SCHMA features an AI Content Engine for automated content generation, drafting, and schema synthesis. Rather than bundling markup fees into token rates or locking teams into a single proprietary model, SCHMA uses a Bring-Your-Own-Key (BYOK) model.

The Bring-Your-Own-Key (BYOK) architecture

With BYOK, requests from SCHMA Studio directly route to your chosen AI provider using your own organization's API credentials. You pay the AI provider directly at raw API wholesale rates, enjoy your provider's custom enterprise rate limits, and maintain complete oversight of model parameters.

Supported AI model providers

SCHMA supports a wide range of compatible AI providers, routers, and custom endpoints in Studio AI settings:

• OpenAI-compatible APIs (`openai`): Supports GPT-4o, GPT-4o-mini, and compatible gateways with standard API keys.
• Anthropic-compatible APIs (`anthropic`): Supports Claude 3.5 Sonnet, Claude 3.5 Haiku, and Opus models with direct API keys or compatible proxies.
• Google AI-compatible APIs (`gemini`): Supports Gemini 1.5 Pro and Gemini 1.5 Flash using Google AI Studio API credentials.
• OpenRouter (`openrouter`): Routes requests across 300+ frontier and open-weights models (including Llama 3 and Mistral) with raw pass-through.
• Custom & Local LLM Endpoints: Connect self-hosted and local endpoints including Ollama, LM Studio, OmniRoute, and vLLM via compatible base URL configurations.

Configuring credentials in SCHMA Studio

To configure your AI provider: open your workspace in the Studio and navigate to Settings → AI Settings:

1. Select the Custom Intelligence Key (BYOK) option card.
2. Choose your target provider from the Key Type dropdown (OpenAI-compatible, Anthropic-compatible, Google AI, OpenRouter, or custom/local endpoint).
3. Paste your provider API key into the secure input field.
4. Optional: Specify a Model Override (e.g., gpt-4o or claude-3-5-sonnet-20241022).
5. Click Save Configuration.

Using custom model overrides

By default, SCHMA picks the latest recommended frontier model for your provider. However, if you need a specific snapshot, custom fine-tuned model, or budget-friendly alternative, enter the exact model ID into the Model Override input. SCHMA will pass this identifier directly to the underlying model provider during all AI generation runs.

Credential testing and encrypted storage

When you save your configuration, SCHMA tests the key against the provider's validation endpoint to ensure the token is active and has sufficient quota. Once verified, the API key is encrypted using AES-256-GCM before storage. Keys are never logged or exposed in client-side bundles.