LangChain and LangGraph

<aside> 💡

References

LangGraph Doc - https://langchain-ai.github.io/langgraph/

LangChain Doc - https://python.langchain.com/docs/introduction/

Additional Pages

Multi Agent LangGraph Tutorial

</aside>

LangGraph is a tool or framework designed to create dynamic and flexible workflows for natural language processing (NLP) and large language models (LLMs). It leverages graph-based structures, enabling users to define nodes and edges to represent steps in a pipeline, such as input processing, decision-making, or integration with external tools. LangGraph is particularly useful for building complex, modular systems where each node represents a specific task or function, making it easy to adapt and scale workflows for diverse use cases in NLP and AI applications.

LangChain is a powerful framework designed for building applications powered by large language models (LLMs) like OpenAI's GPT. It provides tools to chain together LLMs with other functionalities such as prompt engineering, memory, APIs, and external data sources. LangChain enables developers to create end-to-end systems for tasks such as question-answering, summarization, and generative pipelines. By integrating components like data loaders, vector stores, and LLM tools, LangChain simplifies the development of robust, scalable AI applications tailored to complex real-world scenarios.

Orchestration

For complex orchestrations, LangGraph would be a suitable option. Main reason being, Chains vs Graphs.

• Flow → Chains have sequential and straightforward pipeline, where as Graphs have a parallel flow. → They allow branching, looping and more complex control flows
• Modularity → Chains provide pre-built modules. Graphs are often complex but offers SubGraphs. Where each graph can be considered a single module as well.

Core Components

1. Nodes

• Represents individual actions or computations within the graph.
• Can be simple functions, LLM Calls, or even sub-graphs. → What this means is the scope of a node is based on perspective.

1. Edges

• Define the connections between nodes, specifying the flow of data and control.

1. State

• The internal state of the application, which is updated as nodes execute.
• Persistent data that is passed between nodes.