AgentField Hackathon: Swarm Quit-Job Intelligence

This project now has:

• AgentField reasoners for due diligence + swarm coordination
• Feedback loop that reweights specialist agents
• Flask frontend for a live demo
• Optional OpenAI opinions (if OPENAI_API_KEY set)
• Optional Tavily news search (if TAVILY_API_KEY set)

What the swarm does

Specialist agents:

• finance_risk_agent
• career_market_agent
• family_stability_agent
• linkedin_positioning_agent
• peer_opinion_agent
• job_search_agent
• news_agent
• knowledge_synth_agent

Memory:

• stored in /Users/lorky/Documents/New project 3/swarm_memory.json
• feedback updates weights so future decisions are influenced by real outcomes

Setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

LLM and news (optional) in .env:

• OPENAI_API_KEY=...
• OPENAI_MODEL=gpt-4o-mini
• TAVILY_API_KEY=...

Run AgentField backend

Terminal A:

af server

Terminal B:

source .venv/bin/activate
python main.py

AgentField reasoners

Import profile

curl -X POST "http://localhost:8080/api/v1/execute/quit-job-due-diligence-agent.import_from_singpass" \
  -H "Content-Type: application/json" \
  -d @sample_singpass_import.json

Swarm recommendation

curl -X POST "http://localhost:8080/api/v1/execute/quit-job-due-diligence-agent.recommend_with_memory" \
  -H "Content-Type: application/json" \
  -d @sample_input.json

Submit outcome feedback

curl -X POST "http://localhost:8080/api/v1/execute/quit-job-due-diligence-agent.submit_feedback" \
  -H "Content-Type: application/json" \
  -d @sample_feedback.json

Run Flask demo frontend

source .venv/bin/activate
python frontend.py

Open:

• http://127.0.0.1:5050

Frontend flow:

1. Your Details: connect LinkedIn + Singpass, run own-agent opinion.
2. Simulated Personas: paste LinkedIn URLs for boss/coworker opinions.
3. Jobs + News Agents: job search and news horizon (Tavily) with opinion.
4. Agentic Swarm + Memory: merges self + peers + jobs + news; memory in swarm_memory.json.

Manual side-investment inputs:

• other_investments_usd
• expected_investment_monthly_income_usd

Notes

• This is decision support, not financial/legal advice.
• For a hackathon demo, this local JSON memory loop is enough to show collective learning.