Monitoring

The repository ships no metrics server or alerting stack. Monitoring is log- and state-based; this page lists the concrete signals.

#Keeper logs

The keeper logs each pipeline stage per fill: observed → session decision → risk decision (Decision, allowed size, reason) → execution result (tx hash, dry-run calldata, or skip). Grep for:

• REJECT / reason codes (ZERO_NOTIONAL, SESSION_CLOSED, cap names, DRAWDOWN_HALT) — expected skips.
• dry-run — live mode without SEAT_SUBMIT_TX; calldata is logged for inspection.
• Executor errors — RPC failures, revert reasons.

#Fill tape

keeper/data/fills.json (gitignored, overwritten per run) records every processed fill with an honest source label: fixture or chain. A tape that stops growing while the leader is active means the fill source is stale or the RPC is down.

#On-chain health checks

# NAV per share — should move only with P&L and fees
cast call $DESK "navPerShare()(uint256)" --rpc-url $RPC

# high-water mark vs NAV — approaching from below after losses is normal;
# a drawdown halt trips at 20% below HWM (default)
cast call $DESK "highWaterMark()(uint256)" --rpc-url $RPC

# withdrawal queue depth — growth means redemptions are outpacing cash
cast call $DESK "withdrawalQueueLength()(uint256)" --rpc-url $RPC

# pause flag
cast call $DESK "paused()(bool)" --rpc-url $RPC

(Function names per the contract sources; see DeskVault.)

#App-level signals

The app (@seat/app) renders the fill tape and vault state from the same chain reads. If the app and cast disagree, trust cast and investigate the app's RPC (NEXT_PUBLIC_*_RPC_URL).

#What "healthy" looks like