Crate agent_chain_core 

Agent Chain Core - A Rust implementation of LangChain core library.

This crate provides:

• Message types for LLM conversations (human, AI, system, tool)
• Tool trait and #[tool] macro for function calling
• Chat model abstractions and provider integrations
• Support for multiple providers (Anthropic, OpenAI, etc.)

Architecture

The architecture follows LangChain’s pattern:

• Core layer ([chat_model]): Base ChatModel trait that all providers implement
• Message layer (messages): Message types for conversations
• Tools layer (tools): Tool definitions and the #[tool] macro

Quick Start

use agent_chain_core::{init_chat_model, HumanMessage};

// Initialize a model - provider is inferred from name
let model = init_chat_model("claude-sonnet-4-5-20250929", None)?;

// Or specify explicitly
let model = init_chat_model("my-custom-model", Some("openai"))?;

// Use the model
let messages = vec![HumanMessage::new("Hello!").into()];
let response = model.generate(messages, None).await?;

Feature Flags

• default: Includes all providers
• specta: Specta derive support

Re-exports

pub use error::Error;

pub use error::Result;

pub use chat_models::BoundChatModel;

pub use chat_models::ChatChunk;

pub use chat_models::ChatModel;

pub use chat_models::ChatModelExt;

pub use chat_models::ChatResult;

pub use chat_models::ChatResultMetadata;

pub use chat_models::ChatStream;

pub use chat_models::DynBoundChatModel;

pub use chat_models::DynChatModelExt;

pub use chat_models::LangSmithParams;

pub use chat_models::ToolChoice;

pub use chat_models::UsageMetadata;

pub use messages::AIMessage;

pub use messages::AnyMessage;

pub use messages::BaseMessage;

pub use messages::ContentPart;

pub use messages::HasId;

pub use messages::HumanMessage;

pub use messages::ImageDetail;

pub use messages::ImageSource;

pub use messages::MessageContent;

pub use messages::SystemMessage;

pub use messages::ToolCall;

pub use messages::ToolMessage;

pub use tools::Tool;

pub use tools::ToolDefinition;

Modules

chat_models

Core ChatModel trait and related types.

error

Error types for agent-chain.

messages

Message types for LLM interactions.

tools

Tools module for LLM function calling.

Attribute Macros

async_trait

tool

Marks a function as a tool that can be used by an LLM.