InterDeepResearch: Enabling Human-Agent Collaborative Information Seeking through Interactive Deep Research

A prototype interacive research system that enables human-agent collaborative information seeking. Observe, steer, and navigate the research process in real-time with visual sensemaking and evidence tracing.

🔥 Paper: InterDeepResearch: Enabling Human-Agent Collaborative Information Seeking through Interactive Deep Research

System Interface & Demo

Benchmark Performance

To validate the basic information-seeking capabilities of InterDeepResearch, we evaluate it (without user intervention) on two representative text-based deep research benchmarks: Xbench-DeepSearch-v1 and Seal-0. The results demonstrate that InterDeepResearch achieves competitive performance in automated information seeking, even surpassing popular commercial deep research systems like Perplexity Deep Research and Gemini Deep Research.

Getting Started

Prerequisites

• Python 3.11 or higher
• Node.js 16 or higher
• npm or yarn package manager

Backend Setup

1. Clone the repository and navigate to the project root:

git clone https://github.com/bopan3/InterDeepResearch.git
cd InterDeepResearch

2. Create and activate a conda environment:

conda create -n InterDeepResearch python=3.11
conda activate InterDeepResearch

3. Install Python dependencies:

pip install -r requirements.txt
pip install 'crawl4ai[pdf]'

4. Install Playwright browsers:

playwright install

5. Configure API keys and settings:

Edit configs/default_config.yaml to set up your API keys:

# Serper API Key Pool (for web search), you can obtain free key from https://serper.dev/.
SERPER_API_KEY_POOL:
  - "your_serper_api_key_here"

# LLM Configuration. You should use the model name standard vy liteLLM (https://www.litellm.ai/) to specify the model (recommand to use claude-sonnet-4.5 by defualt). 
llm_config:
  agent_config:
    root_agent_config:
      agent_type: "ResearchAgent"
      model: "openrouter/anthropic/claude-sonnet-4.5"
      temperature: 0.2
      top_p: 0.1

# Custom API configuration. You should set up the base_url and api_key from your api platform (recommand to use openrouter for flexibility and stability). 
  customized_base_url: "your_api_base_url"
  customized_api_key: "your_api_key"

# System language: "English" (fixed to English)
system_language: "English"

# Server mode: "normal" (fixed to normal)
server_mode: "normal"

6. Start the backend server:

python IDR_backend/server.py --host 0.0.0.0 --port 5001

Optional arguments:

• --host: Server host (default: 0.0.0.0)
• --port: Server port (default: 5001)
• --global_config: Path to config file (default: configs/default_config.yaml)

The backend server will start on http://localhost:5001

Frontend Setup

1. Navigate to the frontend directory:

cd IDR_frontend

2. Install dependencies:

npm install

3. Configure backend connection (if needed):

The frontend connects to http://localhost:5001 by default. If you changed the backend port, update IDR_frontend/src/api.tsx:

this.socket = io("http://localhost:5001", {
  // ... connection options
});

4. Start the development server:

npm start

The frontend will open automatically at http://localhost:3000

Usage

1. Start the backend server (see Backend Setup step 6)
2. Start the frontend development server (see Frontend Setup step 4)
3. Open your browser and navigate to http://localhost:3000
4. Enter your research goal in the input box and start exploring!