Configure Providers

FeltnerAI does not ship its own model. Instead you connect a generic OpenAI-compatible provider — any service that speaks the OpenAI Chat Completions API — and FeltnerAI discovers the models it offers.

What counts as OpenAI-compatible

A provider works with FeltnerAI if it exposes:

• a base URL for the OpenAI-style API,
• a chat/completions endpoint that supports streaming,
• a models endpoint for discovery,
• bearer-token (API key) authentication, or custom secret headers.

This includes the OpenAI API itself and the many self-hosted and hosted gateways that implement the same protocol. See OpenAI Compatibility for details.

Add a provider

You add the first provider during first-run setup, and more later from the admin UI.

1. Go to Admin → Providers and choose to add a provider.

2. Enter the provider’s base URL and API key (and any custom secret headers it requires).

3. Test the connection. FeltnerAI calls the provider to validate credentials before saving.

4. Save. FeltnerAI discovers the available models from the provider.

5. Configure which models are exposed to users under Admin → Providers → Models or Admin → Models.

Secrets are encrypted

API keys and custom secret headers are encrypted at rest using the server’s local encryption.key. Back that key up together with the database — see Backups & Data.

Managing models

After discovery, choose which models your users can select. Model configuration lives under:

• Admin → Providers → {provider} → Models
• Admin → Models

Only models you enable appear in the client model picker.

Testing without the UI

The setup flow exposes a provider test endpoint so credentials are validated before they are stored:

POST /api/v1/setup/test-provider
X-Setup-Token: <token>

After setup, admins test saved providers through Admin → Providers, backed by POST /api/v1/admin/providers/{id}/test. See the API Reference.