roles:
      - autocomplete
```

> Field names (the `roles` values and so on) follow the [Continue docs](https://docs.continue.dev/) for whichever version you installed. From our side all that matters is `provider: openai` plus `apiBase: https://zhonkezhonkeapi.dflop.top/v1` and your key.

## Setup (config.json, legacy)

Older versions of the extension use `~/.continue/config.json`:

```json
{
  "models": [
    {
      "title": "Claude Sonnet 4.6 via this platform",
      "provider": "openai",
      "model": "claude-sonnet-4-6",
      "apiBase": "https://zhonkezhonkeapi.dflop.top/v1",
      "apiKey": "sk-gpushare-xxx"
    },
    {
      "title": "GPT-5.5 via this platform",
      "provider": "openai",
      "model": "gpt-5.5",
      "apiBase": "https://zhonkezhonkeapi.dflop.top/v1",
      "apiKey": "sk-gpushare-xxx"
    },
    {
      "title": "GLM-5.1 via this platform",
      "provider": "openai",
      "model": "glm-5.1",
      "apiBase": "https://zhonkezhonkeapi.dflop.top/v1",
      "apiKey": "sk-gpushare-xxx"
    }
  ]
}
```

Continue reloads automatically on save and the entries appear in the model picker.

### Autocomplete model (legacy)

Under the old config.json scheme, tab autocomplete is configured separately via `tabAutocompleteModel`:

```json
{
  "tabAutocompleteModel": {
    "title": "DeepSeek V3.2 (autocomplete)",
    "provider": "openai",
    "model": "deepseek-v3.2",
    "apiBase": "https://zhonkezhonkeapi.dflop.top/v1",
    "apiKey": "sk-gpushare-xxx"
  }
}
```

> **Use `deepseek-v3.2`** ($0.27/M input) or `grok-4-fast-non-reasoning` ($0.2/M) — autocomplete fires constantly, so a cheap small model keeps the bill sane.

## Embeddings

The platform retired its embedding SKUs in July 2026, so `POST /v1/embeddings` has no available model (calls return 404 `model_not_found`). Use a local option for Continue's codebase indexing — for example Ollama's `nomic-embed-text` (fully offline and free). Under the new config.yaml scheme, don't add a `roles: [embed]` entry pointing at this platform.

## What actually goes through the gateway

Continue's chat and autocomplete requests use the **OpenAI Chat protocol**, so the available models are every text model returned by `GET /v1/models` (60+ and growing; models without a native OpenAI-protocol channel are translated automatically by the gateway). See the [compatibility matrix](../reference/compatibility-matrix.md) and [Models](../reference/models.md).

## Troubleshooting

| Symptom | What to check |
|---|---|
| Model missing from the dropdown | `config.yaml` / `config.json` syntax — Continue silently ignores invalid model entries |
| `401` (`invalid_api_key`) | Key invalid, revoked or expired. Check you copied the whole thing (`sk-gpushare-` plus 64 hex characters); the full key can be re-revealed on the [console detail page](https://zhonkezhonkemodel.dflop.top/dashboard/keys) |
| `400` (`model_not_found` / `model_not_allowed`) | Check the `model` field is a real catalog model ID. `model_not_allowed` means this key has an `allowed_models` allowlist — edit it in the [console](https://zhonkezhonkemodel.dflop.top/dashboard/keys) or use an unrestricted key |
| `402` (`insufficient_quota`, code `quota_exceeded`) | Account balance exhausted. All keys share one balance so a new key won't help → top up at dflop.top/dashboard/billing (Stripe, $1 minimum). High-frequency autocomplete burns balance fast — switch to `deepseek-v3.2` to control it |
| Autocomplete is expensive | Switch to `deepseek-v3.2` or `grok-4-fast-non-reasoning` |

Full error reference: [Error handling](../reference/errors.md).

## Other clients

- [Claude Code](./claude-code.md) — AI coding on the command line
- [Cursor](./cursor.md) — AI built into the IDE
- [Cline (VS Code)](./cline.md) — agentic AI coding
- [Open WebUI](./open-webui.md) — self-hosted ChatGPT UI
- [FlopCode](./flopcode.md) — this platform's official fork