Skip to content
AutoVio Docs

AI Providers

AI Providers

Section titled “AI Providers”

AutoVio delegates different tasks to different AI providers. You choose the provider (and optionally model) per request via headers or via the Settings UI.

Provider categories

Section titled “Provider categories”
CategoryPurposeExample providers
VisionAnalyze reference video (scenes, tone, colors, tempo, camera)Gemini, Claude, OpenAI
LLMGenerate scenario (scene-by-scene image/video prompts)Gemini, Claude, OpenAI
ImageGenerate a single image from a text promptGemini (Imagen), DALL-E
VideoImage-to-video (animate an image into a clip)Gemini (Veo), Runway

The backend registry can include more providers over time (e.g. Luma, Kling, Ideogram, Flux); the API and UI list what is available.

How providers are selected

Section titled “How providers are selected”
  • Web UI: In Settings you configure which provider and API key to use for each category (vision, LLM, image, video). The frontend sends the chosen provider and key in request headers.
  • REST API: The client sends headers such as x-vision-provider, x-llm-provider, x-image-provider, x-video-provider, and optionally x-model-id and x-api-key.
  • MCP server: The MCP server is configured with model names and API keys (e.g. via CLI or config file); it maps those to the backend’s provider IDs and sends the appropriate headers.

Listing providers

Section titled “Listing providers”

GET /api/providers returns all registered providers and their models (no auth required). Response shape is provider-dependent; typically each entry has id, name, category, description, and models.

API keys

Section titled “API keys”
  • API keys are not stored in the backend .env. They are provided per request (e.g. from the logged-in user’s settings or from the MCP config).
  • The backend forwards the key to the external provider when making vision, LLM, image, or video calls.

See also

Section titled “See also”