Back to Projects
AIQA AutomationMulti-chainPlaywright
Levana Test Assistant
AI-powered QA copilot for multi-chain perpetual trading validation
The Challenge
Levana's multi-chain architecture introduced significant testing complexity. Identical UI flows behaved differently depending on chain, wallet, and environment (testnet vs mainnet). Wallet integrations (Keplr, Leap, Metamask, WalletConnect) required precise sequencing and conditional logic. Trading logic involved dynamic pricing, slippage thresholds, liquidation risks, and real-time state changes. Traditional QA documentation could not keep pace with protocol evolution, network differences, and weekly product changes.
Approach & Solution
- Designed the Levana Test Assistant as a domain-aware QA copilot, not a generic chatbot
- Built structured test case generation aligned with Levana's actual UI, wallet flows, and trading mechanics (connect wallet, open/close positions, limit orders, error states)
- Implemented chain-aware and wallet-aware logic where test cases dynamically adjusted based on network (Osmosis, Injective, Neutron), supported wallets per chain, and testnet vs mainnet behavior differences
- Added selector-level precision — test cases referenced concrete UI elements and HTML selectors to reduce ambiguity and automation failures
- Structured output to feed directly into Playwright-based automation, minimizing rework between manual QA and automated execution
- Built explicit handling for edge-cases: price deviation warnings, wallet rejection and disconnects, insufficient balance and margin errors, network latency and state desynchronization
Results
01Standardized test case quality across complex multi-chain trading flows
02Reduced time required to design and update test coverage as protocols evolved
03Improved confidence in wallet connection, trading execution, and error handling flows
04Enabled faster transition from manual validation to reliable automation
05Lowered regression risk in high-impact trading features without slowing delivery