Strike account

Nine read-only endpoints. They are the desk's only view of its own money, and the only acceptable one: a position from a brief an hour ago is not a position.

Signing, the API wallet and the strike_request.py helper are in strike-auth.

1. Live state

cd /workspace/strikegrok
python3 scripts/strike_request.py GET /v2/account      # equity, margin usage, maxWithdrawAmount
python3 scripts/strike_request.py GET /v2/balances     # per-asset balances
python3 scripts/strike_request.py GET /v2/positions    # size, entry, leverage, liquidation, uPnL
python3 scripts/strike_request.py GET /v2/openOrders   # resting orders, with the numeric ids cancels need
python3 scripts/strike_request.py GET /v2/portfolio    # account-level portfolio view

Read account, positions and open orders before sizing anything: equity sets the risk budget, positions set what is already at risk, and resting orders are exposure that has not happened yet.

GET /v2/openOrders is the protection check. A position with no matching reduce-only trigger is unprotected - an incident under desk-incident-response, reported the moment it is seen, not a footnote.

Remember order status 5 = untriggered. A resting stop that has not fired lives there, not in 2 open.

2. History

python3 scripts/strike_request.py GET /v2/closedPositions --query symbol=ADA-USD --query limit=25
python3 scripts/strike_request.py GET /v2/history/order  --query symbol=ADA-USD --query limit=50
python3 scripts/strike_request.py GET /v2/history/fill   --query symbol=ADA-USD --query limit=50
python3 scripts/strike_request.py GET /v2/history/funding --query symbol=ADA-USD
python3 scripts/strike_request.py GET /v2/history/transaction

GET /v2/history/funding is what the desk actually paid or received in funding - a real read, not a derivation from the current rate. Use it for the holding-cost line in every review; a rate multiplied by a guess at hours is not a fee figure.

GET /v2/history/transaction types: 1 deposit, 2 withdraw, 3 fee, 4 and up per the spec. Fees appear here, which is how the desk corrects the taker-rate assumption recorded in risk-limits.md against what it was really charged.

Time filters are Unix milliseconds. Page until the window you were asked about is covered, and say how many pages you read. A review that silently stopped at the first page is a wrong review.

3. A single order, authoritatively

python3 scripts/strike_request.py GET /v2/order --query client_order_id=SG-20260910-01-entry
python3 scripts/strike_request.py GET /v2/order --query order_id=1839201122

Either id works. This is the endpoint that makes an unknown send result recoverable: because the desk chose the client_order_id before sending, it can ask the exchange directly whether that order exists rather than inferring from a list. Use it first in any reconciliation.

4. Reconstructing a trade

One order can produce many fills; one position can span many orders.

1. GET /v2/closedPositions for the position and its realised PnL.
2. GET /v2/history/order over the same window for every order that touched the symbol.
3. GET /v2/history/fill for the actual execution prices, sizes and fees.
4. GET /v2/history/funding for what the position cost to hold.
5. Compare against the ticket in /workspace/trading-desk/proposals/SG-*.md, matched by client_order_id.

Report the exchange's numbers. Show the arithmetic where a figure was derived. Where the record is incomplete, say unavailable and name what is missing - never interpolate a fill price.

5. Rules

• Live or nothing. Call the endpoints; never answer account questions from memory or from another Bot's message.
• Every figure carries a UTC time.
• Equity, margin usage and distance to liquidation are the Risk Manager's inputs and are re-read for every ticket.
• Nothing here writes. If a job needs a change, it belongs to the Execution Trader and strike-orders.