an open-source orchestration library that helps developers connection LLMs for building complex applications_

![[Pasted image 202509117.png.webp]]

It provides a standard interface for various components and abstractions:

  • LLMs: The backbone of LangChain, LLMs like OpenAI’s GPT-3 or GPT-4 provide the core capabilities for understanding and generating language. They are trained on vast datasets to produce coherent and contextually relevant text.
  • Prompt Templates: These templates structure the input to LLMs, maximizing their effectiveness in understanding and responding to queries. By designing effective prompts, developers can guide the LLMs to produce desired outputs.
  • Output Parsers: These components refine the language generated by LLMs into formats that are useful and relevant to specific tasks, enhancing the overall user experience.
  • Vector Store: This component handles the embedding of words or phrases into numerical vectors, which is essential for tasks involving semantic analysis and understanding language nuances.
  • Agents: Agents are decision-making components that determine the best course of action based on input, context, and available resources. They enable LLMs to interact intelligently with their environment.

Chain#

Chain of Thought (CoT), the implementation on minimum manageable steps

  • explicitly guiding the LLM to think step-by-step through a problem
  • split the overall task into subtasks and run the subtasks sequentially
  • the output of one step becomes the input of the next

![[Pasted image 202509118.png.webp]]

chain = prompt | model **|** outputParser

LangChain Ecosystem#

!Image Description

LangGraph#

LangGraph is an extension of LangChain that focuses on building more complex and controllable agent workflows using a graph-based approach

LangChain vs LangGraph (DAG vs Graph)#

!Image Description

Key components:

  • State: A shared data structure that represents the current snapshot of your application.
  • Nodes: Python functions that encode the logic of your agents. They receive the current State as input, perform some computation or side-effect, and return an updated State.
  • Edges: Python functions that determine which Node to execute next based on the current State. They can be conditional branches or fixed transitions.

Support cases such as time travel/human in the loop/parallelization/multi agents(stateful) architecture

Deciding Between LangChain and OpenAI API#

Use Direct API (such as OpenAI) When:#

  • Simplicity and performance are top priorities.

  • Tasks are straightforward and don’t require chaining operations.

  • Answering single-turn questions

  • Running one-off analyses

  • Performance-critical applications

Use LangChain When:#

  • You need to chain multiple operations conveniently.

  • Development speed and maintainability are important.

  • Your project involves integrating tools, external APIs, or maintaining state across interactions.

  • Ease of Chaining Tasks: You can easily combine multiple operations (e.g., answering questions based on external data).

  • Built-in Tools and Agents: LangChain offers agents to manage complex workflows with minimal code.

  • Extensibility: It integrates with external systems like document stores, APIs, and Python functions.

LangSmith#

![[Pasted image 2025091111.png.gif]]

Appendix#

!Image Description