Lorsque plusieurs agents IA partagent une clé API de recherche, les coûts peuvent rapidement augmenter sans savoir quel agent dépense quoi. Ce tutoriel construit un suivi centralisé des coûts qui étiquette chaque appel de recherche avec l'ID de l'agent, agrège les dépenses par agent et par période, et alerte lorsqu'un agent dépasse son budget. Fonctionne avec n'importe quel fournisseur de recherche à n'importe quelle échelle.
Prérequis
- Python 3.9+ installé
- bibliothèque requests installée
- Une clé API Scavio depuis scavio.dev
Parcours
Étape 1: Construire le suivi centralisé des coûts
Créez un suivi qui intercepte tous les appels de recherche, les étiquette avec les IDs des agents et enregistre les données de coût.
Python
import time, json
from datetime import datetime, timedelta
from collections import defaultdict
from typing import Dict, List
class AgentCostTracker:
def __init__(self):
self.records: List[dict] = []
self.budgets: Dict[str, float] = {}
self.alerts: List[str] = []
def set_budget(self, agent_id: str, monthly_budget: float):
self.budgets[agent_id] = monthly_budget
def record(self, agent_id: str, cost: float, query: str, provider: str = 'scavio'):
self.records.append({
'agent_id': agent_id,
'cost': cost,
'query': query[:100],
'provider': provider,
'timestamp': datetime.now().isoformat()
})
# Check budget
monthly_spend = self.get_monthly_spend(agent_id)
budget = self.budgets.get(agent_id, float('inf'))
if monthly_spend > budget * 0.8:
alert = f'Agent {agent_id}: {monthly_spend/budget*100:.0f}% of budget used'
if alert not in self.alerts:
self.alerts.append(alert)
print(f'ALERT: {alert}')
def get_monthly_spend(self, agent_id: str) -> float:
month_start = datetime.now().replace(day=1, hour=0, minute=0, second=0)
return sum(r['cost'] for r in self.records
if r['agent_id'] == agent_id and r['timestamp'] >= month_start.isoformat())
tracker = AgentCostTracker()
tracker.set_budget('research-bot', 5.00)
tracker.set_budget('lead-gen-bot', 10.00)
tracker.set_budget('monitor-bot', 2.00)
print('Cost tracker initialized with 3 agent budgets')Étape 2: Créer des fonctions de recherche suivies par agent
Enveloppez l'appel API de recherche avec un suivi automatique des coûts. Chaque agent obtient sa propre fonction de recherche qui enregistre l'utilisation.
Python
import requests, os
SCAVIO_KEY = os.environ['SCAVIO_API_KEY']
COST_PER_QUERY = 0.005
def make_tracked_search(agent_id: str, tracker: AgentCostTracker):
def search(query: str, count: int = 10) -> list:
resp = requests.post('https://api.scavio.dev/api/v1/search',
headers={'x-api-key': SCAVIO_KEY, 'Content-Type': 'application/json'},
json={'query': query, 'country_code': 'us', 'num_results': count})
results = resp.json().get('organic_results', [])
tracker.record(agent_id, COST_PER_QUERY, query)
return [{'title': r['title'], 'link': r['link']}
for r in results]
return search
# Create per-agent search functions
research_search = make_tracked_search('research-bot', tracker)
lead_gen_search = make_tracked_search('lead-gen-bot', tracker)
monitor_search = make_tracked_search('monitor-bot', tracker)
# Simulate agent activity
research_search('python best practices 2026')
research_search('react server components guide')
lead_gen_search('plumber Austin TX')
lead_gen_search('electrician Denver CO')
lead_gen_search('HVAC contractor Portland OR')
monitor_search('scavio brand mentions')
print(f'Total records: {len(tracker.records)}')Étape 3: Générer un rapport de tableau de bord des coûts
Construisez un tableau de bord qui montre les dépenses par agent, l'utilisation du budget et les tendances des coûts.
Python
def dashboard(tracker: AgentCostTracker):
print('\n' + '=' * 60)
print(' Search API Cost Dashboard')
print('=' * 60)
agents = set(r['agent_id'] for r in tracker.records)
total_cost = sum(r['cost'] for r in tracker.records)
total_queries = len(tracker.records)
print(f'\n Total queries: {total_queries}')
print(f' Total cost: ${total_cost:.3f}')
print(f' Agents tracked: {len(agents)}')
print(f'\n {"Agent":<20} {"Queries":>8} {"Spent":>8} {"Budget":>8} {"Used":>6}')
print(' ' + '-' * 54)
for agent_id in sorted(agents):
agent_records = [r for r in tracker.records if r['agent_id'] == agent_id]
spent = sum(r['cost'] for r in agent_records)
budget = tracker.budgets.get(agent_id, 0)
pct = (spent / budget * 100) if budget > 0 else 0
queries = len(agent_records)
print(f' {agent_id:<20} {queries:>8} ${spent:>6.3f} ${budget:>6.2f} {pct:>5.1f}%')
if tracker.alerts:
print(f'\n Alerts:')
for alert in tracker.alerts:
print(f' {alert}')
print(f'\n Avg cost/query: ${total_cost/total_queries:.4f}' if total_queries > 0 else '')
print('=' * 60)
dashboard(tracker)Exemple Python
Python
import requests, os
from collections import defaultdict
SCAVIO_KEY = os.environ['SCAVIO_API_KEY']
spending = defaultdict(float)
counts = defaultdict(int)
def tracked_search(agent_id, query, cost=0.005):
resp = requests.post('https://api.scavio.dev/api/v1/search',
headers={'x-api-key': SCAVIO_KEY, 'Content-Type': 'application/json'},
json={'query': query, 'country_code': 'us', 'num_results': 10})
spending[agent_id] += cost
counts[agent_id] += 1
return resp.json().get('organic_results', [])
tracked_search('research-bot', 'python frameworks 2026')
tracked_search('lead-gen', 'plumber Austin TX')
tracked_search('lead-gen', 'electrician Denver CO')
for agent in spending:
print(f'{agent}: {counts[agent]} queries, ${spending[agent]:.3f}')Exemple JavaScript
JavaScript
const SCAVIO_KEY = process.env.SCAVIO_API_KEY;
const spending = {};
const counts = {};
async function trackedSearch(agentId, query, cost = 0.005) {
const resp = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST',
headers: { 'x-api-key': SCAVIO_KEY, 'Content-Type': 'application/json' },
body: JSON.stringify({ query, country_code: 'us', num_results: 10 })
});
spending[agentId] = (spending[agentId] || 0) + cost;
counts[agentId] = (counts[agentId] || 0) + 1;
return (await resp.json()).organic_results || [];
}
await trackedSearch('research-bot', 'python frameworks 2026');
await trackedSearch('lead-gen', 'plumber Austin TX');
Object.keys(spending).forEach(a => console.log(`${a}: ${counts[a]} queries, $${spending[a].toFixed(3)}`));Sortie attendue
JSON
Cost tracker initialized with 3 agent budgets
Total records: 6
============================================================
Search API Cost Dashboard
============================================================
Total queries: 6
Total cost: $0.030
Agents tracked: 3
Agent Queries Spent Budget Used
------------------------------------------------------
lead-gen-bot 3 $ 0.015 $ 10.00 0.2%
monitor-bot 1 $ 0.005 $ 2.00 0.3%
research-bot 2 $ 0.010 $ 5.00 0.2%
Avg cost/query: $0.0050
============================================================