Debate With Judge Example

import requests
import os

API_BASE_URL = "https://api.swarms.world"
API_KEY = os.environ.get("SWARMS_API_KEY", "your_api_key_here")

headers = {
    "x-api-key": API_KEY,
    "Content-Type": "application/json"
}

def run_debate(topic: str, max_loops: int = 1) -> dict:
    """Run a structured debate with judge evaluation."""

    swarm_config = {
        "name": "Tech Strategy Debate",
        "description": "Structured debate with progressive refinement",
        "swarm_type": "DebateWithJudge",
        "task": topic,
        "agents": [
            {
                "agent_name": "Pro Debater",
                "description": "Argues in favor of the proposition",
                "system_prompt": "You are an expert debater arguing IN FAVOR of the given proposition. Present well-structured arguments with evidence, data, and concrete examples. Anticipate counterarguments and address them. In refinement rounds, strengthen your case based on the judge's feedback.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.5
            },
            {
                "agent_name": "Con Debater",
                "description": "Argues against the proposition",
                "system_prompt": "You are an expert debater arguing AGAINST the given proposition. Identify weaknesses, risks, and unintended consequences. Challenge assumptions with data and real-world counterexamples. In refinement rounds, sharpen your opposition based on the judge's feedback.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.5
            },
            {
                "agent_name": "Debate Judge",
                "description": "Evaluates arguments and provides synthesis",
                "system_prompt": "You are an impartial judge evaluating a structured debate. Assess both sides on logical coherence, evidence quality, and persuasiveness. Identify the strongest points from each side. Provide a clear verdict with justification, synthesizing the best elements from both arguments into a balanced conclusion.",
                "model_name": "gpt-4o",
                "max_loops": 1,
                "temperature": 0.3
            }
        ],
        "max_loops": max_loops
    }

    response = requests.post(
        f"{API_BASE_URL}/v1/swarm/completions",
        headers=headers,
        json=swarm_config,
        timeout=180
    )

    return response.json()

# Define the debate topic
topic = """
Should startups build their initial product on a microservices architecture
or a monolithic architecture? Consider development speed, scalability,
operational complexity, team size constraints, and long-term maintainability.
"""

# Run the debate
result = run_debate(topic)

# Display the debate
for output in result.get("output", []):
    role = output["role"]
    content = output["content"]

    print(f"\n{'='*60}")
    print(f"{role}")
    print(f"{'='*60}")

    if isinstance(content, list):
        content = ' '.join(str(item) for item in content)

    print(str(content)[:800] + "...")

============================================================
Pro Debater
============================================================
Startups should absolutely begin with a monolithic architecture.
Here's the evidence-based case:

1. DEVELOPMENT SPEED: A monolith eliminates inter-service
   communication overhead. Shopify scaled to $5.6B revenue on a
   Ruby on Rails monolith. Instagram had 30M users on a Django
   monolith before acquisition.

2. TEAM SIZE: Startups typically have 2-8 engineers. Microservices
   require dedicated DevOps — a luxury most early-stage teams
   cannot afford. The overhead of managing service discovery,
   API contracts, and distributed debugging slows a small team...

============================================================
Con Debater
============================================================
The monolith-first argument ignores critical modern realities:

1. CLOUD-NATIVE TOOLING: In 2024, frameworks like Next.js API
   routes, AWS Lambda, and Vercel serverless functions make
   microservices nearly as easy to deploy as monoliths. The
   operational overhead argument is outdated.

2. SCALING BOTTLENECKS: Monoliths force you to scale everything
   together. If your search feature needs 10x the compute of
   your auth service, you're paying for 10x across the board...

============================================================
Debate Judge
============================================================
Both sides present compelling arguments. Here is my evaluation:

PRO STRENGTHS: The development speed argument is well-supported
by real examples (Shopify, Instagram). The team size constraint
is a practical reality for most early-stage startups.

CON STRENGTHS: The point about modern cloud tooling reducing
microservices overhead is valid. The scaling cost argument has
merit for compute-intensive applications.

VERDICT: For most startups, begin with a modular monolith...

# Run 2 loops — debaters refine their arguments based on judge feedback
deep_result = run_debate(
    topic="Should AI companies open-source their foundation models? Consider innovation speed, safety risks, competitive moats, and societal benefit.",
    max_loops=2
)

# Show the final contributions after 2 rounds of refinement
for output in deep_result.get("output", []):
    print(f"\n{output['role']}:")
    content = output["content"]
    if isinstance(content, list):
        content = ' '.join(str(item) for item in content)
    print(str(content)[:600] + "...")