LangGraphAgenticAI: Building Stateful AI Agents for SDLC Feedback and AI News Summarization

Abstract

LangGraph is a Python library for creating stateful, multi-step AI workflows using Large Language Models (LLMs). This paper presents LangGraphAgenticAI, a project that leverages LangGraph to build an agent for SDLC feedback and AI news summarization. The agent utilizes graph-based workflows, state management, and prebuilt templates to enable dynamic decision-making and interaction with tools/APIs. The results demonstrate the effectiveness of LangGraph in building complex and versatile AI agents.

1. Introduction

Stateful AI agents are becoming increasingly important for applications like customer service, information retrieval, and software development. However, building and deploying such agents requires managing state, orchestrating multiple actors, and handling complex workflows. LangGraph provides a solution by enabling developers to create stateful, multi-step AI workflows using LLMs. This paper introduces LangGraphAgenticAI, a project that showcases these capabilities through an agent designed for SDLC feedback and AI news summarization.

2. Background

LangGraph allows the creation of graph-based architectures where workflows are defined as nodes (tasks) and edges (connections). Inspired by systems like Pregel and Apache Beam, it's used by companies like LinkedIn, Uber, and Elastic. Key features include:

Graph-Based Workflows: Supports conditional routing between tasks.
State Management: Enables persistence of conversation history, tool outputs, etc.
Multi-Actor Support: Supports collaborative agents.
Prebuilt Agents: Offers customizable templates like create_react_agent.

3. Project Description

LangGraphAgenticAI includes:

SDLC Feedback Node: Collects and processes SDLC-related feedback.
AI News Summarization Node: Summarizes AI-related news.
Streamlit UI: Provides a simple frontend for interaction.

4. Implementation

The implementation includes three key Python files:

sdlc_node.py - Implements SDLC feedback logic
ai_news_node.py - Implements AI news summarization logic
app.py - Hosts the Streamlit-based UI

Steps to run:

git clone https://github.com/aitiwari/langgraph.git
cd langgraph
pip install -r requirements.txt
export GROQ_API_KEY="your-api-key-here"
streamlit run app.py

Access the UI at: http://localhost:8501

🌟 Use Cases

1. Basic Chatbot

Prompt: What is generative AI?

2. Chatbot with Tool

Prompt: Latest news from the India vs Australia match?

Prompt2: search latest news from india

3. Appointment Receptionist

Prompt 1: Book an appointment for priti.

Prompt 2: Yes or No.

4. Customer Support

5. Travel Planner

6. AI News Summarization

6. Templates & Community

Explore prebuilt templates using:

langgraph new

Template	Description	Link
ReAct Agent	Tool-calling agent with memory	GitHub
Retrieval Agent	RAG with dynamic context fetching	Docs

7. Conclusion

This paper presented LangGraphAgenticAI, an implementation of a stateful agent using LangGraph. It demonstrated how LangGraph can be used to create complex and versatile agents for SDLC feedback and AI news summarization. Future work includes extending it to more use cases and improving performance.

8. References