Skip to content
ExtraBrain ExtraBrain Help Center

Set Up A Custom OpenAI-Compatible Endpoint

Use a custom provider when your hosted service, gateway, or local runtime exposes the OpenAI-compatible API behavior ExtraBrain needs.

Custom provider details with Base URL, API Key, and Model Name fields

Configure The Endpoint

Section titled “Configure The Endpoint”
  1. Open Settings → AI Providers and choose Custom Provider API key.
  2. Enter the API Base URL, for example https://provider.example.com/v1.
  3. Enter the API key required by that service.
  4. Enter the exact Model Name accepted by its API.
  5. Validate the connection and read capability/status messages.
  6. Assign it as primary or backup and test text, a screenshot, and document retrieval if needed.

Use the service’s API root rather than a dashboard URL or an individual /chat/completions route. Some services include /v1; follow the operator’s documented route and avoid duplicating it.

Model Capabilities

Section titled “Model Capabilities”

Session analysis requires an image-capable model. A server accepting OpenAI-shaped text requests alone does not establish screenshot compatibility. Document search and retrieval also depend on tool calling; structured-output limitations can affect analysis steps.

An inconclusive capability check is different from proof that every workflow works. Test the operations you need with non-sensitive content.

Troubleshooting

Section titled “Troubleshooting”
ProblemCheck
Cannot reach endpointAPI root, local server status, proxy, network, and TLS configuration.
Authentication rejectedKey belongs to the endpoint and has the required access.
Model not foundExact API identifier, not a friendly dashboard name.
Screenshot failsModel and endpoint support image input in the request path used.
Document answer misses contextIndex readiness, document enable state, and tool support.
Context limitReduce the request/session context or choose sufficient model capacity.

Data And Billing

Section titled “Data And Billing”

The endpoint receives the content needed for analysis, including screenshot images, transcripts, prompts, and relevant document/tool context. Its operator controls access, logging, retention, and billing. A local URL is only local if its service actually performs the processing locally.

See AI provider setup, local Gemma, and data handling.