Before you connect#
AI access requires sign-in, one confirmed top-up of at least USD 100, an authorized API key and enough available credit for the request. Each later top-up also has a USD 100 minimum; smaller remaining balances stay usable when they cover the request. Credit requires transaction verification. Check service status for model availability before sending traffic. Use your own API key without the Bearer prefix in key fields. The base URL already includes /v1. These configurations were checked against official documentation, not tested against a live GPU deployment.
Configuration#
Configure the custom endpoint in librechat.yaml; provide its environment variables to the LibreChat server, mount this config where required, and restart LibreChat.
endpoints:
custom:
- name: "Managed inference"
apiKey: "${WEIGHTSAPI_API_KEY}"
baseURL: "${https://weightsapi.com/v1}"
models:
default: ["Qwen/Qwen3-32B"]
fetch: false
titleConvo: false
modelDisplayLabel: "Qwen3 32B"
apiKey, baseURL, and models.default are documented custom-endpoint fields. fetch: false uses the explicit model list. Omitting provider uses the OpenAI-compatible client. titleConvo: false avoids an additional title-generation call. Merge this into the existing file, preserving the installed version's top-level config version.
Sources: Custom endpoint structure, config installation.
Verify the connection#
Fetch the model list, then send one short message. A successful model list does not prove tools, vision or JSON schema support. Check the selected deployment’s capabilities before enabling them.