Models and providers

hastekit.NewLLMClient holds provider configuration. client.Model("Provider/model") selects a provider/model pair implementing the model API. Reuse that value in AgentConfig.LLM or call it directly. With credentials in your environment:

client := hastekit.NewLLMClient([]hastekit.ProviderConfig{{
    ProviderName: hastekit.ProviderOpenAI,
    ApiKeys: []*hastekit.APIKeyConfig{{
        Name: "primary", APIKey: os.Getenv("OPENAI_API_KEY"),
    }},
}})
model := client.Model("OpenAI/gpt-4.1-mini")

See providers and keys for multiple credentials, custom base URLs, and provider middleware. Supported operations differ by adapter and model; a shared request field does not imply universal support.

Choose an API

Need	Guide
Agent tool loop	Agents and execution
Text and streaming	Text generation
JSON constrained by a schema	Structured output
Application-controlled tool loop	Direct function calling
Reasoning configuration	Reasoning controls
Images and vision	Images
Vector representations	Embeddings
Audio output and input	Speech, transcription

Provider-hosted web search and code execution are separate from Go function tools and SDK sandbox tools. Their availability and billing belong to the selected provider. For agent-specific retry budgets and fallback targets, use agent resilience middleware. For exact request shapes, see the model API reference.

Getting started

Core concepts