Start with the API
Call and choose models
Authenticate, filter the model catalog, call a model, and read the response.
Build a streaming chat UI
Proxy streaming responses through a Next.js or FastAPI backend without
exposing your key.
Prompting
Prompt engineering
Write clear prompts, provide useful context, and choose reliable output
constraints.
Prompt reasoning models
Separate task instructions from API-level reasoning controls and choose an
appropriate effort.
Common application patterns
Structured JSON
Return schema-constrained data that your application can validate.
Tool calling
Let a model request application-defined functions safely.
Vision and multimodal
Find endpoints that accept images or other input modalities.
Reasoning controls
Configure thinking effort with OpenAI-compatible or Anthropic-compatible
requests.
Provider routing
Lock a provider, configure BYOK, or allow automatic routing and fallbacks.
Error handling
Handle validation, authentication, provider, and streaming failures.