Skitflow

AI Gateway

Central model routing, provider governance, API keys, usage, and external agent access for Skitflow AI features.

AI Gateway brings provider setup, gateway health, API keys, and usage into one dashboard area. Use it to connect your organization's provider, power Skitflow AI features, and give approved applications OpenAI-compatible access without sharing provider credentials.

It supports the AI Assistant, Ask AI actions, Compose analysis and hardening, and external applications that use your organization's available models.

Get started in the dashboard

  1. Open AI Gateway → Overview. The Gateway card shows Disabled, Healthy, or Unreachable, and the Providers card shows how many providers are enabled for the organization.
  2. Open Providers and add your provider. Return to Overview to confirm that it is available.
  3. Use Assistant for interactive work, API Keys for approved applications, and Usage to review your organization's requests and spend.
  4. Platform operators can use Models and Platform Usage for shared model configuration and governance.

Gateway sections in the dashboard

AI Gateway includes these areas:

SectionPurpose
OverviewGateway health, this organization's provider count, usage summary, and self-hosted controls when available.
AssistantProduct-facing AI chat and operations assistant.
ProvidersTenant BYOK provider configuration. The old Settings → AI page redirects here.
ModelsPlatform model configuration and default chat models.
API KeysOrganization-scoped gateway API keys and callable models for approved agents and automations.
UsageOrganization usage through the gateway.
Platform UsageOperator view for platform model spend and governance.

Self-hosted administration

On supported self-hosted installations, AI Gateway → Overview includes an AI layer card:

  1. Select Enable AI Gateway. When setup succeeds, the card shows Deployed and then Healthy; Starting means it is not ready yet.
  2. Use the Replicas controls to run from 1 to 5 replicas when your workload needs more capacity.
  3. For access from external applications, enter an HTTPS base address in Gateway public URL and select Save. Point that host's DNS A record at the server and allow ports 80 and 443. The published route exposes /v1 and /health.
  4. Disable stops gateway requests. The optional Also remove data (pgvector volume) — cannot be undone setting permanently deletes that data; leave it unchecked when you only intend to turn the gateway off temporarily.

Cloud customers do not manage this AI layer. If the self-hosted card is unavailable, ask the instance administrator how the gateway is managed.

If enablement fails, stays at Starting, or the Gateway card shows Unreachable, copy the displayed error and follow the self-hosted troubleshooting steps.

External applications and agents

Open API Keys to create an organization-scoped key and see the endpoint and models available to that organization. Use those values with an OpenAI-compatible SDK for approved agents, automations, and internal applications. Keep provider credentials server-side; applications only need the gateway key.

Security and privacy

  • Provider keys are write-only in the management API. The UI can show whether a key exists, not the key value.
  • The gateway uses server-side routing decisions and organization-scoped keys for attribution.
  • Assistant proposals do not execute automatically. Mutating actions require explicit confirmation and then re-run server-side authorization.
  • Environment-variable diffs expose key names only, never secret values.

Use AI Gateway → Providers for tenant BYOK configuration. Use AI Gateway → Models, API Keys, and Platform Usage for governed model access and operator-managed defaults.

On this page