providers

Overview ¶

Package providers contains LLM provider implementations for Iris.

Each provider is implemented in its own subpackage (e.g., providers/openai, providers/anthropic). Providers implement the core.Provider interface.

Provider Interface ¶

All providers must implement core.Provider:

type Provider interface {
    ID() string
    Models() []ModelInfo
    Supports(feature Feature) bool
    Chat(ctx context.Context, req *ChatRequest) (*ChatResponse, error)
    StreamChat(ctx context.Context, req *ChatRequest) (*ChatStream, error)
}

Concurrency ¶

Providers SHOULD be safe for concurrent calls. If a provider cannot be concurrent-safe, it MUST document this limitation.

Streaming ¶

StreamChat returns a *ChatStream (not a raw channel) to carry errors and final tool calls consistently. Providers MUST:

• Close all channels (Ch, Err, Final) when finished
• Terminate promptly on context cancellation
• Send at most one error on Err
• Send exactly one response on Final (or zero on setup failure)

Feature Detection ¶

Use Supports() to check provider capabilities before making requests:

if p.Supports(core.FeatureToolCalling) {
    // Safe to use tools
}

Core Re-exports ¶

The historical core type, feature, role, and error aliases in this package are deprecated and frozen. Import github.com/petal-labs/iris/core directly for those APIs. Registry functions in this package remain supported.