# Sequential execution: 12 seconds
Agent → Task 1 (4s) → Task 2 (5s) → Task 3 (3s) = 12s

# Parallel execution with subagents: 8 seconds (33% faster!)
Agent → [SubAgent 1 (4s) ∥ SubAgent 2 (5s)] → Task 3 (3s) = ~8s

import asyncio
from agent_framework import Agent

async def main():
    # Create an agent
    agent = Agent(model_name="claude-3-5-sonnet-20241022")

    # Run a query
    result = await agent._run_async(
        query="What are the top 3 customer retention strategies?",
        user_id="user_001"
    )

    # Get results
    print(f"Output: {result['output']}")
    print(f"Cost: ${result['metrics']['cost']:.4f}")
    print(f"Tokens: {result['metrics']['tokens_used']}")

if __name__ == "__main__":
    asyncio.run(main())

import asyncio
from agent_framework import Agent

async def main():
    # Create an agent
    agent = Agent(model_name="claude-3-5-sonnet-20241022")

    # Run with automatic subagent parallelization
    result = await agent.run_with_subagents(
        query="Analyze customer churn: fetch data, predict churn, recommend actions",
        user_id="user_001",
        allow_parallelization=True
    )

    # Get parallel results
    print(f"Subagent results: {result['subagent_results']}")
    print(f"Total cost: ${result['metrics']['cost']:.4f}")
    print(f"Duration: {result['metrics']['duration_ms']}ms")

if __name__ == "__main__":
    asyncio.run(main())

# Query: "Analyze customer churn and recommend actions"
#
# Automatic decomposition:
#   Task 1: Query database (independent)
#   Task 2: Run ML model (depends on Task 1)
#   Task 3: Generate recommendations (depends on Task 2)
#
# Execution:
#   - SubAgent A runs Task 1 (4 seconds)
#   - SubAgent B runs Task 2 in parallel? No, it depends on Task 1
#   - Parent waits for Task 1 → runs Task 2 (5 seconds)
#   - Parent runs Task 3 (3 seconds)
#   Total: 12 seconds sequential
#
# With independent tasks:
#   Query: "Analyze churn, predict demand, assess sentiment"
#   - All 3 are independent!
#   - SubAgent A: Churn (4s) ∥ SubAgent B: Demand (5s) ∥ SubAgent C: Sentiment (3s)
#   Total: ~5 seconds (parallel) = 58% faster!

{
  "query": "Analyze customer churn",
  "reasoning": "...",
  "tasks": [...],
  "outcome": "success",
  "cost": 0.15,
  "timestamp": "2025-07-10T12:34:56Z"
}

{
  "text": "How to reduce customer churn",
  "embedding": [0.1, 0.2, ...],
  "topic": "retention",
  "success_rate": 0.92
}

{
  "request_id": "req_abc123",
  "layers": [
    {
      "layer": 1,
      "name": "Input",
      "duration_ms": 5,
      "cost": 0.0,
      "tokens": 0
    },
    {
      "layer": 3,
      "name": "Reasoning",
      "duration_ms": 450,
      "cost": 0.00225,
      "tokens": 250
    },
    ...
  ],
  "total_cost": 0.07089,
  "total_tokens": 1200,
  "total_duration_ms": 3420,
  "status": "success"
}

import asyncio
import aiohttp

async def stream_chat():
    async with aiohttp.ClientSession() as session:
        async with session.post(
            "http://localhost:8000/api/chat",
            json={
                "user_id": "user_001",
                "query": "What are the top 3 products by revenue?"
            }
        ) as resp:
            async for line in resp.content:
                print(line.decode())

asyncio.run(stream_chat())

from agent_framework import Agent, AgentConfig

config = AgentConfig(
    model_name="claude-3-5-sonnet-20241022",
    max_iterations=10,
    max_tokens=4096,
    timeout_ms=30000,
    max_cost_usd=1.0,
    enable_memory=True,
    enable_guardrails=True,
    allow_subagents=True,
    max_subagents=10,
)

agent = Agent(config=config)