Agregar búsqueda web a un agente de DeerFlow le brinda acceso a datos en vivo para tareas de investigación en lugar de depender de conocimientos de capacitación estáticos. DeerFlow es un marco de investigación profundo que organiza flujos de trabajo de investigación de varios pasos, pero necesita una herramienta de búsqueda lista para usar para recopilar evidencia externa. Este tutorial registra una herramienta de búsqueda Scavio en un agente DeerFlow, la conecta al flujo de investigación y analiza los datos SERP estructurados en el formato de contexto del agente.
Requisitos previos
- Python 3.10+
- DeerFlow instalado (pip install deerflow)
- Clave API de Scavio de scavio.dev
- Comprensión básica del registro de herramientas de agentes
Guia paso a paso
Paso 1: Crear la función de herramienta de búsqueda
Defina una función de búsqueda que llame a la API de Scavio y devuelva resultados en el formato que espera DeerFlow.
Python
import os, requests
H = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}
def web_search(query: str, num_results: int = 5) -> list[dict]:
'''Search the web and return structured results.'''
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=H, json={'query': query, 'country_code': 'us'}).json()
results = []
for r in data.get('organic_results', [])[:num_results]:
results.append({
'title': r.get('title', ''),
'url': r.get('link', ''),
'content': r.get('snippet', ''),
})
return resultsPaso 2: Registre la herramienta con DeerFlow
Registre la función de búsqueda como herramienta en la configuración del agente DeerFlow para poder llamarla durante los flujos de investigación.
Python
from deerflow import DeerFlowAgent, Tool
search_tool = Tool(
name='web_search',
description='Search the web for current information. Use for fact-checking, finding recent data, and research.',
function=web_search,
parameters={
'query': {'type': 'string', 'description': 'The search query', 'required': True},
'num_results': {'type': 'integer', 'description': 'Number of results to return', 'default': 5},
}
)
agent = DeerFlowAgent(
name='research_agent',
tools=[search_tool],
model='gpt-4o',
system_prompt='You are a research agent. Use web_search to find current information before answering.',
)Paso 3: Agregar capacidad de búsqueda multiplataforma
Amplíe la herramienta para realizar búsquedas en Google, Reddit y YouTube para obtener una cobertura de investigación completa.
Python
def multi_search(query: str, platforms: list[str] = None) -> dict:
'''Search across multiple platforms for comprehensive research.'''
if platforms is None:
platforms = ['google', 'reddit']
all_results = {}
for platform in platforms:
params = {'query': query, 'country_code': 'us'}
if platform != 'google':
params['platform'] = platform
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=H, json=params).json()
all_results[platform] = [{
'title': r.get('title', ''),
'url': r.get('link', ''),
'content': r.get('snippet', ''),
} for r in data.get('organic_results', [])[:3]]
return all_results
multi_search_tool = Tool(
name='multi_search',
description='Search multiple platforms (google, reddit, youtube) for research.',
function=multi_search,
parameters={
'query': {'type': 'string', 'required': True},
'platforms': {'type': 'array', 'items': {'type': 'string'}, 'default': ['google', 'reddit']},
}
)Paso 4: Ejecutar una tarea de investigación
Ejecute una consulta de investigación y observe al agente utilizando herramientas de búsqueda para recopilar evidencia antes de sintetizar una respuesta.
Python
async def run_research():
agent = DeerFlowAgent(
name='research_agent',
tools=[search_tool, multi_search_tool],
model='gpt-4o',
system_prompt='You are a research agent. Always search before answering. Cite sources.',
)
result = await agent.run(
'What are the top 3 LLM agent frameworks in 2026 and how do they compare?'
)
print(result.answer)
print(f'\nTools called: {len(result.tool_calls)}')
for call in result.tool_calls:
print(f' - {call.tool_name}({call.arguments.get("query", "")})')
import asyncio
asyncio.run(run_research())Ejemplo en Python
Python
import os, requests, asyncio
from deerflow import DeerFlowAgent, Tool
H = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}
def web_search(query: str, num_results: int = 5) -> list[dict]:
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=H, json={'query': query, 'country_code': 'us'}).json()
return [{'title': r.get('title', ''), 'url': r.get('link', ''),
'content': r.get('snippet', '')}
for r in data.get('organic_results', [])[:num_results]]
def multi_search(query: str, platforms: list[str] = None) -> dict:
platforms = platforms or ['google', 'reddit']
results = {}
for p in platforms:
params = {'query': query, 'country_code': 'us'}
if p != 'google': params['platform'] = p
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=H, json=params).json()
results[p] = [{'title': r.get('title',''), 'url': r.get('link',''),
'content': r.get('snippet','')} for r in data.get('organic_results',[])[:3]]
return results
async def main():
agent = DeerFlowAgent(
name='researcher',
tools=[
Tool(name='web_search', description='Search the web', function=web_search,
parameters={'query': {'type':'string','required':True}}),
Tool(name='multi_search', description='Multi-platform search', function=multi_search,
parameters={'query': {'type':'string','required':True},
'platforms': {'type':'array','default':['google','reddit']}}),
],
model='gpt-4o',
system_prompt='Research agent. Search before answering. Cite sources.',
)
result = await agent.run('Compare LangGraph vs CrewAI vs AutoGen in 2026')
print(result.answer)
asyncio.run(main())Ejemplo en JavaScript
JavaScript
const H = {'x-api-key': process.env.SCAVIO_API_KEY, 'Content-Type': 'application/json'};
async function webSearch(query, numResults = 5) {
const data = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST', headers: H,
body: JSON.stringify({query, country_code: 'us'})
}).then(r => r.json());
return (data.organic_results || []).slice(0, numResults).map(r => ({
title: r.title || '', url: r.link || '', content: r.snippet || ''
}));
}
async function multiSearch(query, platforms = ['google', 'reddit']) {
const results = {};
for (const p of platforms) {
const params = {query, country_code: 'us'};
if (p !== 'google') params.platform = p;
const data = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST', headers: H, body: JSON.stringify(params)
}).then(r => r.json());
results[p] = (data.organic_results || []).slice(0, 3).map(r => ({
title: r.title, url: r.link, content: r.snippet
}));
}
return results;
}
// Register with DeerFlow (JS SDK)
// const agent = new DeerFlowAgent({tools: [{name: 'web_search', fn: webSearch}]});
console.log('DeerFlow search tools ready');
webSearch('LangGraph vs CrewAI 2026').then(r => console.log(\`\${r.length} results\`));Salida esperada
JSON
Research agent output:
- LangGraph: Best for stateful, cyclic agent workflows. Most GitHub stars in 2026.
- CrewAI: Best for multi-agent orchestration with role-based agents.
- AutoGen: Best for conversational multi-agent patterns.
Tools called: 3
- web_search(LLM agent frameworks comparison 2026)
- multi_search(LangGraph vs CrewAI vs AutoGen)
- web_search(agent framework github stars 2026)