ResearchMind

Four specialized AI agents collaborate to deliver polished research reports on any topic.

What It Does

ResearchMind is a multi-agent research pipeline that:

• searches the web for reliable, recent information,
• reads and extracts deeper source content,
• writes a structured report,
• critiques the final output with a score and improvement suggestions.

This system is designed as a practical agentic AI workflow using LangChain and Streamlit.

Pipeline Overview

Agent Roles

1. Search Agent
   Uses Tavily-powered web search to find relevant live sources.
2. Reader Agent
   Selects and scrapes high-value pages using BeautifulSoup-based tooling.
3. Writer Chain
   Synthesizes gathered context into a professional, structured report.
4. Critic Chain
   Reviews report quality, assigns a score, and returns targeted feedback.

Execution Flow

flowchart LR
    A[User Topic Input] --> B[Search Agent]
    B --> C[State Memory: Search Results]
    C --> D[Reader Agent]
    D --> E[State Memory: Scraped Content]
    E --> F[Writer Chain]
    F --> G[Draft Report]
    G --> H[Critic Chain]
    H --> I[Scored Final Output + Feedback]

Loading

Runtime Sequence

sequenceDiagram
    participant U as User
    participant S as Search Agent
    participant R as Reader Agent
    participant W as Writer Chain
    participant C as Critic Chain
    participant M as Pipeline State

    U->>S: Provide research topic
    S->>M: Save search results
    M->>R: Pass top results context
    R->>M: Save scraped content
    M->>W: Provide combined research
    W->>M: Save generated report
    M->>C: Send report for review
    C->>U: Return score + feedback

Loading

Architecture at a Glance

Layer	Responsibility	Technology
Interface	Single-page app, pipeline progress, report actions	Streamlit
Agent Orchestration	Multi-step task decomposition and execution	LangChain
Search Tooling	Live web lookup for relevant sources	Tavily
Reading Tooling	URL scraping and content extraction	BeautifulSoup + Requests
Report Synthesis	Structured report generation	LLM chain
Quality Control	Critique, scoring, improvement suggestions	LLM chain

UI Screenshots

Main Interface

Research Results Section

Key Features

• Multi-agent collaboration with clear responsibility boundaries
• Stateful step-by-step pipeline (search -> reader -> writer -> critic)
• Progressive UX with modern Streamlit interface
• Expandable raw outputs for transparency and debugging
• Download/copy support for generated reports

Project Structure

multi_agent_ai_research_system/
├── app.py
├── pipeline.py
├── agents.py
├── tools.py
├── requirements.txt
├── public/
│   └── screenshots/
│       ├── Research-mind-main-page.jpeg
│       └── Research-section_researchmind.jpeg
└── README.md

Quick Start

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
streamlit run app.py

Environment Variables

Add required API keys in .env (for example OpenAI and Tavily), then run the app.

Why This Project

This project demonstrates production-style agentic design:

• modular and composable agent architecture,
• practical state passing between agents,
• quality loop through critic-based review,
• clean UX for real users and portfolio demonstration.