modular-agent-fullstack

# Modular Agent Fullstack
Modular Agent Fullstack is an extended and improved version of the "[Gemini Fullstack LangGraph Quickstart](https://github.com/google-gemini/gemini-fullstack-langgraph-quickstart)" project by the Google ChatGPT Team. While the original project offers a great introduction to building fullstack AI agents using LangGraph, it also comes with several limitations that can hinder beginners and AI enthusiasts from further exploration, customization, or learning.

This project addresses those limitations by refactoring the codebase into a more modular framework, making it easier to extend, customize, develop, and deploy AI agents with your preferred UI and backend architecture.

<div style="display: flex; justify-content: center; gap: 5px;">
  <img src="assets/select-agent.png" width="49%" />
  <img src="assets/rag-ui.png" width="49%" />
</div>

## Limitaions of Gemini's Fullstack LangGraph Project
- **Heavy Dependence on LangGraph**: The original project relies extensively on LangGraph, not just for building the agent, but also for serving frontend components and powering the backend API server. This hides important implementation details and reduces flexibility for developers who prefer to use their own methods for frontend/backend integration.
- **Difficult to Extend with Custom Agents**: The current frontend design is rigid and doesn’t support adding custom agents easily, for example, creating custom dashboards for workflow visualization or adding configurable parameter inputs.
- **Limited to Gemini**
- **Cumbersome Development and Deployment**: Although Docker is used, the setup lacks a dedicated development environment and does not offer a clear or feasible deployment strategy - for example, hosting the project on a public URL is not straightforward.

## Improvements in Modular Agent Fullstack
- **Reduced LangGraph Dependency**: LangGraph dependencies `@langchain/langgraph-sdk` have been removed from the frontend. The backend and Docker configu