XRPL in Time
Conversational AI assistant for XRPL AMM liquidity management. Analyze portfolio risk, generate quantitative trading strategies with visual PnL profiles.
Demo Video
Project Information
At a Glance
Conversational AI assistant for XRPL AMM liquidity management. Analyze portfolio risk, generate quantitative trading strategies with visual PnL profiles.
Description
XRPL in Time turns complex DeFi liquidity management into a simple conversation. Users describe their goal in natural language and the system classifies intent locally on-device, fetches their XRPL positions, and computes quantitative risk metrics including impermanent loss, delta exposure, Value at Risk, and Sharpe ratio.
Claude Sonnet then generates three risk-ranked strategies with PnL visualizations and break-even analysis. The user reviews, selects, and signs a single transaction via Xaman or Crossmark. No private keys are ever held by the backend, and slippage is hard-capped at 1% at the smart contract level.
The architecture is fully documented with risk models, LLM prompts, and smart contract interfaces defined.
Technical Details
The system is built across six modules. The frontend runs on Next.js with TypeScript, Tailwind CSS, and Recharts. Intent routing uses a local Llama 3.2 3B model via Ollama, delivering sub-100ms classification with no data leaving the device.
A Python FastAPI backend orchestrates XRPL data fetching and delegates strategy generation to Claude Sonnet via the Anthropic API with structured JSON output. Quantitative risk models (IL, VaR, Sharpe, delta) run in a dedicated Rust/Axum service. On-chain execution is handled by a Bedrock smart contract in Rust/WASM targeting the XRPL native AMM.
XRPL ledger data is streamed and indexed by a Go-based firehose module. An optional MCP server exposes VEGA tools to Claude Desktop and compatible AI clients.
Team
3Alex Havlin
Paul Quesnot
Alejandro Meredith Romero
Hackathon
Stablecoins & Payments Hackathon
Duration
Mar 21, 8:00 AM - Mar 22, 4:00 PM UTC