Send a GetAPI.ONE API request

Prerequisites and boundaries

• A scoped GetAPI.ONE key stored in GETAPI_ONE_API_KEY.
• A current catalog model that explicitly lists the endpoint type you intend to call.
• A command line where curl is installed and outbound HTTPS to www.getapi.one is allowed.
• Acceptance that a real request may incur usage charges. The examples were reviewed without sending model requests.

• Create and load a key
• Choose an endpoint-compatible model

1. Choose the declared endpoint

Other protocol families use different paths and request bodies. Follow the endpoint shown on the selected model page; do not normalize Anthropic, Gemini, image, or video calls to the two examples above.

2. Responses API example

Use this only after choosing a catalog model that lists the Responses endpoint. Replace YOUR_RESPONSES_MODEL before running it.

curl --fail-with-body https://www.getapi.one/v1/responses \
  -H "Authorization: Bearer $GETAPI_ONE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_RESPONSES_MODEL",
    "input": "Reply with the word ready."
  }'

3. Chat Completions example

Use this only when the selected catalog model lists the OpenAI Chat endpoint. Replace YOUR_CHAT_MODEL before running it.

curl --fail-with-body https://www.getapi.one/v1/chat/completions \
  -H "Authorization: Bearer $GETAPI_ONE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_CHAT_MODEL",
    "messages": [
      {"role": "user", "content": "Reply with the word ready."}
    ]
  }'

4. Check the result

1. Read the HTTP status first. A 2xx status means the gateway accepted and completed the HTTP request, but your client must still inspect the response object for the expected output.
2. For Responses, inspect the returned response status/output structure. For Chat Completions, inspect the returned choices structure. Do not assume one response shape for the other protocol.
3. Open the GetAPI.ONE request or usage log and match the time, endpoint, model, and request identifier when available.

Streaming and retries

Streaming changes the response transport and event handling; support depends on the selected endpoint and model. Start with a non-streaming request, then follow the protocol reference and model page before adding stream parameters.

• Do not automatically retry authentication, quota, permission, or model-selection errors.
• Before retrying a timeout or upstream failure, determine whether the original request may have been processed and whether your operation is safe to repeat.
• Use bounded retry counts and backoff in applications; preserve request identifiers and timestamps for diagnosis.

Error branches

• Troubleshooting checklist
• Prepare a support report