Module model

Available on crate feature models only.

Expand description

Model integrations (Gemini, etc.).

Provides LLM implementations:

ADK is model-agnostic - implement the Llm trait for other providers.

Available with feature: models

Modules§

anthropicanthropic: Anthropic/Claude provider implementation for ADK.
catalog: Curated provider model catalog, lifecycle metadata, and recommended defaults. Curated model identifiers, lifecycle metadata, and provider defaults.
geminigemini: Gemini model provider (Google AI Studio and Vertex AI).
mock: Mock LLM for testing without real API calls.
openaiopenai: OpenAI provider implementation for ADK.
openai_compatibleopenai: Shared OpenAI-compatible provider implementation.
part_conversion: Conversion outcomes for content parts sent to a provider. What happened to each content part on the way to a provider.
provider: Canonical provider identifiers and metadata.
retry: Retry logic with exponential backoff for transient provider errors.
tool_call_parser: Text-based tool call parser for models that emit tool calls as text tags.
usage_tracking: Stream wrapper that records token usage on the active tracing span.

AnthropicClientanthropic: Anthropic client for Claude models.
AzureConfigopenai: Configuration for Azure OpenAI Service.
AzureOpenAIClientopenai: Azure deployment client with API-key authentication and streaming support.
GeminiModelgemini: Gemini model client wrapping the adk-gemini crate for the Llm trait.
MockLlm: A mock LLM implementation for testing without real API calls.
OpenAIClientopenai: OpenAI client for standard OpenAI API and OpenAI-compatible APIs.
OpenAICompatibleopenai: Shared OpenAI-compatible client implementation.
OpenAICompatibleConfigopenai: Configuration for OpenAI-compatible providers.
OpenAIConfigopenai: Configuration for OpenAI API.
RetryConfig: Configuration for automatic retry with exponential backoff.
ServerRetryHint: Hint from the server about when to retry.

ModelProvider: Canonical provider identifiers and metadata shared across ADK crates.
ReasoningEffortopenai: Reasoning effort level for OpenAI reasoning models (e.g., o1, o3).
ReasoningReplayFieldopenai: Assistant-message field accepted by an OpenAI-compatible provider for reasoning replay.