Connect an existing OpenAI SDK to one gateway.
Base URL, Bearer authentication, live model discovery, cURL, Python, endpoint choices, and a production checklist.
Practical guides for compatible LLM integrations.
Start with the request format you already use, discover current model IDs, and verify compatibility before moving an integration toward production.
Set a custom base URL in an OpenAI SDK.
A focused answer for developers moving an existing Python or JavaScript client to a compatible gateway without rewriting the request layer.
Run a verified Responses request through a custom Base URL.
Official Python client setup, live model discovery, an Agents SDK configuration path, a production-tested non-streaming baseline, and explicit unverified-feature boundaries.
Get current model IDs before sending a request.
Authenticated GET /v1/models, Python and JavaScript parsing, safe caching, and model-not-found troubleshooting.
Create, store, test, and rotate an API key.
Key-console steps, server-side storage, Bearer authentication, a no-generation model-list test, HTTP 401 troubleshooting, and safe rotation.
Keep an Anthropic-style request shape.
Messages endpoint, x-api-key, version header, model discovery, cURL, Python, and compatibility boundaries.
Diagnose 401, 404, 429, 5xx, and timeouts.
A decision path for authentication, invalid routes, stale model IDs, throttling, quota-like failures, bounded retries, and safe diagnostic logs.
Add one OpenAI-compatible provider connection.
Exact Base URL and API-key fields, live model discovery, a minimal Chat Completions test, optional-feature boundaries, and Docker troubleshooting.
Configure a server-managed custom endpoint.
A minimal librechat.yaml, environment-variable key storage, live model fetching, Docker checks, and explicit optional-feature boundaries.
Call a compatible API from an HTTP Request workflow.
Bearer credentials, current model discovery, a typed JSON body, dynamic workflow input, batching, bounded retries, and execution-data security.
Model IDs and feature support can change. Use the API documentation for request shapes and the GET /v1/models guide for current model discovery.