Quick Start¶

This guide will help you run your first negotiation simulation with AgenticPay.

Running the Example Script¶

The quickest way to try AgenticPay is to run the provided example script:

python agenticpay/examples/single_buyer_product_seller/Task1_basic_price_negotiation.py

This runs a simple negotiation between a buyer and a seller for a single product.

Basic Negotiation Tutorial¶

Step 1: Import Required Modules¶

from agenticpay import make
from agenticpay.agents.buyer_agent import BuyerAgent
from agenticpay.agents.seller_agent import SellerAgent

# For local models
from agenticpay.models.sglang_lm import SGLangLM
# Or: from agenticpay.models.vllm_lm import VLLMLM

Step 2: Initialize the Model¶

# Option 1: SGLang (recommended for single GPU)
model_path = "agenticpay/models/download_models/Qwen3-8B-Instruct"
model = SGLangLM(model_path=model_path)

# Option 2: vLLM (for multi-GPU setups)
# from agenticpay.models.vllm_lm import VLLMLM
# model = VLLMLM(
#     model_path=model_path,
#     trust_remote_code=True,
#     gpu_memory_utilization=0.9,
#     tensor_parallel_size=4,  # Number of GPUs
# )

Step 3: Create Agents¶

Each agent has a confidential “bottom price” that influences their negotiation strategy:

# Buyer's maximum acceptable price (won't pay more than this)
buyer_max_price = 120.0

# Seller's minimum acceptable price (won't sell below this)
seller_min_price = 80.0

# Create agents
buyer = BuyerAgent(model=model, buyer_max_price=buyer_max_price)
seller = SellerAgent(model=model, seller_min_price=seller_min_price)

Step 4: Create the Environment¶

Use the make() function to create a negotiation environment:

env = make(
    "Task1_basic_price_negotiation-v0",
    buyer_agent=buyer,
    seller_agent=seller,
    max_rounds=20,
    initial_seller_price=150.0,
    buyer_max_price=buyer_max_price,
    seller_min_price=seller_min_price,
    environment_info={
        "temperature": "warm",
        "season": "summer",
        "weather": "sunny",
    },
    price_tolerance=0.0,
)

Step 5: Initialize and Run Negotiation¶

# Optional: Define user profile for personalized negotiation
user_profile = "User prefers business/professional style and likes to compare prices."

# Reset environment with product information
observation, info = env.reset(
    user_requirement="I need a high-quality winter jacket",
    product_info={
        "name": "Premium Winter Jacket",
        "brand": "Mountain Gear",
        "price": 180.0,
        "features": ["Waterproof", "Insulated", "Windproof"],
        "condition": "New",
    },
    user_profile=user_profile,
)

# Run negotiation loop
done = False
while not done:
    # Buyer responds first
    buyer_action = buyer.respond(
        conversation_history=observation["conversation_history"],
        current_state=observation
    )

    # Update conversation history
    updated_history = observation["conversation_history"].copy()
    if buyer_action:
        updated_history.append({
            "role": "buyer",
            "content": buyer_action,
            "round": observation.get("current_round", 0)
        })

    # Seller responds
    seller_action = seller.respond(
        conversation_history=updated_history,
        current_state=observation
    )

    # Execute step
    observation, reward, terminated, truncated, info = env.step(
        buyer_action=buyer_action,
        seller_action=seller_action
    )
    done = terminated or truncated

    # Display current state
    env.render()

# Print results
print(f"Negotiation ended: {info['status']}")
print(f"Final price: ${info.get('seller_price', 'N/A')}")
env.close()

Understanding the Output¶

The negotiation can end in several ways:

Deal Reached: Both parties agreed on a price
No Deal: Parties couldn’t agree within max rounds
Buyer Rejected: Buyer explicitly rejected the offer
Seller Rejected: Seller explicitly rejected the offer

The info dictionary contains detailed results:

{
    "status": "deal_reached",      # Outcome status
    "seller_price": 105.0,         # Final agreed price
    "buyer_savings": 15.0,         # Buyer saved from max price
    "seller_profit": 25.0,         # Seller profit above min price
    "rounds_used": 5,              # Number of negotiation rounds
}

Next Steps¶

Learn about Core Concepts to understand the framework architecture
Explore different Environments for various negotiation scenarios
Check out more Examples for advanced use cases