Crate opi_ai 

Unified multi-provider LLM API with streaming support.

Provides a standardized interface for interacting with multiple LLM providers: Anthropic, OpenAI Chat Completions, OpenAI Responses, Google Gemini, plus OpenAI-compatible profiles for OpenRouter and Mistral.

Re-exports

pub use config::Config;

pub use config::Error;

pub use model::Model;

pub use provider::Provider;

pub use registry::ProviderRegistry;

pub use registry::RegistrationError;

pub use registry::RegistryError;

pub use stream::AssistantStreamEvent;

pub use stream::CostBreakdown;

pub use stream::CumulativeUsage;

pub use stream::Pricing;

pub use stream::calculate_cost;

Modules

anthropic

Anthropic Messages SSE provider (S8.1).

azure_openai

Azure OpenAI provider profile (task 3.2).

bedrock

AWS Bedrock provider (task 3.1).

config

Provider configuration and authentication.

gemini

Google Gemini streamGenerateContent SSE provider (S8.1).

http

Shared HTTP client with connection pooling and proxy support (tasks 3.13, 3.12).

message

Provider-facing message types (S7.1).

mistral

Mistral provider profile - routes through the OpenAI-compatible adapter.

model

Model discovery and configuration.

openai_chat

OpenAI-compatible chat completions SSE provider (S8.1).

openai_responses

OpenAI Responses API SSE provider (S8.1).

openrouter

OpenRouter provider profile - routes through the OpenAI-compatible adapter.

provider

LLM provider abstraction (S8.1).

registry

Provider registry — resolves provider:model specs to provider + model info.

retry

Retry/backoff/rate-limit support (task 2.15).

stream

Streaming response events (S7.3).

vertex

Google Vertex AI provider (task 3.3).

Enums

ApiKind