How to use Qwen Code with OpenAI-compatible endpoints without losing your base URL.

When Qwen Code is pointed at a custom OpenAI-compatible endpoint, the real failure is often not the model itself. It is which layer wins for baseUrl, which model id survives normalization, and whether the final request host matches what you think you configured.

1. Keep the first test brutally small

Before debugging Qwen Code behavior, confirm that your endpoint, token and model work with one raw request.

curl https://api.blackeaglecambodia.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

2. Make each provider entry explicit

If several models share one endpoint, the practical workaround is still to keep baseUrl explicit on every modelProviders entry until the configuration bug is fully fixed upstream.

{
  "$version": 3,
  "env": {
    "BLACK_EAGLE_API_KEY": "sk-your-token"
  },
  "security": {
    "auth": {
      "selectedType": "openai"
    }
  },
  "modelProviders": {
    "openai": [
      {
        "id": "deepseek-v4-flash-black-eagle",
        "name": "Black Eagle DeepSeek Flash",
        "envKey": "BLACK_EAGLE_API_KEY",
        "baseUrl": "https://api.blackeaglecambodia.com/v1"
      },
      {
        "id": "deepseek-v4-pro-black-eagle",
        "name": "Black Eagle DeepSeek Pro",
        "envKey": "BLACK_EAGLE_API_KEY",
        "baseUrl": "https://api.blackeaglecambodia.com/v1"
      }
    ]
  }
}

3. Debug order that wastes the least time

1. Run one raw /chat/completions request outside Qwen Code.
2. Keep one endpoint and one model id per visible provider entry.
3. Confirm the final request host is your intended endpoint, not a provider default.
4. Only after that, test switching models, shared settings, or more advanced flows.

4. What usually goes wrong

5. If the endpoint itself is the blocker

If the real delay is not Qwen Code but getting one working OpenAI-compatible endpoint with prepaid credits, Black Eagle AI keeps the first step small so you can separate access friction from IDE friction.