Back
Tavily Research Agent

Tavily Research Agent

by Joel Samuel

Deploy w/ Agentuity
PythonWeb SearchTavilyReal-time DataInformation RetrievalSearch APIAgentuity SDK

Tavily Research Agent

What This Agent Does

The Tavily Research Agent is a powerful hybrid research system that combines:

  • Real-time web search using Tavily's advanced search, extract, and crawl tools
  • Internal vector search through your company's CRM data
  • LangGraph ReAct workflow for intelligent multi-step research
  • Comprehensive responses with proper citations and sources

Perfect for business intelligence, competitive analysis, and research tasks that need both public web information and internal enterprise knowledge.

Quick Start Guide

Step 1: Prerequisites

Before you begin, ensure you have:

Step 2: Get API Keys

  • Go to Tavily
  • Sign up for an account
  • Get your API key from the dashboard

Step 3: Authentication

Authenticate with Agentuity:

agentuity login

This opens a browser for you to log in to your Agentuity account.

Step 4: Install Dependencies

The project uses UV for dependency management:

# Install all dependencies automatically
uv sync

This installs all required packages including:

  • agentuity - Agentuity Python SDK
  • langchain - LangChain framework
  • langchain-openai - OpenAI integration
  • langchain-tavily - Tavily search tools
  • langchain-chroma - Vector database for CRM data
  • langgraph - ReAct agent framework

Step 6: Run in Development Mode

Start the agent in development mode:

agentuity dev

This will:

  • Start the agent server on http://localhost:3500
  • Open the Agentuity DevMode in your browser
  • Provide a public URL for testing

How to Use

Via DevMode (Recommended)

  1. After running agentuity dev, click the DevMode URL
  2. In the web interface, type your research question: 
    What's Apple's latest AI strategy?
  3. Hit Send and watch the agent work!

Via API

You can also test via direct API calls:

curl -X POST http://localhost:3500/agents/tavily_agent \
  -H "Content-Type: application/json" \
  -d '{"message": "What are the latest developments in Microsoft Azure AI services?"}'

Example Research Queries

Try these example queries to see the agent in action:

  • "What's Apple's latest AI strategy?"
  • "Find recent earnings reports for Microsoft and check our internal notes"
  • "Research Amazon's cloud revenue growth trends"
  • "What are Google's recent enterprise AI announcements?"
  • "Compare Tesla's autonomous driving progress with competitors"

How It Works

Research Workflow

  1. Input Processing: Accepts natural language research questions
  2. Web Search: Uses Tavily to find current web information
  3. Internal Search: Queries CRM vector database for company data
  4. Analysis: LangGraph ReAct agent processes and synthesizes information
  5. Response: Returns comprehensive answer with citations

Architecture

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│  User Question  │───▶│  Research Agent  │───▶│  Final Response │
└─────────────────┘    └──────────────────┘    └─────────────────┘
                              │
                    ┌─────────┼─────────┐
                    ▼         ▼         ▼
              ┌──────────┐ ┌──────┐ ┌─────────┐
              │ Tavily   │ │ CRM  │ │ LangGraph│
              │ Web      │ │ Vector│ │ ReAct   │
              │ Search   │ │ Store │ │ Agent   │
              └──────────┘ └──────┘ └─────────┘

Project Structure

tavily_agent/
├── agentuity-agents/           # Agent implementations
│   └── tavilyAgent/
│       └── agent.py           # Main agent code                 
├── .env                      # Environment variables (create this)
├── pyproject.toml           # Python dependencies
├── agentuity.yaml          # Agentuity configuration
├── server.py               # Server entry point
└── README.md              # This file

Changing the AI Model

To use a different OpenAI model:

# In agent.py, change the model:
model = ChatOpenAI(
    model="gpt-4o-mini",   # or "gpt-3.5-turbo", etc.
    temperature=0
)

Deployment

When ready to deploy to production:

agentuity deploy

This deploys your agent to the Agentuity Cloud, making it accessible via:

  • REST API endpoints
  • Agentuity Console
  • Integration with other services

Getting Help

If you encounter issues:

  1. Check the Agentuity Documentation
  2. Join our Discord Community
  3. Contact Agentuity support

Monitoring & Logs

Development Logs

When running agentuity dev, you'll see detailed logs:

[INFO] Received text input: What's Apple's latest AI strategy?
[INFO] Processing research request: What's Apple's latest AI strategy?
[INFO] Starting research agent execution...
[INFO] Agent step: HumanMessage
[INFO] Agent step: AIMessage  
[INFO] Research completed successfully

Production Monitoring

Once deployed, monitor your agent through the Agentuity Console:

  • Request/response logs
  • Performance metrics
  • Error tracking
  • Usage analytics

Next Steps

  • Customize the prompts to match your specific research needs
  • Add more data sources to the vector store
  • Integrate with your existing systems via API
  • Deploy to production for team access

📖 Additional Resources

Built with ❤️ using Agentuity - Build Agents, Not Infrastructure