Crate cognate_core 

Cognate Core — HTTP client, traits, and base types for LLM providers.

This crate provides the foundational abstractions for building provider-agnostic LLM applications with type-safe interfaces and zero-cost abstractions.

Quick start

use cognate_core::{Provider, Request, Message};

async fn run<P: Provider>(provider: &P) -> cognate_core::Result<()> {
    let response = provider
        .complete(
            Request::new()
                .with_model("gpt-4o-mini")
                .with_message(Message::user("Hello!")),
        )
        .await?;
    println!("{}", response.content());
    Ok(())
}

Re-exports

pub use error::Error;

pub use error::Result;

pub use middleware::Layer;

pub use middleware::Middleware;

pub use middleware::ProviderExt;

pub use mock::MockProvider;

pub use ratelimit::TokenBucket;

Modules

error

Error types for Cognate.

middleware

Tower-inspired middleware system for Cognate providers.

mock

Mock provider for use in tests.

ratelimit

Token bucket rate limiting implementation

types

Re-exports of core types for ergonomic imports.

Structs

Choice

A single completion choice within a Response.

Chunk

A single chunk in a streaming response.

Delta

The incremental content delta inside a Chunk.

Message

A single message in a conversation.

ProviderConfig

Configuration shared by all provider clients.

Request

A completion request sent to a provider.

Response

A completed response from a provider.

ResponseFormat

Structured output format specifier.

ToolCall

A tool invocation requested by the model.

ToolCallFunction

The function component of a ToolCall.

Usage

Token usage statistics for a request.

Enums

Role

The role of a message sender.

Traits

EmbeddingProvider

Trait for providers that can generate embedding vectors.

Provider

Core trait for all LLM providers.