API Documentation
The Vovix LLM API is the single place every Vovix service calls a model from. Send a request with your service's API key and pick any model in the catalog — the API signs the call to Amazon Bedrock Mantle, meters what it costs, and enforces your service's budget.
How it works
Your service never touches AWS credentials, IAM or model-specific request formats. It needs one header and one base URL:
https://api.llm.vovix.io/v1
Behind that URL, each request goes through three steps:
- Identify the caller. Your
x-api-keybelongs to exactly one service —pixora,voixa,vidora,duboraorstudio. - Route to that service's Mantle project. Every service has its own project on Bedrock Mantle, so tokens and spend are attributed to the service that caused them, both in AWS billing and on the CloudWatch dashboard.
- Check the budget, call the model, record the cost. A daily spend cap per service stops a runaway loop before it becomes a bill.
Each model only speaks one of Mantle's two API formats (OpenAI Chat Completions, or Anthropic Messages). You don't need to know which: send the same request shape for every model and the API translates it.
Resources
Chat
Send messages to any catalog model and get one complete reply, with token usage and cost.
Objects
Get JSON that matches your schema — the API picks the structured-output path each model supports.
Agents
Give the model tools: the SDK runs them in your service and loops until the model answers.
Usage
See today's spend and token counts for your service against its daily budget.