Community · E2 · artifact verified

Paper-trade crypto on typed market state

A Next.js dashboard that streams live crypto data, computes fifteen moving averages and eleven oscillators into one typed market state, and lets a Jev agent trade a simulated 100k portfolio - paper only, no broker connected.

01 · Role in the system

What Jev does here

Each trading tick sends the full technical picture as fitted state - oscillators from RSI to Williams %R, moving averages from EMA-10 to SMA-200 with Ichimoku, the open position, cash balance, unrealized PnL, and a configured risk appetite - and asks typed questions in one call: an action Choice among buy, hold, and sell whose criteria are written per risk appetite (aggressive may act on promising setups, conservative requires strong confirmation), a trend-structure Choice comparing price against three moving-average tiers, and further Nouls and scores for the trade desk. Code executes the decision against a local SQLite portfolio with automatic take-profit and stop-loss; the README and the code are explicit that no broker or live order API is connected - every fill is simulated.

The full technical picture as one typed state

Each tick compresses the dashboard's entire analytical weight — fifteen moving averages including Ichimoku, eleven oscillators from RSI to Williams %R, the open position, cash, unrealized PnL, and a configured risk appetite — into one fitted state, and asks in a single call: a buy-hold-sell Choice whose criteria are written differently per risk appetite, a trend-structure Choice against three moving-average tiers, and further Nouls and scores for the desk.

Risk-aware criteria are the subtle design win. Aggressive appetites may act on promising setups while conservative ones require strong confirmation — the trading policy lives in the question text, versioned and readable, rather than buried in prompt evolution. Execution stays deterministic: a local SQLite portfolio with automatic take-profit and stop-loss.

Paper by construction, and honest about signal quality

No broker or live order API is connected — the README and the code say so explicitly, every fill is simulated, and simulated fills are not slippage-accurate. The record also declines to claim an edge: strategy quality is the authored question criteria, not a backtested result.

One-minute timeframes on streamed Yahoo Finance data bound the signal quality from below, which the limitations acknowledge. As the trading cluster's middle record, it shows what the judgment loop looks like with real indicators and real risk vocabulary — before anyone connects real money.

02 · Control boundary

Where Jev sits

Technical indicators into one typed state; buy-hold-sell Choice with risk-aware criteria plus trend and risk Nouls; deterministic execution against a simulated portfolio.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

  • Paper trading only by design: no broker or live order API is connected, and simulated fills are not slippage-accurate.
  • Strategy quality is the authored question criteria, not a backtested edge; no performance claims are made in the README.
  • One-minute timeframes on streamed Yahoo Finance data bound the signal quality.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.