🔍 Research Assistant AI Agent

An AI-powered research assistant built using LangChain, OpenAI, and tool-calling agents that can search the web, fetch Wikipedia data, and save structured research outputs to a file.

🚀 Features

• 🌐 Web search using DuckDuckGo
• 📚 Wikipedia-based research
• 🧠 AI-generated structured responses
• 💾 Save research outputs to a .txt file
• 🛠️ Tool-calling agent using LangChain
• 📄 Structured output using Pydantic

---

🛠️ Tech Stack

• Python
• LangChain
• OpenAI (GPT-4o-mini)
• Pydantic
• DuckDuckGo Search
• Wikipedia API
• dotenv

---

⚙️ Setup Instructions

1. Clone the repo

git clone <your-repo-link>
cd ResearchAssistantAgent

2. Create virtual environment

python3 -m venv .venv
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Add your API key

Create a .env file:

OPENAI_API_KEY=your_api_key_here

---

▶️ Usage

Run the agent:

python main.py

Then enter a query like:

What is MCP server? ⚡ Example Output

{
  "topic": "MCP server",
  "summary": "An MCP server is a system that connects AI models to external tools and data...",
  "source": ["Wikipedia", "Web Search"],
  "tools_used": ["search", "wiki"]
}

---

🧠 How It Works

The agent uses multiple tools:

• search → Fetches real-time web results
• wiki → Retrieves summarized Wikipedia content
• save_text_to_file → Stores research output in a text file

The AI agent:

• Understands the query
• Decides which tools to use
• Collects information
• Returns structured output
• Optionally saves results to a file

---

💾 Output Saving

All saved outputs are appended to:

research_output.txt

Each entry includes:

• Timestamp
• Structured research data

---

⚠️ Notes

• .env is ignored for security
• Do not expose API keys
• DuckDuckGo search may fail on older Python SSL setups (use Python 3.11+)

---

🔮 Future Improvements

• Add PDF export
• Add streaming responses
• Improve error handling
• Add UI (Streamlit / React)
• Support multiple research topics at once