🤖 AI Research Agent

An autonomous research assistant that searches the web, queries Wikipedia, and delivers structured research summaries — powered by LangGraph + Groq LLaMA 3.3-70B.

---

What it does

Give it any research topic and the agent autonomously:

1. Searches the web via DuckDuckGo for recent information
2. Queries Wikipedia for background knowledge
3. Synthesizes both sources using LLaMA 3.3-70B (via Groq's free API)
4. Returns structured JSON output with summary, sources, and tools used
5. Saves results to a local text file for later reference

$ python main.py

What can I help you research? > Generative AI

{
  "topic": "Generative AI",
  "summary": "Generative AI refers to algorithms that can create new content...",
  "sources": ["Wikipedia", "DuckDuckGo"],
  "tools_used": ["search_tool", "wiki_tool"]
}

---

Architecture

User Input
    │
    ▼
┌─────────────────────────────────────┐
│           LangGraph Agent           │
│                                     │
│  ┌──────────┐    ┌───────────────┐  │
│  │ DuckDuck │    │   Wikipedia   │  │
│  │  Search  │    │    Lookup     │  │
│  └──────────┘    └───────────────┘  │
│         \              /            │
│          ▼            ▼             │
│     ┌──────────────────────┐        │
│     │  Groq LLaMA 3.3-70B  │        │
│     │   (Synthesis Layer)  │        │
│     └──────────────────────┘        │
└─────────────────────────────────────┘
    │
    ▼
Structured JSON + Saved Output File

---

Tech Stack

Layer	Technology
Agent framework	LangGraph 0.2
LLM	Groq API — LLaMA 3.3-70B Versatile
Web search	DuckDuckGo Search (no API key needed)
Knowledge base	Wikipedia via langchain-community
Output schema	Pydantic v2
Language	Python 3.11

---

Project Structure

AI_agent/
├── main.py              # Agent graph definition and entry point
├── tools.py             # Tool definitions (search, wiki, save)
├── requirements.txt     # All dependencies
├── .env                 # API keys — NOT committed to GitHub
└── research_output.txt  # Generated research outputs

---

Getting Started

Prerequisites

• Python 3.11+
• A free Groq API key → console.groq.com

Installation

# 1. Clone the repo
git clone https://github.com/Shreyaaaaaak/AI_agent.git
cd AI_agent

# 2. Create and activate a virtual environment
python -m venv venv
venv\Scripts\activate       # Windows
source venv/bin/activate    # Mac/Linux

# 3. Install dependencies
pip install -r requirements.txt

# 4. Add your Groq API key
echo "GROQ_API_KEY=your_key_here" > .env

# 5. Run the agent
python main.py

Note: Groq's API is completely free for this use case — no credit card required.

---

Example Output

Input: "What is LangGraph?"

{
  "topic": "LangGraph",
  "summary": "LangGraph is a library for building stateful, multi-actor applications with LLMs. It extends LangChain by enabling cyclic computation graphs, making it ideal for agentic workflows where an LLM must reason, act, and observe in a loop.",
  "sources": ["Wikipedia", "DuckDuckGo"],
  "tools_used": ["search_tool", "wiki_tool", "save_tool"]
}

---

Why LangGraph over vanilla LangChain?

LangGraph allows the agent to loop — it can decide to search again if the first result isn't sufficient, rather than executing tools in a fixed linear chain. This makes the research behavior genuinely agentic rather than scripted.

---

Limitations & Future Improvements

• Add a Streamlit or Gradio web UI
• Support multi-turn conversations with memory
• Integrate arXiv search for academic topics
• Add citation formatting (APA / MLA)
• Deploy as a web API using FastAPI

---

Built by

Shreya Kaushik — ECE @ KIIT University (2027)
GitHub · LinkedIn

---

License

MIT — feel free to fork and build on this.