Open-DeepThink-Researcher

Open source deep researcher that also thinks based on `OpenAI DeepResearch. This automates online research using DeepSeek-R1 via Nebius AI Studio and Exa search APIs. It continuously generates search queries, extracts relevant content, evaluates information and produces a comprehensive research report based on all relevant information collected.

Watch demo video on Youtube

You can run this deep researcher using this Google Colab

Features

• Automated Research: Uses AI to generate search queries and evaluate results.

• Iterative Search: Runs multiple iterations to refine and improve the quality of results.

• Content Extraction & Evaluation: Identifies and extracts relevant information from search results.

• Final Report Generation: Summarizes findings into a well-structured report.

• Gradio UI: Provides a user-friendly web interface for easy interaction.

Potential Improvements

• User can choose any LLM models available via Nebius Studio to power DeepThink-Researcher.
• Increase the number of iterations to refine responses further.
• Extend the text/context length from 2,000 to 5,000 or even 20,000 words/characters — but this will increase processing time and API costs.
• Optimize response times for production and public use to enhance performance.
• Export whole notebook code into hosted python web app using tools like reflex.dev or streamlit.
• Add option for users to download final report.

Installation

To run this project, install the required dependencies:

!pip install nest_asyncio gradio aiohttp openai exa_py

Configuration

Before running the researcher, set up your API keys in the notebook:

NEBIUS_API_KEY="your_nebius_api_key"# Replace with your Nebius API keyEXA_API_KEY="your_exa_api_key"# Replace with your EXA API key

Parameters

• User Query: The research topic or question you want to investigate.

• Iteration Limit: The maximum number of iterations the AI should perform while refining search queries.

Example Queries

Try out the following example queries:

• "What are the latest advancements in quantum computing?"

• "How does intermittent fasting impact metabolism?"

• "Best practices for deploying large-scale AI models."

• "Comparison of cloud AI providers: AWS vs. GCP vs. Azure."

Project Structure

• async_research(): Handles the asynchronous execution of research.

• call_nebius_async(): Calls Nebius AI for generating responses.

• perform_search_async(): Uses Exa API to fetch search results.

• is_page_useful_async(): Evaluates the relevance of a webpage.

• extract_relevant_context_async(): Extracts meaningful content from pages.

• generate_final_report_async(): Compiles all information into a final research report.

• gradio_run(): Wraps everything into a Gradio UI.

Contributing

Contributions are welcome! Please feel free to open issues or submit pull requests.

License

This project is licensed under the MIT License.

Acknowledgments

• This researcher is a extended fork of OpenDeepResearcher
• DeepSeek Model is used via Nebius AI Studio