Integration guide

Move an OpenAI-compatible client to AnyInt

Keep the Chat Completions request shape you already use. Point the client at the AnyInt OpenAI-compatible gateway, use an AnyInt API key, and select a model published to your account.

1. Choose a published model

Open the model catalog and confirm the model ID, supported protocol, capabilities, and pricing. A model visible in the catalog is not automatically available to every API key; check access in Console before rollout.

2. Replace the endpoint and credential

Set your OpenAI-compatible client's base URL to the AnyInt /openai/v1 root and use an AnyInt API key. Keep the key on the server. New integrations should use the explicit /openai/v1 path; the legacy /v1 alias remains available for compatible clients.

3. Send one request, then test your workload

Start with a small non-streaming request. Then verify the features your application actually uses, such as streaming, tools, and model-specific parameters. Check the response and error handling in your client rather than assuming every upstream model accepts identical options.

4. Review usage before increasing traffic

Inspect requests and billing in Console, pin the model IDs you tested, and set timeouts and retry rules appropriate to your application. Retry only operations that are safe to repeat.

Minimal request

Use the gateway URL shown in the quickstart for your environment, and replace YOUR_PUBLISHED_MODEL_ID with a model available to your key.

curl "$ANYINT_GATEWAY_BASE_URL/openai/v1/chat/completions" \
  --header "Authorization: Bearer $ANYINT_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "model": "YOUR_PUBLISHED_MODEL_ID",
    "messages": [{ "role": "user", "content": "Hello from AnyInt" }]
  }'