How it uses Jev
For each isolated coin sleeve, the bot sends Jev a market-facing state with book, trades, returns, indicators, asset data, and open-position fields. Jev chooses a long/short bias, leverage rung, and an open/close/hold intent appropriate to whether the position is flat; the bot handles order execution.
What Jev decides
Choose long or short for this asset, given the market, tick timing, position stance, and compact market context.
A representative BTC sleeve market snapshot, including order book, recent trades, indicators, and current position. This question is one of the bounded questions generated for the snapshot.
- long
- short
App workflow
- Read each sleeve’s market data
Collect order-book, trade, indicator, and position state for each configured coin.
- Ask Jev for bounded decisions
Request direction, leverage, and position intent for that sleeve.
- Translate the decision to an order
Apply the chosen intent and direction through the bot’s venue execution code; hold can issue no order.
- Update the live desk
Show the latest call and read position, balance, and PnL from Hyperliquid.
Why it is interesting
The model’s request excludes wallet fills and lifetime PnL, and separates action selection from venue execution. Each sleeve receives its own call, and hold is an explicit option that can result in no order.