Personal Deep Researcher

This is a personal Deep Research tool built using Langgraph. It automates the process of researching and compiling information on various topics by leveraging multiple search tools and language models.

Features

• Generates a detailed report on a topic following the outline with sources.
• Integrates three graphs, Orchestrator graphs uses Researcher graph which uses Searcher graphs.
• Uses Interrupt API from langgraph to take human input for planning.
• Integrates GPT-4o, 4o-mini, Gemini-flash-1.5 and deepseek-r1-distill-qwen-32b effectively to reduce the cost per million tokens.
• Uses Send and Command API from Langgraph to simultaneously execute Research graph and Searcher graphs for each section.
• Searcher graph decides between Tavily, Wikipedia and Arxiv search engines to get the most relevant information.
• Keeps track of the search queries to avoid querying the same term multiple times.
• Research graph decides whether more web search is needed or not.

Project Structure

Installation

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install the required dependencies:

   pip install -r requirements.txt

4. Set up environment variables by creating a .env file in the root directory with the following content:

   OPENAI_API_KEY = "your_openai_api_key"
   ANTHROPIC_API_KEY = "your_anthropic_api_key"
   GEMINI_API_KEY = "your_gemini_api_key"
   HF_TOKEN = "your_hf_token"
   TAVILY_API_KEY = "your_tavily_api_key"
   GOOGLE_API_KEY = "your_google_api_key"
   GOOGLE_CSE_ID = "your_google_cse_id"
   NEWSAPI_API_KEY = "your_newsapi_api_key"
   GROQ_API_KEY = "your_groq_api_key"
   GEMINI_API_KEY = "your_gemini_api_key"

Usage

1. To run the deep research agent, execute:

   python deep_researcher/deep_research_agent.py

2. To run the search graph, execute:

   python deep_researcher/search_graph.py

3. To run the researcher graph, execute:

   python deep_researcher/researcher_graph.py

Project Components

• deep_research_agent.py: Main script to run the deep research agent.
• researcher_graph.py: Script to build and run the researcher graph.
• search_graph.py: Script to build and run the search graph.
• utils/nodes.py: Contains the nodes (functions) used in the graphs.
• utils/state.py: Defines the state models used in the graphs.
• utils/tools.py: Contains the tools used for searching information.

License

This project is licensed under the MIT License.