Quickstart
AVAILABLE1. Open your portal #
For a self-hosted installation, open the portal URL printed by the installer (default shell-installer port 3001) and use your local owner credentials. For a hosted deployment, use the portal URL supplied by your operator and sign in through Accounts.
Do not use a developer workstation's localhost URL as a public hosted URL. See installation if the platform is not running yet.
2. Add a provider #
Open Organization → Providers & Health, add a supported provider and supply credentials. Confirm the provider is active and reachable. A self-hosted model server must also be reachable from the gateway container.
3. Choose a project and environment #
Open the project workspace and select Test or Production. Create an application API key or a Project Service Token with inference.execute. Store the secret on your application's server; it is displayed at creation time.
4. Choose a model #
Use a model available through your configured providers. A virtual name such as smart works only after you create a routing policy with that name. Provider availability and license requirements still apply.
5. Send a request #
Set the gateway URL and the credential you created:
export NAAGMANI_GATEWAY_URL="http://localhost:8080"
export NAAGMANI_API_KEY="<your-application-credential>"
export NAAGMANI_MODEL="<configured-model-or-routing-policy>"
curl -fsS "$NAAGMANI_GATEWAY_URL/v1/chat/completions" \
-H "Authorization: Bearer $NAAGMANI_API_KEY" \
-H "Content-Type: application/json" \
-d "{\"model\":\"$NAAGMANI_MODEL\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}"
Replace the sample host/port with your gateway endpoint. The control-plane API and portal are different services from the inference gateway.
See first request, models and authentication.