AskElira Trader - Autonomous Prediction Market Trading

Built on the AskElira Framework - Multi-agent orchestration with swarm intelligence

Autonomous trading system for Polymarket & Kalshi prediction markets using 5 AI agents + MiroFish swarm intelligence.

🎯 Results

• 65% accuracy overall (200+ predictions)
• 75% accuracy on high-confidence trades (≥80% confidence)
• $0.015 cost per prediction
• Beat market baseline by 12%

🏗️ Framework

AskElira Trader is the trading use case of the AskElira Framework.

Other AskElira Applications:

• AskElira Marketing - Viral marketing campaigns
• AskElira Framework - Build your own automation

Want to build your own? Fork the framework and adapt the 5-agent pattern to your domain.

---

🔮 AskElira

Ask Elira anything. She predicts binary outcomes using 5 AI agents + swarm intelligence.

Will the Lakers win? Will Bitcoin hit $100K? Who wins the election?

Elira researches, simulates crowd behavior, validates, and gives you predictions. Optionally, she can auto-trade on her predictions.

Open source (MIT). Built for prediction markets, adaptable for sports, crypto, futures, forex, or any yes/no outcome.

---

🎯 What Elira Does

Ask her any binary question:

"Ask Elira: Will the Lakers beat the Warriors?"
"Ask Elira: Will Trump win in 2028?"
"Ask Elira: Will Bitcoin reach $100K by June?"
"Ask Elira: Will the Fed cut rates in March?"

How she answers:

1. 🔍 Researches (web search, news, data)
2. 🧠 Simulates (thousands of AI agents predict via swarm intelligence)
3. 🛡️ Validates (quality checks, catches bad logic)
4. 🎯 Predicts (gives you confidence % + direction)
5. 💰 Trades (optional: auto-execute on prediction markets/brokers)

---

⚡ Quick Start

# 1. Clone
git clone https://github.com/jellyforex/askeliratrader.git
cd askeliratrader

# 2. Install
pip install -r requirements.txt

# 3. Setup
cp .env.example .env
# Add your ANTHROPIC_API_KEY to .env

# 4. Start MiroFish (swarm intelligence engine)
cd MiroFish/Mirofish && docker-compose up -d && cd ../..

# 5. Ask Elira anything
./start_paper_trading.sh --once

First prediction takes ~5-8 minutes. Results appear in data/active_positions.json

---

🤖 Meet Elira's Team

Elira orchestrates 5 specialized AI agents:

Agent	Role	What They Do
Alba	Research Analyst	Web search, market scan, news gathering, data collection
David	Simulation Engineer	Runs MiroFish swarm intelligence (thousands of AI agents predict)
Vex	Quality Auditor	Validates predictions, catches bad logic, blocks flawed reasoning
Elira	Decision Maker	Coordinates team, makes final call, talks to you
Steven	Executor	Places trades (if enabled), tracks positions, manages exits

+ MiroFish: Swarm intelligence engine (simulates crowd behavior to predict outcomes)

---

🎨 Two Modes

Mode 1: Prediction Only (Default)

ELIRA_MODE=predict

• Get predictions with confidence scores
• No trading, just insights
• Free (except API costs ~$0.01/prediction)

Example:

You: "Will Bitcoin hit $100K by June?"
Elira: "68% likely YES (based on swarm simulation of 1000 trader agents)"

Mode 2: Auto-Trade (Optional)

ELIRA_MODE=trade
TRADING_MODE=paper  # or 'live' for real money

• Elira auto-executes trades on her predictions
• Paper trading (safe) or real money (requires broker API)
• Confidence-based position sizing

Example:

You: "Will Bitcoin hit $100K by June?"
Elira: "68% YES. Current odds: 2.4x payout. Placing $25 bet (tier 1)."

---

📊 Dashboard

Live pipeline visualization at: http://localhost:3000

cd ~/Desktop/quantjellyfish-dashboard
npm install
npm run dev

Features:

• Real-time agent status
• MiroFish swarm embed
• Activity log (last 50 events)
• Position tracker

---

🏆 Performance

Paper Trading Results (200+ predictions):

• Overall accuracy: 65%
• High-confidence (≥80%): 75%
• Cost per prediction: $0.015
• Beat market baseline: +12%

Latest test run:

• Market: "Will CPI rise more than 0.6% in March 2026?"
• MiroFish result: 80% YES
• Vex verdict: FAIL (semantic drift detected)
• Result: Correctly blocked deployment ✅

---

🔧 Configuration

Environment variables (.env):

# Required
ANTHROPIC_API_KEY=sk-ant-...

# Trading mode
TRADING_MODE=paper  # paper or live
ELIRA_MODE=predict  # predict or trade

# Optional
PINECONE_API_KEY=...
KALSHI_API_KEY_ID=...
MIROFISH_URL=http://localhost:5001

Adjust confidence thresholds:

• elira.py line 106: confidence >= 0.70 (60% for fast testing)
• david.py line 106: min_runs=3 (1 for fast testing)

---

🚀 Architecture

5-agent pipeline:

Alba (Research) 
  ↓
David (MiroFish Simulation)
  ↓
Vex (Adversarial Audit)
  ↓
Elira (Orchestration)
  ↓
Steven (Execution)

Elira's 6-gate validation:

1. Confidence ≥70%
2. Vex verdict = PASS/PASS-WITH-WARNINGS
3. Calendar = CLEAR
4. Liquidity >$500
5. No single-actor override
6. Alba uncertainty ≠ HIGH

Capital tiers:

• Tier 1: $25 (60-79% confidence)
• Tier 2: $50 (80-89% confidence)
• Tier 3: $100 (≥90% + Vex HIGH verdict)

---

📚 Documentation

• Architecture
• Agent Details
• MiroFish Integration
• Dashboard Setup
• Custom Use Cases

---

🤝 Contributing

Contributions welcome! See CONTRIBUTING.md

Areas we need help:

• New market integrations (sports, forex, futures)
• Agent optimization
• Cost reduction strategies
• Alternative swarm implementations

---

📝 License

MIT License - see LICENSE

---

🔗 Links

• Framework: github.com/AskElira/askelira
• Marketing: github.com/AskElira/askeliramarketing
• Website: askelira.com

---

Built with 🧠 by the AskElira Team