Group Chat 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_product_brainstorm(product_brief: str) -> dict:
    """Run a collaborative product strategy discussion."""

    swarm_config = {
        "name": "Product Strategy GroupChat",
        "description": "Cross-functional product brainstorm",
        "swarm_type": "GroupChat",
        "task": f"""Collaborate on a go-to-market strategy for the following product:

{product_brief}

Each participant should contribute their domain expertise, build on others' ideas, and identify blind spots. Converge on a concrete action plan.""",
        "agents": [
            {
                "agent_name": "Product Manager",
                "description": "Drives product vision, roadmap, and prioritization",
                "system_prompt": """You are a Senior Product Manager. In this group discussion:
- Define the core value proposition and target personas
- Prioritize features for the launch MVP
- Identify key metrics and success criteria
- Challenge assumptions and ask clarifying questions

Be concise and action-oriented. Build on what other participants say.""",
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "temperature": 0.5
            },
            {
                "agent_name": "Growth Marketer",
                "description": "Designs acquisition channels and launch campaigns",
                "system_prompt": """You are a Growth Marketing Lead. In this group discussion:
- Propose acquisition channels ranked by expected ROI
- Design the launch campaign strategy
- Suggest pricing and positioning tactics
- Estimate CAC benchmarks for each channel

Be data-driven and specific with numbers. React to others' suggestions.""",
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "temperature": 0.5
            },
            {
                "agent_name": "Engineering Lead",
                "description": "Assesses technical feasibility and delivery timelines",
                "system_prompt": """You are an Engineering Lead. In this group discussion:
- Evaluate technical feasibility of proposed features
- Flag complexity risks and dependencies
- Propose a realistic MVP scope and timeline
- Suggest technical shortcuts for faster time-to-market

Be pragmatic. Push back on scope creep and offer alternatives.""",
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "temperature": 0.4
            },
            {
                "agent_name": "Customer Success Lead",
                "description": "Represents the customer voice and retention strategy",
                "system_prompt": """You are a Customer Success Lead. In this group discussion:
- Advocate for the end-user experience
- Identify onboarding friction points
- Propose retention and engagement hooks
- Share common objections and how to address them

Ground the discussion in real customer needs. Challenge ideas that overlook usability.""",
                "model_name": "gpt-4.1",
                "role": "worker",
                "max_loops": 1,
                "temperature": 0.5
            }
        ],
        "max_loops": 1
    }

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

    return response.json()

# Product brief
brief = """
PRODUCT: TaskPilot - AI-powered project management tool for remote teams
TARGET: Small-to-mid engineering teams (5-30 people)
KEY FEATURES:
- AI auto-assigns tasks based on team capacity and skill matching
- Natural language standup summaries generated from activity
- Smart sprint planning that predicts delivery risk
- Integrations with GitHub, Slack, and Jira

STAGE: Pre-launch, MVP built, 12 beta users
BUDGET: $50K for initial go-to-market
TIMELINE: 8 weeks to public launch
"""

# Run brainstorm
result = run_product_brainstorm(brief)

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

    print(f"\n{'='*50}")
    print(f"{participant.upper()}")
    print(f"{'='*50}")
    print(content[:800] + "...")

==================================================
PRODUCT MANAGER
==================================================
Based on the brief, here's my take on the launch strategy:

VALUE PROPOSITION: "Stop managing tasks. Let AI manage them for you."
Target persona: Engineering managers at remote-first startups (5-30 devs)
who currently use Jira/Linear but struggle with sprint planning accuracy.

MVP PRIORITY (Must-have for launch):
1. AI task assignment - this is the core differentiator
2. Slack integration - lowest friction adoption path
3. Standup summaries - immediate daily value

DEFER post-launch:
- Jira migration (complex, slow ROI)
- Delivery risk prediction (needs more data)

SUCCESS METRICS:
- 50 teams onboarded in 8 weeks
- 60% WAU after first month
- NPS > 40 from beta cohort...

==================================================
GROWTH MARKETER
==================================================
Love the positioning. Here's my channel strategy for the $50K budget:

CHANNEL ALLOCATION:
1. Product Hunt launch ($2K) - Expected: 500-1K signups, best ROI
2. Dev-focused content marketing ($15K) - Blog posts, Twitter threads
3. Slack community partnerships ($8K) - Sponsor remote-work communities
4. LinkedIn ads targeting eng managers ($15K) - Est. CAC: $35-50
5. Beta user referral program ($10K) - Give existing 12 users incentives

PRICING:
- Free tier: up to 5 users (growth loop)
- Pro: $12/user/month (target ACV ~$2,500)
- Launch discount: 40% off annual for first 3 months...

==================================================
ENGINEERING LEAD
==================================================
A few reality checks on the proposed scope:

FEASIBLE IN 8 WEEKS:
- AI task assignment: Core algo exists from MVP, needs polish. 3 weeks.
- Slack integration: Standard OAuth + webhooks. 2 weeks.
- Standup summaries: We can ship a v1 using activity logs. 1 week.

RISKS:
- GitHub integration adds 2-3 weeks if we want bidirectional sync.
  Recommend: start with read-only GitHub status sync (1 week).
- AI assignment accuracy depends on historical data. New teams
  will have a cold-start problem. Suggest: manual override + learning mode.

TIMELINE: 6 weeks eng + 2 weeks buffer = doable if we freeze scope NOW...

==================================================
CUSTOMER SUCCESS LEAD
==================================================
From our 12 beta users, here's what I'm seeing:

TOP ONBOARDING FRICTION:
1. Teams don't trust AI assignments on day one - need a "suggestion mode"
   before fully autonomous mode
2. Connecting Slack takes 3 clicks but users expect 1-click setup
3. No clear "aha moment" in first 5 minutes

RETENTION HOOKS I'd prioritize:
- Daily standup digest in Slack (passive value, no effort needed)
- Weekly "time saved" report (quantifies ROI for the buyer)
- Team leaderboard for sprint completion (engagement driver)

COMMON OBJECTION: "How is this different from Jira automation rules?"
Answer: We need a clear comparison page and a 2-minute demo video...