Deep Research Agent

An advanced AI-powered research assistant that creates comprehensive, well-researched blog posts on any topic by combining web search, content analysis, and automated writing capabilities.

Features

• 🔍 Intelligent web searching across multiple sources
• 📝 Automated content extraction and analysis
• ✍️ Well-structured blog post generation
• 📚 Proper citation and source attribution
• 💾 Organized research storage

Prerequisites

Before you begin, ensure you have the following installed:

• Node.js (v16 or higher)
• pnpm (recommended) or npm
• Ollama (if using default LLM provider)

Installation

1. Clone the repository:

1git clone <repository-url>
2cd deep-research-agent

2. Install dependencies:

1pnpm install
2# or
3npm install

Environment Variables

The agent uses environment variables for LLM configuration. Create a .env file in the root directory:

1# LLM Provider Configuration (Optional)
2LLM_PROVIDER=ollama    # Options: 'ollama' (default) or 'bedrock'
3LLM_MODEL_ID=llama3.2  # Default model for Ollama
4
5# If using AWS Bedrock (only needed if LLM_PROVIDER=bedrock)
6AWS_ACCESS_KEY_ID=your_aws_access_key
7AWS_SECRET_ACCESS_KEY=your_aws_secret_key
8AWS_REGION=us-west-2

By default, the agent uses Ollama with the llama3.2 model. Make sure you have Ollama installed and the model downloaded if using the default configuration.

Usage

Run the agent with a research query:

1pnpm start -- "your research query"
2# or
3npm start -- "your research query"

Examples:

1pnpm start -- "What are the latest developments in quantum computing?"
2pnpm start -- "What is the current state of renewable energy?"

How It Works

The Deep Research Agent follows a sophisticated process to create high-quality blog posts:

1. Web Search:

   • Searches for the most relevant and authoritative sources
   • Identifies key resources and references

2. Content Analysis:

   • Scrapes and analyzes content from selected sources
   • Extracts important information and insights
   • Maintains source attribution

3. Content Generation:

   • Synthesizes information from multiple sources
   • Creates well-structured blog posts
   • Includes proper citations and references

4. Output:

   • Saves the generated blog post in the research directory
   • Provides progress updates during the research process
   • Shows token usage statistics upon completion

Output Structure

The agent creates a research directory containing:

• Generated blog post in markdown format
• Source attribution and references
• Metadata about the research process

Notes

• The research process may take several minutes depending on the topic complexity
• Internet connection is required for web search and content analysis
• The quality of results depends on the availability of reliable sources
• Token usage is tracked and displayed after completion

License

MIT