Skip to content

README.md

Latest commit

 

History

History
253 lines (190 loc) · 7.39 KB

README.md

File metadata and controls

253 lines (190 loc) · 7.39 KB

YUCLAW

Open Financial Intelligence Platform

MIT License Python 3.10+ PyPI DGX Spark Ethereum SSRN

The open-source financial intelligence OS. Real backtests. ZKP audit trail. Local 120B AI. Like OpenClaw, but strictly built for institutional finance.

Live Dashboard · Academic Paper · Twitter · PyPI

Quick Start

pip install yuclaw
yuclaw today

Output:

YUCLAW Daily Brief — 2026-03-27
==================================================

MARKET: CRISIS (90% confidence)
   Maximum cash
   TLT + GLD only

TOP BUY SIGNALS:
   LUNR   STRONG_BUY   score:+0.650 price:$19.23
   ASTS   STRONG_BUY   score:+0.687 price:$87.00
   DELL   STRONG_BUY   score:+0.674 price:$176.91

TRACK RECORD: Day 5 accuracy 67%

PORTFOLIO ACTION:
   Hold 80%+ cash. Buy only highest conviction signals.
   Max position size: 5% per ticker

Full Command Interface

yuclaw today          # What should I do today? (START HERE)
yuclaw watchlist      # All signals with prices and actions
yuclaw portfolio      # Kelly-optimal allocation for your capital
yuclaw track          # 30-day verified track record
yuclaw ask "..."      # Ask Nemotron 120B any financial question
yuclaw verify LUNR    # Verify signal proof on Ethereum
yuclaw brief          # Latest institutional brief
yuclaw signals        # Raw signal list
yuclaw regime         # Market regime only
yuclaw risk           # Portfolio risk metrics
yuclaw dashboard      # Open live dashboard

Live Dashboard

yuclawlab.github.io/yuclaw-brain

Updates every 30 minutes. Real signals. Real risk. Real model.

Why YUCLAW?

We replace LLM estimates with rigorous, adversarial quantitative validation.

Feature OpenClaw Claude YUCLAW
Real Backtests No No Calmar 3.055
ZKP Audit Trail No No On-chain Sepolia
Real VaR/Kelly No No Historical Sim
Local 120B Model No No Nemotron DGX
Verified Track Record No No 30-day building
Adversarial Red Team No No Kills bad strategies
Dark Pool Engine No No ZKP per match
Governance Token No No YCT live
Academic Paper No No SSRN #6461418

Verifiable Track Record (Day 6 of 30)

Every signal is cryptographically secured on the Ethereum blockchain.

Signal Result Return ZKP Block
LUNR STRONG_BUY CORRECT +14.68% 10515603
ASTS STRONG_BUY CORRECT +10.44% 10515603
MRVL STRONG_BUY CORRECT +6.59% 10515603
DELL STRONG_BUY CORRECT +4.01% 10515603

Verify any signal yourself:

yuclaw verify LUNR
# On-chain: Ethereum Sepolia
# Block: 10515603
# Explorer: https://sepolia.etherscan.io/tx/651aa6b4...

System Architecture

Pipeline Flow

graph TD
    A[Market Data - 104+ Tickers] --> B[Factor Library - RSI MACD Bollinger Momentum]
    B --> C[Signal Aggregator - Adversarial Red Team]
    C --> D[Evidence Graph v2 - 4-Layer Validation]
    D --> E[Nemotron 3 Super 120B - Local 19 tok/s Zero cloud]
    E --> F[ZKP Proof - Ethereum Sepolia]
    F --> G[Execution Manager - Levels 0-4]
    G --> H[Institutional Brief - Dashboard - Track Record]
Loading

Directory Structure

yuclaw/
  modules/       Signal aggregator, macro regime detection
  factors/       Factor library — RSI, MACD, Bollinger
  risk/          VaR, CVaR, Kelly criterion
  brain/         Evidence Graph v2, financial NER
  memory/        Portfolio memory, pattern learning
  validation/    Validation Studio — 5-stage Red Team
  darkpool/      Dark Pool matching engine + ZKP
  marketplace/   FinSkills Marketplace
  yct_token_pkg/ YCT governance token
  edge/          FIX gateway, execution levels 0-4
  trust/         ZKP vault, on-chain proofs
  openclaw/      OpenClaw skill + MCP server
  api/           REST API — port 8000

Hardware & Operations

Running 24/7 on NVIDIA DGX Spark GB10 — zero cloud dependency.

Engine Frequency Status
Signal loop — 104 tickers Every 30 min LIVE
Factor engine — RSI/MACD/Bollinger Every 30 min LIVE
Risk engine — VaR/CVaR/Kelly Every 1 hr LIVE
Macro regime — CRISIS detection Every 1 hr LIVE
Nemotron brief — 120B analysis Every 15 min LIVE
ZKP vault — Ethereum proofs Every 1 hr LIVE
Dark Pool — order matching Every 30 min LIVE
Track record — verified performance Every 1 hr LIVE
Portfolio memory — pattern learning Every 1 hr LIVE

Hardware specs:

  • Model: Nemotron 3 Super 120B (Q4_K_M, 81GB)
  • GPU: Grace Blackwell GB10
  • VRAM: 81GB on GPU (89/89 layers)
  • Speed: 18.9 tok/s continuous
  • Uptime: 96+ hours

OpenClaw Integration

YUCLAW works as an OpenClaw skill and MCP server.

# Install as OpenClaw skill
bash <(curl -s https://raw.githubusercontent.com/YuClawLab/yuclaw-brain/main/yuclaw/openclaw/install.sh)

# Then in any chat:
/yuclaw today
/yuclaw signals
/yuclaw backtest NVDA
/yuclaw verify LUNR

Or as MCP server:

python3 yuclaw/openclaw/mcp_server.py
# Add http://localhost:8002 to OpenClaw config

Dark Pool Engine

Private order matching with ZKP proof per match.

from yuclaw.darkpool import DarkPoolEngine

engine = DarkPoolEngine()
matches = engine.run_from_signals(signals)
# Every match gets a ZKP proof
# Midpoint execution — no market impact

YCT Governance Token

Earn YCT by contributing to YUCLAW ecosystem.

Action YCT Reward
Correct signal 10 YCT
Strategy approved 100 YCT
Governance vote 1 YCT
Skill installed 10x 50 YCT

Total supply: 100,000,000 YCT

Academic Foundation

CRT Lock-Free Concurrent Scheduler for Financial Systems SSRN Abstract #6461418 | DGX Spark GB10 | 1.37ms latency

Read the full paper on SSRN

Community & Contributing
Dashboard yuclawlab.github.io/yuclaw-brain
Twitter @Vincenzhang2026
GitHub YuClawLab
PyPI pypi.org/project/yuclaw
Paper SSRN #6461418

Released under the MIT License — free for everyone.

Built on NVIDIA DGX Spark GB10 - Nemotron 3 Super 120B - Zero cloud dependency

pip install yuclaw