Build Smart and Autonomous Apps with LangChain & Large Language Models
Table of contents
Large Language Models (LLMs) have the ability to understand human language and generate human-like responses to queries.
LLMs are pre-trained and fine-tuned on massive datasets that make them great for problems that require the understanding of human languages and and text generation, including, for example, question-answering and text summarization.
For more specialized use cases that involve data the LLM has not been trained on, we need to provide the LLM with:
- External data that the LLM can use to solve the problem
- Tools that offer capabilities and features to assist the LLM in problem-solving
- Autonomous problem-solving features to enable the LLM to solve problems independently
LangChain is an open-source library that enables software engineers to develop applications that connect LLMs, like ChatGPT, with the data and tools necessary to perform advanced automation tasks.
Core LangChain Features
Based on LangChain 0.0.157
LangChain is a comprehensive framework for developing applications powered by language models. The core features that LangChain provides are the following:
Schema: data structures used for storing data, including text, document, and chat message
Models: the core functionality that takes input and generates output, including the embedding, language, and chat models
Prompts: the queries (input) given to the LLM by the user
Indexes: data structures that allow us find documents that are relevent to the query
Memory: short-term and long-term memory for applications, e.g., chat messages and history
Chains: provides a way of splitting a complex task into more optimal subtasks
Agents: a component that can autonomously solve problems through the use of external tools and the other features of LangChain
Structured tool: https://blog.langchain.dev/structured-tools/
LangChain provides you with a set of tools that allow you to create autonomous agents powered by large language models that have access to your internal and external data sources.