🔍 Deep Researcher

Deep Researcher is a powerful research agent built with LangChain and LangGraph that helps you conduct comprehensive research on any topic by leveraging multiple data sources.

💡 Features

• Answer research queries using AI-powered reasoning
• Search the web using multiple search engines:
  • SerpAPI (Google search results)
  • DuckDuckGo (privacy-focused search)
  • Tavily (AI-powered search optimized for research)
• Search academic papers from arXiv and Google Scholar
• Find relevant YouTube videos
• Discover insights from Twitter and LinkedIn posts
• Upload local documents to include in research
• Send research results via email
• Interactive visualization of the agent's reasoning process
• Customizable search options - enable/disable specific search features

🔧 Installation

For detailed installation instructions and troubleshooting tips, please see the Installation Guide.

Quick start:

1. Clone this repository:

git clone https://github.com/Sallyliubj/Deep-Researcher.git
cd Deep-Researcher

2. Create and activate a virtual environment:

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

3. Install required dependencies:

pip install -r requirements.txt

4. Set up Ollama locally for the LLM:

# Install Ollama if not already installed
curl -fsSL https://ollama.com/install.sh | sh

# Pull the Gemma model
ollama pull gemma3:1b

5. Create a .env file in the root directory with your API keys (see .env.example).

🔭 Usage

1. Start the Streamlit app:

export PYTHONPATH=$(pwd)
streamlit run app/main.py

Or use the provided shell script:

./run.sh

2. Open your browser and go to http://localhost:8501

3. Enter your research query, select the search options you want to enable, and click "Research"

4. View the results, LangGraph visualization, and send the results to an email if desired

🌐 Web Search Options

Deep Researcher offers multiple web search options:

• SerpAPI: Provides Google search results (requires API key)
• DuckDuckGo: Privacy-focused search engine (no API key required)
• Tavily: AI-powered search engine optimized for research (requires API key)

You can enable or disable any of these search engines from the sidebar.

🧩 Architecture

This project uses:

• LangChain for orchestrating the various AI components and tools
• LangGraph for creating a dynamic research workflow with self-reflection capabilities
• Ollama (gemma3:1b) for local AI model inference
• Streamlit for the web interface

📺 Video Demonstrations

Search Agent Demo

System Overview

RAG Implementation Demo

MCP Demo

---

This software is for personal, non-commercial use only. Redistribution or modification is strictly prohibited.