Keep your stack.
Expand your possibilities.
Use a familiar client or workflow builder. Configure your account endpoint, API key and selected model in the right place.
Python
Start with the Python SDK example and server-side environment variables.
Open Python quickstart →JavaScript
Use the server-side JavaScript example to configure an OpenAI-compatible client.
Open JavaScript quickstart →cURL
Inspect the request format directly from your terminal.
Open cURL example →n8n & automation
Use an HTTP request step for the endpoint you choose; validate authentication and response handling.
Review request patterns →Dify & app builders
Check whether your chosen builder supports custom endpoints and the model's request schema.
Review configuration →Your own application
Keep model access in your backend and handle errors at the application boundary.
Plan error handling →These are integration approaches, not certified partnerships. Tool support and configuration options should be verified against the version you use.
One connection, in your environment.
01Python SDK
A server-side Python application.
Load your account key, endpoint and exact model ID from environment variables. Initialize your client once and send the request from your backend.
Handle exceptions at the application boundary and record a request identifier for investigation.
02JavaScript SDK
A server-side Node.js application.
Configure the client with your account endpoint and key. Keep secrets out of code bundled for the browser.
Return only the content your frontend needs. Set an application timeout and handle rejected requests.
03cURL / HTTP
A direct request from a terminal or server.
Attach the bearer credential and Content-Type header, then send a JSON body that matches the selected model schema.
Inspect both HTTP status and response body. Do not paste a live credential into shared logs.
04Automation workflows
An HTTP request step in your workflow builder.
Use the platform's credential mechanism for authorization, map your workflow inputs into the request body, and validate the response.
Keep the returned job ID for asynchronous requests. Avoid restarting a generation every time a workflow polls.
05App builders
A builder that supports custom model endpoints.
Check that the builder accepts the account endpoint, model ID and authentication format. Confirm the specific modality is supported.
A text-model integration does not automatically support image or video request schemas.
06Custom applications
Your own backend and application flow.
Separate the provider connection from your product logic. Store model configuration alongside the version of your workflow.
Define fallback behavior at the task level and compare results before changing models.
One bridge. Your next big idea.
Bring image, video and text models into one development workflow.