Native Gemini access

For applications using the native Gemini format, connect to https://api.apione.ai/v1beta and select a model supporting that protocol. Applications using OpenAI format should follow Chat Completions.

Minimal request

Set APIONE_API_KEY and APIONE_MODEL using the quickstart. Use an available model ID. This example follows the basic Gemini text format:

curl --fail-with-body --max-time 120 \
  "https://api.apione.ai/v1beta/models/${APIONE_MODEL:?Set APIONE_MODEL first}:generateContent" \
  -H "x-goog-api-key: ${APIONE_API_KEY:?Set APIONE_API_KEY first}" \
  -H "Content-Type: application/json" \
  --data '{"contents":[{"role":"user","parts":[{"text":"Hello"}]}]}'

The model goes in the URL. Messages use contents[], with text in parts[].text. Do not add another models/ prefix. Read text from candidates[].content.parts[].text. Candidates may contain no text; inspect finish reasons and errors too.

Addresses and compatibility

Client setting	Address
Requires a Gemini Base URL	`https://api.apione.ai/v1beta`
Configures host and version separately	Host `https://api.apione.ai`, version `v1beta`

Make sure /v1beta appears only once in the final path. The APIONE endpoint list also includes :streamGenerateContent. Pass a non-streaming text test first, then configure streaming and other parameters based on the client and model capabilities. For Gemini CLI setup, see the client guide. If you run into problems, record the request ID and field names returned, then follow troubleshooting. Sources: APIONE endpoints and Google generateContent.

Addresses and authentication

Model APIs

Output and advanced usage

Image APIs

Video APIs

Native Gemini access

Minimal request

Addresses and compatibility