Skip to content

Readme.md

Latest commit

 

History

History
76 lines (48 loc) · 2.44 KB

Readme.md

File metadata and controls

76 lines (48 loc) · 2.44 KB

DeepResearch

DeepResearch is a multi-agent AI research assistant platform leveraging CrewAI to conduct comprehensive topic research using coordinated agents for web searching, analysis, and writing.

Features

  • Multi-agent orchestration with specialized agents for web search, research analysis, and technical writing.
  • Integration with LinkUp API for live web search results.
  • Interactive Streamlit UI for submitting research queries and displaying structured results.
  • Benchmark logging of request success and elapsed time in benchmark_results.csv.
  • Load testing capabilities to evaluate Streamlit server performance via test.py.

Project Structure

  • agents.py: Implements multi-agent coordination for the research pipeline.
  • linkup.py: Defines the LinkUpSearchTool using LinkUp search API.
  • app.py: Streamlit app for user queries, result display, and benchmarking.
  • research_server.py: MCP server exposing the research tool for external calls.
  • test.py: Load test script for the Streamlit server.
  • benchmark_results.csv: Logs performance metrics per research query.
  • __pycache__/: Bytecode cache (ignored in usage).

Getting Started

Prerequisites

  • Python 3.12 or higher
  • Required packages: CrewAI, Streamlit, Pydantic, etc.
  • API keys (export as environment variables):
    • GEMINI_API_KEY for Gemini LLM
    • LinkUp API key inside linkup.py

Installation

pip install -r requirements.txt

Running the Application

Start the Streamlit interface:

streamlit run app.py

Access the web UI at http://localhost:8501 to enter research questions.

Running the MCP Server

python research_server.py

Starts the research server for tool integration.

Load Testing

python test.py

Runs concurrent requests to benchmark Streamlit server throughput.

Usage

  • Enter a research query in the Streamlit app.
  • Multi-agent system fetches and analyzes web data, then writes a structured, cited document.
  • Performance metrics are logged automatically.

Environment Variables

  • GEMINI_API_KEY: Gemini LLM API key (required).
  • LinkUp_API_KEY: LinkUp API key (required).

Notes

  • LinkUp API integration requires complete client implementation.
  • Agents run in sequential process ensuring detailed and factual research output.
  • Streamlit UI provides server status indicators and debugging tools.

License

Specify your license here (e.g., MIT License).