Les agents Hermes dépendent de navigateurs sans tête ou de SearXNG pour la recherche web, ce qui est lent, fragile et gourmand en ressources. Ce tutoriel remplace la recherche basée sur un navigateur par un appel API direct. L'agent obtient des résultats structurés en 300 ms au lieu de 3 à 5 secondes, sans Puppeteer, sans Chrome et sans dépendance Docker. Chaque recherche coûte 0,005 $.
Prérequis
- Python 3.8+
- bibliothèque requests
- Une clé API Scavio depuis scavio.dev
- Framework agent Hermes installé
Parcours
Étape 1: Créer l'outil de recherche sans navigateur
Construire une fonction de recherche qui renvoie des résultats structurés sans aucune dépendance de navigateur.
Python
import os, requests, json
API_KEY = os.environ['SCAVIO_API_KEY']
SH = {'x-api-key': API_KEY, 'Content-Type': 'application/json'}
def web_search(query, num_results=5):
"""Browser-free web search for Hermes agents.
Returns structured results in ~300ms vs 3-5s with headless Chrome.
"""
resp = requests.post('https://api.scavio.dev/api/v1/search',
headers=SH, json={'query': query, 'country_code': 'us'}, timeout=10)
resp.raise_for_status()
data = resp.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', ''),
})
# Include People Also Ask for richer context
paa = [q.get('question', '') for q in data.get('people_also_ask', [])[:3]]
return {'results': results, 'related': paa, 'count': len(results)}
# Test - no browser needed
result = web_search('python async database tutorial 2026')
print(f'Results: {result["count"]} (no browser, no Chrome, no Docker)')
for r in result['results'][:3]:
print(f' {r["title"][:50]}')
print(f' {r["url"]}')
if result['related']:
print(f'\nRelated questions:')
for q in result['related']:
print(f' - {q}')
print(f'\nCost: $0.005/search')Étape 2: Enregistrer en tant qu'outil Hermes
Envelopper la fonction de recherche en tant qu'outil que Hermes peut découvrir et invoquer.
Python
HERMES_SEARCH_TOOL = {
'type': 'function',
'function': {
'name': 'web_search',
'description': 'Search the web for current information. Use when you need to find documentation, verify facts, check APIs, or answer questions requiring up-to-date data. No browser required.',
'parameters': {
'type': 'object',
'properties': {
'query': {
'type': 'string',
'description': 'The search query. Be specific for better results.'
},
'num_results': {
'type': 'integer',
'description': 'Number of results to return (1-10, default 5)',
'default': 5
}
},
'required': ['query']
}
}
}
def handle_hermes_tool(name, args):
"""Handle tool calls from Hermes agent."""
if name == 'web_search':
query = args.get('query', '')
num = args.get('num_results', 5)
result = web_search(query, num)
return json.dumps(result)
return json.dumps({'error': f'Unknown tool: {name}'})
# Simulate Hermes calling the tool
print('Hermes tool registered: web_search')
print('Dependencies removed: puppeteer, chromium, selenium')
result = handle_hermes_tool('web_search', {'query': 'FastAPI deployment best practices'})
parsed = json.loads(result)
print(f'\nTool returned {parsed["count"]} results')Étape 3: Remplacer la recherche par navigateur dans la configuration Hermes
Mettre à jour la configuration de l'agent Hermes pour utiliser la recherche API au lieu de la recherche par navigateur.
Python
# BEFORE: Hermes with browser search (in hermes_config.py)
# search_tool = BrowserSearchTool()
# Requires: puppeteer, chromium, 500MB+ disk, 3-5s per search
# AFTER: Hermes with API search
# search_tool = web_search # The function we built above
# Requires: requests library only, 300ms per search
def demo_hermes_agent():
"""Demonstrate Hermes agent using API search instead of browser."""
print('Hermes Agent Starting...')
print(' Search backend: Scavio API (no browser)')
print(' Dependencies: requests only')
print(' Avg latency: ~300ms')
# Agent task: research a topic
task = 'Find the best Python web framework for async APIs in 2026'
print(f'\nTask: {task}')
# Step 1: Search
print('\n Step 1: Searching...')
result = web_search(task)
print(f' Found {result["count"]} results in ~300ms')
# Step 2: Process results
print('\n Step 2: Processing top results...')
for r in result['results'][:3]:
print(f' - {r["title"][:50]}')
# Step 3: Follow-up search
if result['related']:
print(f'\n Step 3: Follow-up search...')
followup = web_search(result['related'][0])
print(f' Follow-up: {followup["count"]} more results')
print(f'\n Total cost: $0.010 (2 searches)')
print(f' Browser equivalent: 6-10 seconds, 500MB RAM')
print(f' API equivalent: 600ms, 0MB extra RAM')
demo_hermes_agent()Exemple Python
Python
import os, requests
SH = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}
def hermes_search(query):
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=SH, json={'query': query, 'country_code': 'us'}, timeout=10).json()
return [{'title': r['title'], 'url': r['link']} for r in data.get('organic_results', [])[:5]]
for r in hermes_search('python async database'):
print(f'{r["title"][:50]}')
print('No browser. $0.005/search.')Exemple JavaScript
JavaScript
const SH = { 'x-api-key': process.env.SCAVIO_API_KEY, 'Content-Type': 'application/json' };
const data = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST', headers: SH,
body: JSON.stringify({ query: 'python async database', country_code: 'us' })
}).then(r => r.json());
(data.organic_results || []).slice(0, 5).forEach(r => console.log(r.title));
console.log('No browser needed.');Sortie attendue
JSON
Results: 5 (no browser, no Chrome, no Docker)
FastAPI with Async SQLAlchemy - Complete Guide
https://fastapi.tiangolo.com/advanced/async-sql/
Python Async Database Tutorial 2026 - Real Python
https://realpython.com/async-database-python/
Hermes Agent Starting...
Search backend: Scavio API (no browser)
Dependencies: requests only
Avg latency: ~300ms
Task: Find the best Python web framework for async APIs in 2026
Total cost: $0.010 (2 searches)
Browser equivalent: 6-10 seconds, 500MB RAM
API equivalent: 600ms, 0MB extra RAM