vequil-sdk 0.2.0

Multi-agent prediction market trading framework. Deploy autonomous agent teams to trade prediction markets.

V E Q U I L

Multi-agent prediction market trading framework.

---

Install

pip install vequil

Quick Start

from vequil import Pipeline

pipeline = Pipeline(bankroll=500.0)
result = await pipeline.evaluate(
    "Will Liverpool win?",
    odds=0.65,
    model_prob=0.78,
)

print(result.recommendation)   # STRONG — edge=13.0%, kelly=20.0%
print(result.position_size_usdc)  # $100.00

CLI

# Scan top Polymarket markets by volume
vequil scan --limit 10

# Evaluate a specific opportunity
vequil evaluate "Will Liverpool win?" --odds 0.65 --model-prob 0.78

# Run a live demo (pulls real Polymarket data)
vequil demo

With Analyst Team (LLM debate)

import os
os.environ["ANTHROPIC_API_KEY"] = "sk-..."

from vequil import Pipeline

pipeline = Pipeline(bankroll=500.0, enable_analysts=True)
result = await pipeline.evaluate("Will Arsenal win?", odds=0.72)

# LLM analysts evaluate in parallel, bull/bear researchers debate
print(result.research.consensus_signal)  # "bullish"
print(result.research.bullish_case)
print(result.research.bearish_case)

---

What is Vequil?

Vequil is a multi-agent prediction market trading framework. Deploy autonomous agent teams to evaluate and trade prediction market opportunities with deterministic signal rules and Kelly-constrained risk management.

No learned models in the execution path. Every decision logged and auditable.

Vequil the framework is what you install. Vequil the firm is how we use it.

---

Architecture

    ┌─────────────┐  ┌──────────────┐   ┌──────────────────┐
    │  Feed Agent  │  │ Sports Agent │   │   Analyst Team   │
    │  tick stream │  │ match events │   │   LLM analysts   │
    └──────┬──────┘  └──────┬───────┘   │  bull/bear debate │
           ▼                │           └────────┬─────────┘
    ┌──────────────┐        │            soft context only
    │Feature Agent │        │           (never overrides Kelly)
    │ vol, z-score │        │                    │
    └──────┬───────┘        │                    │
           ▼                ▼                    ▼
    ┌──────────────────────────────────────────────────┐
    │              DETERMINISTIC DECISION               │
    │  features → closed-form rule → probability shift  │
    │  NO learned model · NO curve fitting              │
    └──────────────────────┬───────────────────────────┘
                           ▼
    ┌──────────────────────────────────────────────────┐
    │                Scanner Agent                      │
    │  Polymarket CLOB → match opportunities → score    │
    └──────────────────────┬───────────────────────────┘
                           ▼
    ┌──────────────────────────────────────────────────┐
    │                  Risk Agent                       │
    │  Kelly criterion · position limits · breaker      │
    └──────────────────────┬───────────────────────────┘
                           ▼
    ┌──────────────────────────────────────────────────┐
    │              Execution Agent                      │
    │  paper / live modes · full order audit trail       │
    └──────────────────────────────────────────────────┘

---

Repository Structure

vequil/                        pip package (what you install)
├── __init__.py                public API
├── pipeline.py                Pipeline — the main evaluation API
├── cli.py                     CLI entry point (vequil scan/evaluate/demo)
├── core/
│   ├── models.py              frozen dataclasses — Tick, Signal, Order, etc.
│   ├── kelly.py               Kelly criterion with fractional cap
│   └── pricing.py             deterministic signal rule
└── agents/
    ├── analyst_team.py        LLM analysts + bull/bear researcher debate
    ├── feed_agent.py          dual exchange WebSocket feed
    ├── feature_agent.py       rolling features: vol, return, jump, z-score
    ├── sports_agent.py        live sports event feed
    ├── scanner_agent.py       Polymarket CLOB matching + edge scoring
    ├── risk_agent.py          position limits, drawdown circuit breaker
    └── execution_agent.py     paper/live orders, SQLite persistence

trading/                       internal orchestration
tests/                         22 tests, all passing
web/static/                    vequil.com

---

Risk Controls

Control	Description
Kelly cap	Fractional Kelly with configurable cap
Min edge	Minimum edge threshold to enter
Max positions	Concurrent position limit
Position cap	Per-position dollar limit
Circuit breaker	Halts trading on daily drawdown
Live gate	Disabled until validated performance threshold

---

Environment Variables

Variable	Required	Description
ANTHROPIC_API_KEY	No	Enables LLM analyst team
API_FOOTBALL_KEY	No	Live sports event data
EXECUTION_MODE	No	paper (default) or live
BANKROLL_USDC	No	Starting bankroll (default: 500.0)

---

Contributing

We're hiring agent systems engineers, quantitative researchers, and LLM architects. See vequil.com/careers.

---

_{Deterministic execution. Kelly-constrained risk. Every trade auditable.}