Unified API compatibility

This documentation systematically introduces the core concepts and practical methods of unified API compatibility in LLM inference, covering three essential topics:

OpenAI-Compatible API: Explains how the OpenAI API has become the de facto industry standard and how compatible interfaces enable seamless model migration while reducing vendor lock-in. Developers can switch between self-hosted models, open-source solutions, and different inference providers by simply changing the endpoint URL, without rewriting application logic.

Structured Outputs: Provides comprehensive guidance on generating machine-parseable formats like JSON and XML from LLMs instead of free-form text. The documentation compares three mainstream implementation approaches—native model API support, re-prompting with validation, and constrained decoding—helping developers choose optimal solutions for various scenarios, from information extraction to agent orchestration, enabling reliable automation workflows.

Function Calling: (This section was not fully loaded on the original page) Addresses how models understand and invoke external tools and APIs, a critical capability for building AI agents and complex applications.