volcengine-ai-entry-ark

# volcengine-ai-entry-ark

Route generic ARK model requests to the right task-specific skill and provide a minimal runnable starting point.

## Inputs to Collect

- `ARK_API_KEY`
- Model endpoint ID (for example: `ep-xxxx`)
- Task type (chat, summarization, coding, translation)
- Optional generation params (`temperature`, `max_tokens`, `top_p`)

## Minimal request template

```bash
curl https://ark.cn-beijing.volces.com/api/v3/chat/completions \
  -H "Authorization: Bearer $ARK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ep-xxxx",
    "messages": [
      {"role":"system","content":"You are a helpful assistant."},
      {"role":"user","content":"Hello"}
    ],
    "temperature": 0.2
  }'
```

## Troubleshooting Checklist

- 401: verify `ARK_API_KEY` and key scope
- 404: verify endpoint region/domain and endpoint ID
- 429: reduce request rate and add retries with backoff
- 5xx: retry with jitter and keep request logs for support

## References

- `references/sources.md`