LangGraph Workflow Agent –

A powerful, extensible AI workflow agent built using LangGraph that leverages:

• Google Gemini models
• LangGraph Studio for debugging, visualization & time travel
• Tavily for real-time web search
• Workflow creation, review, and saving
• Human-in-the-loop (HITL) editing
• Extensible tool interface with memory

---

🚀 What It Does

This agent functions as a workflow orchestrator. It takes a user query and:

1. Uses the LLM to generate a task-specific workflow
2. Optionally allows the user to review/edit the generated steps
3. Utilizes tools like Tavily Search for real-time queries
4. Saves the workflow to a file using a custom Save Workflow tool
5. Stores interactions in memory for contextual reasoning

---

🛠️ Tech Stack

Component	Description
LLM Backend	Google Gemini (gemini-2.0-flash)
Tooling	Tavily Search, Save-to-File Workflow Tool
LangGraph IDE	Full support for LangGraph Studio
HITL	CLI-based review before finalizing output
Memory	In-memory state storage

---

⚙️ Getting Started

1. Clone the Repository

  git clone https://github.com/sohailz360/Workflow_Agent_langgraph.git
  cd Workflow_Agent_langgraph

2. Set Up Environment Variables

Update the .env file with your credentials and update all credits.

You Can Use a different model other than gemini , just add its api and chnage the model and provider name in configuration.py

3. Install Python Dependencies

uv pip install -e .

Run this in your virtual environment , If you use another LLm not gemini then install specific package for using its api also.

Running the Agent

LangGraph Studio Mode For visualization and debugging:

langgraph dev

Features

Feature	File / Description
Prompt Design	prompts.py – Editable system instructions
Model Loader	configuration.py – Connects to Gemini
Workflow Agent	graph.py – Agent loop with logic
Tavily Search Tool	tools.py – Web search integration
Save Workflow Tool	tools.py – Writes agent workflow to disk
Human Review	HITL step in agent routing logic
Memory System	Simple in-memory message store

---

File Structure

src/react_agent/
│
├── graph.py              # Main workflow agent logic
├── tools.py              # Tavily + Save Workflow tools
├── prompts.py            # Prompt templates
├── state.py              # Agent input/output schema
├── configuration.py      # LangGraph configuration
└── utils.py              # Helper functions