# -*- coding: utf-8 -*-
"""A2A agent generator for load testing."""
import random
import uuid
from datetime import datetime
from typing import Generator, List
from mcpgateway.db import A2AAgent
from ..utils.distributions import exponential_decay_temporal, normal_distribution
from .base import BaseGenerator
class A2AAgentGenerator(BaseGenerator):
"""Generate A2A Agent records with realistic distribution."""
def get_count(self) -> int:
"""Get total number of A2A agents to generate."""
user_count = self.get_scale_config("users", 100)
avg_agents = self.get_scale_config("a2a_agents_per_user_avg", 2)
return int(user_count * avg_agents)
def get_dependencies(self) -> List[str]:
"""A2A agents depend on users and teams."""
return ["UserGenerator", "TeamGenerator"]
def generate(self) -> Generator[A2AAgent, None, None]:
"""Generate A2A agent records.
Yields:
A2AAgent instances
"""
user_count = self.get_scale_config("users", 100)
min_agents = self.get_scale_config("a2a_agents_per_user_min", 0)
max_agents = self.get_scale_config("a2a_agents_per_user_max", 5)
avg_agents = self.get_scale_config("a2a_agents_per_user_avg", 2)
# Get temporal distribution
start_date = datetime.fromisoformat(self.config.get("temporal", {}).get("start_date", "2023-01-01"))
end_date = datetime.fromisoformat(self.config.get("temporal", {}).get("end_date", datetime.now().isoformat()))
recent_percent = self.config.get("temporal", {}).get("recent_data_percent", 80) / 100
# Agents per user distribution
agents_per_user = normal_distribution(user_count, min_agents, max_agents, avg_agents)
total_agents = sum(agents_per_user)
if total_agents == 0:
return
timestamps = exponential_decay_temporal(total_agents, start_date, end_date, recent_percent)
agent_idx = 0
# Agent types and their typical configurations
agent_types = [
("openai", "https://api.openai.com/v1", "1.0"),
("anthropic", "https://api.anthropic.com/v1", "1.0"),
("custom", "https://custom-agent.example.com/api", "1.0"),
("huggingface", "https://api-inference.huggingface.co/models", "1.0"),
]
for user_i in range(user_count):
user_email = f"user{user_i+1}@{self.email_domain}"
num_agents = agents_per_user[user_i]
for j in range(num_agents):
if agent_idx >= len(timestamps):
break
agent_type, base_url, protocol = random.choice(agent_types)
agent_id = str(uuid.uuid4())
name = f"{agent_type}_agent_{user_i+1}_{j+1}"
slug = name.lower().replace("_", "-")
# Capabilities vary by agent type
capabilities = {
"text_generation": True,
"code_generation": random.choice([True, False]),
"image_understanding": random.choice([True, False]) if agent_type in ["openai", "anthropic"] else False,
"function_calling": random.choice([True, False]),
}
# Config varies by agent type
config = {
"model": self._get_model_for_type(agent_type),
"temperature": round(random.uniform(0.0, 1.0), 2),
"max_tokens": random.choice([1000, 2000, 4000, 8000]),
}
agent = A2AAgent(
id=agent_id,
name=name,
slug=slug,
description=self.faker.sentence(),
endpoint_url=f"{base_url}/{name}",
agent_type=agent_type,
protocol_version=protocol,
capabilities=capabilities,
config=config,
auth_type=random.choice(["bearer", "api_key", None]),
auth_value=f"sk-{uuid.uuid4().hex}" if random.random() < 0.8 else None,
enabled=random.random() < 0.9, # 90% enabled
reachable=random.choice([True, True, True, False]), # 75% reachable
created_at=timestamps[agent_idx],
updated_at=timestamps[agent_idx],
last_interaction=timestamps[agent_idx] if random.random() < 0.6 else None,
tags=[],
created_by=user_email,
version=1,
visibility=random.choice(["public", "private", "team"]),
)
agent_idx += 1
yield agent
def _get_model_for_type(self, agent_type: str) -> str:
"""Get a realistic model name for the agent type."""
models = {
"openai": ["gpt-4", "gpt-4-turbo", "gpt-3.5-turbo"],
"anthropic": ["claude-3-opus", "claude-3-sonnet", "claude-3-haiku"],
"custom": ["custom-model-v1", "custom-model-v2"],
"huggingface": ["meta-llama/Llama-2-7b", "mistralai/Mistral-7B"],
}
return random.choice(models.get(agent_type, ["default-model"]))
