Hello there, AI enthusiasts! Today, we're going to have a closer look at LangChain Chat Models. LangChain is a fantastic tool that provides a standard interface for interacting with a variety of language models, including both text-based Large Language Models (LLMs) and Chat Models.

The Concept of Models in LangChain

Models are an essential part of LangChain. Rather than being a provider of models, LangChain serves as an interface allowing you to interact with diverse language models from other places. Currently, LangChain supports LLMs, Chat Models, and Text Embedding models.

The primary distinction between LLMs and Chat Models lies in their input and output structure. LLMs operate on a text-based input and output, while Chat Models follow a message-based input and output system.

It's important to note that chat model APIs are a relatively new feature, and they're still determining the most effective abstractions.

Getting Started with Chat Models

Chat Models are a variation of language models. While they use language models underneath, the interface they expose is somewhat different. Instead of a "text in, text out" API, they expose an interface where "chat messages" are the inputs and outputs.

At LangChain, they refer to a ChatMessage as the modular unit of information for a chat model. This usually includes a "text" field, signifying the content of the chat message. There are currently four different classes of ChatMessage supported:

LangChain presently supports the ChatOpenAI model (with gpt-4 and gpt-3.5-turbo), but they anticipate adding more in the future. To get started, all you need to do is use the call method of an LLM implementation, passing in a string input. Here's a simple example of how to initiate a conversation using the ChatOpenAI implementation:

import { ChatOpenAI } from "langchain/chat_models/openai";
import { HumanChatMessage } from "langchain/schema";

export const run = async () => {
  const chat = new ChatOpenAI();
  const response = await chat.call([
    new HumanChatMessage(
      "What is a good name for a company that makes colorful socks?"
    ),
  ]);
  console.log(response);
  // AIChatMessage { text: '\n\nRainbow Sox Co.' }
};

To get a deeper understanding, you can refer to the API references: ChatOpenAI and HumanChatMessage.

Integrations: Chat Models

LangChain offers several Chat Models implementations that integrate with various model providers. These include ChatOpenAI, Azure ChatOpenAI, and ChatAnthropic. Each of these providers has a different way of getting instantiated. For example, to instantiate ChatOpenAI:

import { ChatOpenAI } from "langchain/chat_models/openai";

const model = new ChatOpenAI({
  temperature: 0.9,
  apiKey: "your_openai_api_key"
});

The temperature is a hyperparameter that adjusts the randomness of the model's responses. Higher values result in more diverse outputs, while lower values make the output more deterministic.

The apiKey is your personal OpenAI API key, which authorizes you to use the model.

Similarly, you can instantiate Azure ChatOpenAI and ChatAnthropic:

// Azure ChatOpenAI
import { AzureChatOpenAI } from "langchain/chat_models/azure";

const model = new AzureChatOpenAI({
  temperature: 0.9,
  azureApiKey: "your_azure_api_key"
});

// ChatAnthropic
import { ChatAnthropic } from "langchain/chat_models/anthropic";

const model = new ChatAnthropic({
  temperature: 0.9,
  anthropicApiKey: "your_anthropic_api_key"
});

Note: Make sure to replace "your_openai_api_key", "your_azure_api_key", and "your_anthropic_api_key" with your actual API keys.

Keep in mind that each provider may have specific parameters unique to their implementation. Always refer to the LangChain documentation or the provider's documentation for a comprehensive understanding of the parameters.

Conclusion

LangChain's Chat Models provide a unified approach to interacting with various AI language models. This versatility allows developers to switch between different providers seamlessly. While we've covered the basics of instantiating chat models from OpenAI, Azure, and Anthropic, there's a wealth of other functionality and customization available in LangChain. As an open-source project, LangChain encourages community contributions to drive its growth and evolution. With LangChain, the potential to create powerful language processing applications is at your fingertips.

Subscribe or follow me on Twitter for more content like this!

FAQs

What is LangChain?

How can I contribute to LangChain?

What is the temperature parameter in the instantiation code?

What is an API key, and why do I need it?

Where can I get my API key?

Is LangChain free to use?

Also published here.