🔬 Autonomous Deep Research Agent

Production-Grade Multi-Agent Research System

An AI-powered research assistant that autonomously investigates any topic using specialized agents, parallel web searches, and quality-controlled report generation.

Features • Quick Start • Architecture • Usage • API Reference

---

✨ Features

🤖 Multi-Agent Architecture Four specialized AI agents work together: Planner — Creates targeted search strategies Researcher — Executes parallel web searches Critic — Evaluates quality & completeness Writer — Generates structured reports	⚡ High Performance Parallel Execution — 3-5x faster research Smart Caching — SQLite-based result caching Async I/O — Non-blocking operations Rate Limiting — Respects API limits
🔍 Advanced Research Multi-Provider Search — Tavily + Wikipedia + Serper Quality Scoring — 1-10 relevance ratings Fact Checking — Cross-reference validation Iterative Refinement — Auto-improves weak results	🎨 Modern Interface Streamlit Web UI — Beautiful dark theme Real-time Streaming — Live progress updates CLI Support — Full command-line interface REST API — FastAPI with WebSocket

---

🚀 Quick Start

Prerequisites

• Python 3.11 or higher
• Tavily API Key (free tier available)
• Anthropic or OpenAI API key

Installation

# Clone the repository
git clone https://github.com/yourusername/autonomous-research-agent.git
cd autonomous-research-agent

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure API keys
cp .env.example .env
# Edit .env with your API keys

Run the Application

🎨 Web Interface streamlit run app.py Opens at http://localhost:8501

💻 Command Line python main.py "Your research topic"

🌐 API Server uvicorn api.main:app --reload Opens at http://localhost:8000

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                        🎯 WORKFLOW ORCHESTRATOR                      │
│                    (LangGraph State Machine)                         │
└─────────────────────────────────────────────────────────────────────┘
                                    │
        ┌───────────────────────────┼───────────────────────────┐
        ▼                           ▼                           ▼
┌───────────────┐           ┌───────────────┐           ┌───────────────┐
│  📋 PLANNER   │           │ 🔍 RESEARCHER │           │  🔬 CRITIC    │
│               │           │               │           │               │
│ • Analyze     │     ┌────▶│ • Parallel    │           │ • Score       │
│   topic       │     │     │   search      │           │   quality     │
│ • Generate    │─────┘     │ • Multi-      │──────────▶│ • Check       │
│   queries     │           │   provider    │           │   coverage    │
│ • Strategy    │           │ • Rate &      │           │ • Suggest     │
│   planning    │           │   cache       │           │   refinements │
└───────────────┘           └───────────────┘           └───────┬───────┘
                                                                │
                            ┌───────────────┐                   │
                            │  📝 WRITER    │◀──────────────────┘
                            │               │
                            │ • Structure   │
                            │   report      │
                            │ • Citations   │
                            │ • Formatting  │
                            └───────────────┘

Workflow

1. Planning → Planner breaks topic into 3-5 targeted search queries
2. Research → Researcher executes queries in parallel via Tavily + Wikipedia
3. Evaluation → Critic scores quality (completeness, diversity, consistency)
4. Refinement → If score < 7/10, loops back with improvement suggestions
5. Writing → Writer compiles sources into structured markdown report

---

📁 Project Structure

autonomous-research-agent/
│
├── 🎨 app.py                 # Streamlit Web UI
├── 💻 main.py                # CLI Entry Point
├── 📦 pyproject.toml         # Project configuration
│
├── src/                      # Core Package
│   ├── config.py             # Configuration management
│   ├── state.py              # State definitions
│   ├── graph.py              # LangGraph workflow
│   │
│   ├── agents/               # Specialized Agents
│   │   ├── base.py           # Base agent class
│   │   ├── planner.py        # Research planning
│   │   ├── researcher.py     # Parallel search
│   │   ├── critic.py         # Quality evaluation
│   │   └── writer.py         # Report generation
│   │
│   └── tools/                # Utilities
│       ├── search.py         # Search providers
│       └── cache.py          # SQLite caching
│
├── api/                      # REST API
│   └── main.py               # FastAPI + WebSocket
│
├── tests/                    # Test Suite
├── reports/                  # Generated Reports
└── data/                     # Cache Storage

---

💻 Usage

Web Interface

The Streamlit UI provides the most user-friendly experience:

streamlit run app.py

Features:

• 🌙 Modern dark theme with glassmorphism
• 📊 Real-time quality metrics
• 📋 Live agent activity log
• 📥 One-click report download

Command Line

# Basic usage
python main.py "Impact of quantum computing on cryptography"

# With options
python main.py --output ./my_reports --max-revisions 3 "AI in healthcare"

Options:

Flag	Description	Default
--output, -o	Output directory	reports/
--max-revisions, -r	Max refinement loops	2

Programmatic API

from src.graph import run_research

# Run research
result = run_research("Climate change mitigation strategies")

# Access results
print(result["final_report"])
print(f"Quality: {result['quality_report']['overall_score']}/10")
print(f"Sources: {len(result['sources'])}")

---

🌐 API Reference

REST Endpoints

Method	Endpoint	Description
POST	/api/research/start	Start new research session
GET	/api/research/{id}	Get session status
POST	/api/research/{id}/approve	Approve research plan
GET	/api/research/{id}/report	Get final report

WebSocket

const ws = new WebSocket('ws://localhost:8000/ws/research/{session_id}');

ws.onmessage = (event) => {
  const data = JSON.parse(event.data);
  // Types: 'message', 'status', 'plan', 'quality', 'complete'
  console.log(data.type, data.content);
};

---

⚙️ Configuration

Environment Variables

Variable	Description	Required
TAVILY_API_KEY	Tavily search API	✅
ANTHROPIC_API_KEY	Claude API key	One of these
OPENAI_API_KEY	OpenAI API key	required
LLM_PROVIDER	anthropic or openai	Default: anthropic
SERPER_API_KEY	Google Search (optional)	❌

Advanced Configuration

from src.config import get_config

config = get_config()

# Search settings
config.search.max_results_per_query = 10
config.search.max_parallel_searches = 8

# Quality thresholds
config.quality.min_quality_score = 8.0
config.quality.max_refinement_iterations = 3

# Cache settings
config.cache.ttl_hours = 48

---

🧪 Testing

# Run all tests
pytest tests/ -v

# Run with coverage
pytest tests/ --cov=src --cov-report=html

---

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Built with ❤️ using LangGraph, Streamlit, and Tavily

⭐ Star this repo if you find it useful!