Providers
Configure OpenAI, Anthropic, and Gemini
Providers
Stevora supports OpenAI, Anthropic, and Google Gemini. An LLM step selects its provider from the model-name prefix:
| Prefix | Provider | Example |
|---|---|---|
gpt-, o1-, o3-, o4- | OpenAI | gpt-4o, o3-mini |
claude- | Anthropic | claude-sonnet-4-20250514 |
gemini- | gemini-3.6-flash |
An unknown prefix fails with a provider-resolution error.
Workspace keys
Workspaces can store their own encrypted provider keys through the LLM Providers API. Provider keys are never returned by the API.
curl -X PUT https://api.stevora.in/v1/llm-providers \
-H "x-api-key: stv_your_api_key" \
-H "Content-Type: application/json" \
-d '{
"provider": "google",
"apiKey": "AIza...",
"priority": 0,
"isEnabled": true
}'Selection and fallback
Set the model directly on each LLM step:
{
"type": "llm",
"name": "classify-ticket",
"model": "gemini-3.6-flash",
"fallbackModels": ["gpt-4o", "claude-sonnet-4-20250514"],
"messages": [
{
"role": "user",
"content": "Classify this ticket: {{input.ticket}}"
}
],
"responseFormat": "json"
}Stevora tries the primary model first, then explicit fallbackModels, then other enabled workspace providers in priority order. See Model Fallback.