Tavily Search

Built-in Tavily search tools for AI agents - web search, content extraction, crawling, and site mapping

from praisonaiagents import Agent, tavily

agent = Agent(name="Researcher", tools=[tavily])
agent.start("Search for the latest on agent frameworks")

The user describes a research question; the agent calls Tavily search and returns concise findings.

Prerequisites

• Python 3.10 or higher
• PraisonAI Agents package installed
• tavily-python package installed
• TAVILY_API_KEY environment variable set

Tavily is an AI-powered search API optimized for LLM applications. PraisonAI includes built-in Tavily tools for easy integration.

Quick Start

1. Install and set key

   pip install praisonaiagents tavily-python
   export TAVILY_API_KEY=your_tavily_api_key
   
2. Search with agent

   from praisonaiagents import Agent, tavily
   
   agent = Agent(
       name="SearchAgent",
       instructions="Search the web for accurate, up-to-date information.",
       tools=[tavily],
   )
   
   agent.start("What are the latest AI trends in 2025?")
   

Installation

pip install praisonaiagents tavily-python

Setup

export TAVILY_API_KEY=your_tavily_api_key
export OPENAI_API_KEY=your_openai_api_key

Built-in Tavily Tool

PraisonAI provides a built-in tavily tool that you can import directly:

from praisonaiagents import Agent
from praisonaiagents import tavily

agent = Agent(
    name="SearchAgent",
    role="Web Researcher",
    goal="Find information on the web",
    tools=[tavily]
)

result = agent.start("What are the latest AI trends in 2025?")
print(result)

Available Functions

Function	Description
tavily	Web search (alias for tavily_search)
tavily_search	Search with full parameters
tavily_extract	Extract content from URLs
tavily_crawl	Crawl websites
tavily_map	Get site maps

Basic Usage

Simple Search

from praisonaiagents import tavily

# Simple search
results = tavily("Python programming best practices")
print(results)

Search with Options

from praisonaiagents import tavily_search

results = tavily_search(
    query="AI trends 2025",
    max_results=5,
    search_depth="advanced",  # "basic" or "advanced"
    include_answer=True,      # Get LLM-generated answer
    topic="general"           # "general", "news", or "finance"
)

print(results.get("answer"))  # AI-generated answer
for r in results.get("results", []):
    print(f"- {r['title']}: {r['url']}")

Extract Content from URLs

from praisonaiagents import tavily_extract

content = tavily_extract(
    urls=["https://praison.ai/docs", "https://example.com"],
    extract_depth="basic"  # "basic" or "advanced"
)

for result in content.get("results", []):
    print(f"URL: {result['url']}")
    print(f"Content: {result['raw_content'][:500]}...")

Crawl Websites

from praisonaiagents import tavily_crawl

results = tavily_crawl(
    url="https://praison.ai/docs",
    max_depth=2,
    max_breadth=10,
    limit=20,
    instructions="Find all pages about agents"
)

for page in results.get("results", []):
    print(f"- {page['url']}")

Get Site Map

from praisonaiagents import tavily_map

sitemap = tavily_map(
    url="https://praison.ai/docs",
    max_depth=2,
    limit=50
)

for url in sitemap.get("results", []):
    print(url)

With PraisonAI Agent

from praisonaiagents import Agent
from praisonaiagents import tavily

agent = Agent(
    name="SearchAgent",
    role="Web Researcher",
    goal="Find and analyze information from the web",
    tools=[tavily]
)

result = agent.start("Research the latest developments in quantum computing")
print(result)

Using TavilyTools Class

For more control, use the TavilyTools class directly:

from praisonaiagents import TavilyTools

# Initialize with custom API key (optional)
tools = TavilyTools(api_key="your_api_key")  # or uses TAVILY_API_KEY env var

# Search
results = tools.search("AI news", max_results=5, include_answer=True)

# Extract
content = tools.extract(["https://example.com"])

# Crawl
pages = tools.crawl("https://praison.ai/docs", max_depth=2)

# Map
sitemap = tools.map("https://praison.ai/docs")

Async Usage

import asyncio
from praisonaiagents import Agent, Task, AgentTeam
from praisonaiagents import tavily_search_async

agent = Agent(
    name="AsyncSearchAgent",
    role="Search Specialist",
    tools=[tavily_search_async]
)

task = Task(
    description="Search for AI trends",
    agent=agent,
    async_execution=True
)

agents = AgentTeam(agents=[agent], tasks=[task])
result = asyncio.run(agents.astart())

Search Parameters

Parameter	Type	Description
query	str	Search query
search_depth	str	”basic” or “advanced”
topic	str	”general”, “news”, or “finance”
max_results	int	Max results (1-20)
include_answer	bool	Include LLM-generated answer
include_raw_content	bool	Include full page content
include_images	bool	Include images
include_domains	list	Domains to include
exclude_domains	list	Domains to exclude
time_range	str	”day”, “week”, “month”, “year”

Key Points

• Simple function signature: Tool must accept query: str and return str
• Environment variable: Set TAVILY_API_KEY before running
• Agent decides: The LLM decides when to use the tool based on the query
• Works globally: --query-rewrite works with any PraisonAI command

Best Practices

Use the simple tavily import

from praisonaiagents import tavily is the simplest integration - use it unless you need specific parameters.

Set search_depth based on task

Use search_depth='basic' for quick lookups and 'advanced' for research needing deeper results.

Enable include_answer for summaries

Set include_answer=True to get a Tavily-generated summary alongside raw search results.

Use topic filter for domain-specific search

Set topic='news' for current events or topic='finance' for financial queries.

Related

Custom Tools

Build your own agent tools

Tools Overview

Browse PraisonAI tool documentation